by user

Category: Documents





© RTO/NATO 2001
Single copies of this publication or of a part of it may be made for individual use only. The approval of the
RTA Information Policy Executive is required for more than one copy to be made or an extract included
in another publication. Requests to do so should be sent to the address above.
What Is Essential for Virtual Reality
Systems to Meet Military Human
Performance Goals?
(les Caractéristiques essentielles des systèmes VR pour
atteindre les objectifs militaires en matière de performances
Papers presented at the RTO Human Factors and Medicine Panel (HFM) Workshop held in
The Hague, The Netherlands, 13-15 April 2000.
Published March 2001
Distribution and Availability on Back Cover
This page has been deliberately left blank
Page intentionnellement blanche
What Is Essential for Virtual Reality Systems to
Meet Military Human Performance Goals?
(les Caractéristiques essentielles des systèmes VR pour atteindre les objectifs
militaires en matière de performances humaines)
Papers presented at the RTO Human Factors and Medicine Panel (HFM) Workshop held in
The Hague, The Netherlands, 13-15 April 2000.
The Research and Technology
Organization (RTO) of NATO
RTO is the single focus in NATO for Defence Research and Technology activities. Its mission is to conduct and promote
cooperative research and information exchange. The objective is to support the development and effective use of national
defence research and technology and to meet the military needs of the Alliance, to maintain a technological lead, and to
provide advice to NATO and national decision makers. The RTO performs its mission with the support of an extensive
network of national experts. It also ensures effective coordination with other NATO bodies involved in R&T activities.
RTO reports both to the Military Committee of NATO and to the Conference of National Armament Directors. It comprises a
Research and Technology Board (RTB) as the highest level of national representation and the Research and Technology
Agency (RTA), a dedicated staff with its headquarters in Neuilly, near Paris, France. In order to facilitate contacts with the
military users and other NATO activities, a small part of the RTA staff is located in NATO Headquarters in Brussels. The
Brussels staff also coordinates RTO’s cooperation with nations in Middle and Eastern Europe, to which RTO attaches
particular importance especially as working together in the field of research is one of the more promising areas of initial
The total spectrum of R&T activities is covered by the following 7 bodies:
Applied Vehicle Technology Panel
Human Factors and Medicine Panel
Information Systems Technology Panel
• NMSG NATO Modelling and Simulation Group
Studies, Analysis and Simulation Panel
Systems Concepts and Integration Panel
Sensors and Electronics Technology Panel
These bodies are made up of national representatives as well as generally recognised ‘world class’ scientists. They also
provide a communication link to military users and other NATO bodies. RTO’s scientific and technological work is carried
out by Technical Teams, created for specific activities and with a specific duration. Such Technical Teams can organise
workshops, symposia, field trials, lecture series and training courses. An important function of these Technical Teams is to
ensure the continuity of the expert networks.
RTO builds upon earlier cooperation in defence research and technology as set-up under the Advisory Group for Aerospace
Research and Development (AGARD) and the Defence Research Group (DRG). AGARD and the DRG share common roots
in that they were both established at the initiative of Dr Theodore von Kármán, a leading aerospace scientist, who early on
recognised the importance of scientific support for the Allied Armed Forces. RTO is capitalising on these common roots in
order to provide the Alliance and the NATO nations with a strong scientific and technological basis that will guarantee a
solid base for the future.
The content of this publication has been reproduced
directly from material supplied by RTO or the authors.
Published March 2001
Copyright  RTO/NATO 2001
All Rights Reserved
ISBN 92-837-1057-6
Printed by St. Joseph Ottawa/Hull
(A St. Joseph Corporation Company)
45 Sacré-Cœur Blvd., Hull (Québec), Canada J8X 1C6
What Is Essential for Virtual Reality Systems to Meet
Military Human Performance Goals?
(RTO MP-058 / HFM-058)
Executive Summary
The purpose of the workshop was to:
• identify the functional requirements of potential military applications of Virtual Reality (VR)
• report the state-of-the-art and projected capabilities of VR technologies, and
• propose future research requirements and directions for military applications.
The workshop was organised into three daylong sessions. The first day focused on functional
requirements for military VR applications in the domains of training, robotics, remote operations and
command and control. On the second day, we examined available VR techniques now and in the near
future. Presentations discussed visual, haptic, auditory and motion feedback, navigation interfaces, and
scenario generation, modelling software and rendering hardware. The third day addressed missing VR
capability and future research and concluded with a panel discussion.
During the workshop discussions forty participants from military organisations, academia and industry
put forward their opinions on the biggest bottlenecks and opportunities in the development of military
VR applications.
Virtual Reality technology is of great interest to the military. Its most important application domain is
training. VR for training can reduce cost and risk of casualties and improve flexibility and
performance monitoring. Furthermore, great opportunities are identified in the domains of planning
and mission rehearsal, simulation supported operation, remotely operated systems and product design.
At the same time a number of factors seem to frustrate successful applications in this field. One of the
significant bottlenecks is that VR developments are usually not user driven. Application developers
and designers do not pay enough attention to human factors requirements. Consequently, applications
may fail because of a lack of natural interfaces and motion sickness. So far, user interfaces have been
poorly attuned to natural human skills (crude input devices and inconsistent visual, auditory and
proprioceptive feedback) and to the tasks to be performed in VR. A second bottleneck is the lack of
standardisation causing problems with integrating VR systems and VR software tools. A third is the
lack of behavioural models of people and objects in VR scenarios and facilities for team interactions
(poor visual human representations and communication tools).
In general, better co-ordination between military organisations, industry and academia is essential in
order to identify gaps in current knowledge and co-ordinate research. To this purpose the military
should develop a vision on the use of VR technology and specify their needs more clearly. Industry
should work on standardisation and should substantially implement human factors into their
development process. Academia and research institutes should co-ordinate and accelerate their longterm research efforts to focus on natural interfaces (innovative metaphors) and on how to model
(intelligent) human and object behaviour. In the short term academia should focus on human factors
metrics and metrics for team performance (cognition, communication), and a standard evaluation
A specific suggestion made during the workshop that could contribute to solving the bottlenecks is to
establish a RTO Task Group to (1) identify applications with a high return of investment, user
requirements and technologies for investment by the military and (2) foster development of natural VR
interfaces and behaviourally realistic intelligent agents and models (identify new funding sources).
The enthusiasm of the workshop attendees and the evident willingness to share ideas and to discuss
their findings provide a promising base for a co-operation between military agencies, industry and
academia. Research on the usability of VR technology will enable militaries to be smart buyers. It will
ensure that Virtual Reality hardware and software is capable of meeting the perceptual, fidelity,
transfer of training, and health and safety requirements of applications.
les Caractéristiques essentielles des systèmes VR
pour atteindre les objectifs militaires en matière
de performances humaines
(RTO MP-058 / HFM-058)
L’atelier avait pour objet :
• d’identifier les besoins fonctionnels découlant des applications militaires possibles des technologies
de réalité virtuelle (VR),
• de rendre compte de l’état actuel des connaissances et des capacités anticipées dans ce domaine, et
• de proposer de futurs sujets de recherche et des orientations vers des applications militaires.
L’atelier a été organisé en trois sessions d’une journée : La première journée a été consacrée aux
besoins fonctionnels découlant des applications militaires des technologies VR dans les domaines de
l’entraı̂nement, la robotique, les opérations à distance et le contrôle. Le deuxième jour, nous avons
examiné les techniques VR actuelles et émergentes. Des présentations ont été données sur le bouclage
de l’information dans les domaines visuels, haptiques, auditifs, et cybernétiques, les interfaces de
navigation, la génération de scénarios, les logiciels de modélisation et le matériel de rendu d’image. La
troisième journée a été centrée sur les capacités faisant défaut dans le domaine de la VR, ainsi que les
travaux de recherche futurs, et s’est terminée par une discussion entre les membres de la commission.
Au cours des discussions qui ont eu lieu pendant les trois jours de l’atelier, une quarantaine de
participants venus d’organisations militaires, d’universités et de l’industrie ont exprimé leurs opinions
sur les impasses les plus importantes, ainsi que sur les opportunités offertes de développer de nouvelles
applications VR militaires.
Les technologies de réalité virtuelle sont d’un grand intérêt pour les militaires. Le domaine
d’application le plus important est celui de l’entraı̂nement. L’emploi de techniques VR pour
l’entraı̂nement permettrait de réduire son coût, ainsi que le risque d’accidents corporels, et pourrait
apporter des améliorations au niveau de la flexibilité et du contrôle des performances. En outre, de
grandes possibilités ont déjà été identifiées dans les domaines de la planification et la préparation des
missions, de la conduite des opérations à l’aide de la simulation, de la télécommande des systèmes et
de la conception des produits.
En même temps, un certain nombre de facteurs sembleraient entraver la réussite des applications dans
ce domaine. Le fait que les développements en matière de VR soient rarement orientés par les
utilisateurs représente l’une des principales gênes. Les développeurs d’applications et les concepteurs
ne tiennent pas suffisamment compte des besoins du point de vue des facteurs humains. Par
conséquent, les applications risquent d’échouer du fait du mal des transports et du manque d’interfaces
naturelles. Jusqu’à présent, les interfaces utilisateurs ont été mal adaptées aux capacités humaines
naturelles (des unités d’entrée rustiques et des boucles d’information visuelles, auditives et
proprioceptives incompatibles) ainsi qu’aux tâches à accomplir en VR. Le manque de normalisation,
qui crée des problèmes d’intégration des systèmes et des outils VR représente une deuxième gêne
importante. Enfin, le manque de modèles du comportement humain et d’objets dans les scénarios VR,
ainsi que le manque de possibilités d’interactions interéquipes (représentations visuelles du corps
humain et outils de communication de mauvaise qualité) est la troisième gêne identifiée.
De façon générale, il est indispensable d’assurer une meilleure coordination entre les organisations
militaires, l’industrie et les universités, afin d’identifier les éventuelles lacunes dans les connaissances
et de coordonner les travaux de recherche. Avec cet objectif en vue, les militaires devraient élaborer
une philosophie de mise en oeuvre des technologies VR et exprimer leurs besoins plus clairement.
L’industrie devrait travailler sur la normalisation et faire une large place aux facteurs humains dans
leurs processus de développement. Les universités et les instituts de recherche devrait coordonner et
intensifier leurs efforts de recherche à long terme afin de se concentrer sur les interfaces naturelles
(métaphores novatrices) et sur la modélisation (intelligente) du comportement des objets et des êtres
humains. A court terme, les universitaires devraient privilégier la métrologie des facteurs humains et la
métrologie du travail en équipe (l’approche cognitive, la communication), ainsi que l’élaboration d’une
nouvelle méthodologie normalisée d’évaluation.
L’une des propositions faites au cours de l’atelier, qui pourrait contribuer à l’élimination de impasses,
consisterait à créer un groupe de travail RTO pour (1) identifier des applications ayant un bon
rendement, les besoins des utilisateurs et les technologies méritant des efforts d’investissement de la
part des militaires, et (2) encourager le développement d’interfaces VR naturelles, ainsi que des agents
et des modèles intelligents ayant des comportements réalistes (identification de nouveaux bailleurs de
L’enthousiasme manifesté par les participants durant l’atelier, ainsi que leur volonté évidente de
partager leurs idées et de discuter de leurs conclusions a constitué une base prometteuse pour une
coopération future entre les agences militaires, l’industrie et les universités. Des recherches doivent
être entreprises sur la facilité d’utilisation de ces technologies afin de permettre aux militaires de les
acheter en connaissance de cause. Ils pourraient ainsi s’assurer que le matériel et les logiciels de réalité
virtuelle seraient compatibles avec les exigences de perception, de fidélité, de transfert d’entraı̂nement
et d’hygiène et sécurité demandées pour les applications.
Click inside the blue boxes or on the titles to view the corresponding section
(except for items marked in red, which were not available at the time of production)
Executive Summary
Human Factors and Medicine Panel
Technical Evaluation Report
by P. Werkhoven and R. Breaux
Keynote Address: What is Essential for Virtual Reality Systems to Meet Military Human
Performance Goals?
by R.S. Kalawsky
A Virtual Environment for Naval Flight Deck Operations Training
by V.S.S. Sastry, J. Steel and E.A. Trott
Mission Debriefing System
by B.I. Johansen and B. Fredborg
Mine Clearance in a Virtual Environment
by L. Todeschini, T. Pasquier, P. Hue and P. Gorzerino
Acquiring Real World Spatial Skills in a Virtual World
by B.G. Witmer, B.W. Knerr and W.J. Sadowski Jr.
Advanced Air Defence Training Simulation System (AADTSS). Virtual Reality is Reality in
German Airforce Training
by M. Reichert
“What is Essential for Virtual Reality to Meet Military Performance Goals?” Performance
Measurement in VR
by J. Patrey, R. Breaux, A. Mead and E. Sheldon
Appropriate Use of Virtual Environments to Minimise Motion Sickness
by W. Bles and A.H. Wertheim
Human Computer Interactions in Shared VE
by B. Loftin
Available Virtual Reality Techniques Now and in the Near Future
by G.C. Burdea
Simulating Haptic Information with Haptic Illusions in Virtual Environments
by A. Lécuyer, S. Coquillart and P. Coiffet
Tactile Displays in Virtual Environments
by J.B.F. van Erp
† Paper not available at time of printing.
Virtual Cockpit Simulation for Pilot Training
by K-U. Dörr, J. Schiefele and W. Kubbat
Ergonomic Investigations for Virtual Environments
by C. Meyer
UAV Operations using Virtual Environments
by J.B.F. van Erp and L. van Breda
Productive Application of Virtual Environments
by A. Roessler
The Dangerous Virtual Building, an Example of the Use of Virtual Reality for Training in
Safety Procedures
by M. Lozano, M. Fernandez, J. Casillas, J. Fernández and C. Romero
Visualisation of Geographic Data in Virtual Environments
by T. Alexander
Acquiring Distance Knowledge in Virtual Environments
by E. Heineken and F.P. Schulte
Development of Virtual Auditory Interfaces
by R.D. Shilling and T. Letowski
Educational Conditions for Successful Training with Virtual Reality Technologies
by A. von Baeyer and H. Sommer
Entertainment Technology and Military Virtual Environments
by M.R. Macedonia and P. Rosenbloom
† Paper not available at time of printing.
Human Factors and Medicine Panel
Director, Centre for Human Sciences
F138 Building - Room 204
Farnborough, Hants GU14 0LX
United Kingdom
Tel: 44 1252 393 764
Fax: 44 1252 393 982
Email: [email protected]
P.O. Box 20703
Binckhorstlaan, 135
2500 ES The Hague
The Netherlands
Tel.: 31 70 339 6403
Fax: 31 70 339 7439
Email: [email protected]
Workshop Chairman
TNO Human Factors Research Institute
Dept. of Work Environment
Kampweg 5
3769 ZG Soesterberg, The Netherlands
Tel.: +31 3463 56283 Fax: +31 3463 53977
Email: [email protected]
Neuenahrer Str. 20
53343 Wachtberg-Werthhoven, Germany
Tel.: +49 (0)228 9435 480
Fax: +49 (0)228 9435 508
Email: [email protected]
Pascal HUE
49460 Montreuil Juigne, France
Tel.: +33 241936644
Fax: +33 241936704
Email: [email protected]
Dr. Robert BREAUX
12350 Research Parkway
Orlando, Florida 32826
United States of America
Tel.: +1 407 380 8168
Fax: +1 407 380 4007
Email: [email protected]
Dr. Martin G. KAYE
DERA Centre for Human Sciences
Room 2012, A5 Building
Farnborough, Hampshire, GU14 0LX
United Kingdom
Tel.: +44 1252 393610
Fax: +44 1252 394700
Email: [email protected]
Dr. Stephen L. GOLDBERG
US Army Research Institute
12350 Research Parkway
Orlando, Florida 32826-3276
United States of America
Tel.: +1 407 384 3980
Fax: +1 407 384 3999
Email: [email protected]
Norwegian Defence Research Establishment
P.O.Box 25
N-2007 Kjeller
Tel.: +47 63 80 78 52
Fax: +47 63 80 78 11
Email: [email protected]
continued overleaf
Edison, 4
28006 Madrid
Tel.: +34 1 4115011
Fax: +34 1 4114703
Email: [email protected]
7, rue Roger
Tel.: +33 141087317
Fax: /
email: [email protected]
Danish Defence Research Establishment
Svanemollens Kasserne, Ryvangs Allé 1
DK-2100 Copenhagen OE
Tel.: +45 39 15 18 05
Fax: +45 39 29 15 33
Email: [email protected]
Department of Informatics and Simulation
Royal Military College of Science
Shrivenham, Swindon SN6 8LA
United Kingdom
Tel.: +44 1793 785652
Fax: +44 1793 782753
Email: [email protected]
BP 25 - 7, Rue Ancelle
92201 Neuilly-sur-Seine, France
Tel: +33 1 55 61 22 60
Fax: +33 1 55 61 22 98
Email: [email protected] or [email protected]
“What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”
The Hague, TNO-FEL, 13 – 15 April 2000
Peter Werkhoven
TNO Human Factors
Dept. of Information Processing
PO Box 23
3769 ZG Soesterberg, The Netherlands
[email protected]
Robert Breaux
Naval Air Warfare Centre
Training Systems Division
12350 Research Parkway
Orlando, FL 32826-3275, USA
[email protected]
1.1 Background (history HFM 021, previous activities)
NATO Research Study Group HFM-21 (called RSG-28 before the RTA reorganisation)
was established to explore and evaluate human factors issues that effect the use of virtual
reality technologies for military purposes. The findings of the group are to provide
NATO countries with better understanding of the capabilities and limitations of this new
and sometimes over-hyped technology. The study group has agreed upon the following
definition of virtual reality to establish a common reference point.
Virtual Reality is the experience of being in a synthetic environment and the
perceiving and interacting through sensors and effectors, actively and passively,
with it and the objects in it, as if they were real. Virtual Reality technology allows
the user to perceive and experience sensory contact and interact dynamically with
such contact in any or all modalities.
Virtual reality has great potential in areas such as training, mission rehearsal, concept
development, weapon prototyping, and personnel selection. Many virtual reality
technologies also have application in robotics and remote manipulation applications.
There have been a number of successful research and prototype applications of virtual
reality. They have mostly been in training. Use of virtual reality for ship-handling
training has been successfully demonstrated in both the United States and Canada.
Dismounted soldier simulation has seen considerable research and development activity.
Virtual reality is a relatively new concept and many of the technologies involved in
immersing individuals or teams in virtual environments are evolving and improving
The key to the effectiveness of virtual reality for military purpose is the man-machine
interface or human-computer interaction. Military personnel must be able to perform their
tasks and missions using virtual reality sensory display devices and response devices.
These devices must display an environment that provides the appropriate cues and
responses needed to learn and perform military tasks. Human factors issues include:
determining the perceptual capabilities and limitations of sensory display devices;
designing terrain data bases and other displays to meet task performance needs;
understanding the human and task performance compromises required by current
technologies; evaluating transfer of training and knowledge from the virtual to the real
world; and considering the causes and solutions to simulator sickness that can occur in
virtual reality. The Research Study Group intends to provide information and
recommendations on these issues to military researchers, requirements generators, and
acquisition agencies. The intended benefit is better-informed decisions on application of
virtual reality technologies to meet appropriate military needs.
Previous activities included:
• A one-day workshop titled “The development of a Generic Battery of Human
Performance Metrics for Virtual Environments” (Chertsey, UK, October 14, 1996)
• A three-day workshop titled “The capability of Virtual Reality to meet military
requirements” (Orlando FL, December 1997).
Reviews of these workshops together with a chapter on “Human Computer Interaction
issues in VR” and a chapter on the “State of the art in VR research in NATO countries”
will be published in the FINAL report of HFM-021.
The current workshop is the last activity organised by HFM-02:
• A three-day workshop titled “What is essential for Virtual Reality systems to meet
military human performance goals” (The Hague NL, April 13 – 15, 2000).
1.2 Purpose and scope of the workshop
The focus of the current workshop is on:
• the functional requirements of potential virtual reality military applications,
• the state-of-the-art and projected capabilities of virtual reality technologies, and
• future research requirements and directions for military applications.
In the light of this focus the following military application domains were considered:
• training;
• robotics;
• remote operations;
• command and control.
Within each of these domains VR requirements, capabilities and R&D issues were
considered with respect to the following aspects:
• visual, haptic, auditory and motion feedback;
• navigation interfaces;
scenario generation;
modelling software and rendering hardware.
1.3 Program workshop
The workshop took place over three days with the following structure:
Thursday, 13 April 2000
key-note speaker
Pascal Hue (FR) & Thomas Alexander (GE)
Functional requirements for military VR applications
Prof. Dr. Roy Kalawsky (Advanced Virtual Reality Centre,
Loughborough University, UK)
8 speakers:
• A Virtual Environment for Naval Flight Deck Operations Training (Dr
V.V.S.S. Sastry, UK);
• Debriefing for Pilots in VR (Major B. I. Johanson, DE);
• Probing in VR (Mr. L. Todeschini, FR);
• Acquiring Real Worlds Special Skills in VR (Dr B. G. Witmer, USA);
• Training System STINGER Simulator (Dipl.-Ing. M. Reichert, GE);
• Performance Measurements in VR (Dr J. Patrey, USA);
• Appropriate Use of VR to Minimise Motion Sickness (Dr W. Bles, NL);
• Human Computer Interactions in Shared VE (Prof. Dr B. Loftin, USA).
Friday, 14 April 2000
keynote speaker:
Elizabeth Henderson (UK) & Lisbeth Rasmussen (DE)
Available VR techniques now and in the near future
Prof. Dr Grigore Burdea (Human-Machine Interface
Laboratory, Rutgers University, Piscataway, NJ, USA)
8 speakers:
• Simulating Haptic Information with Haptic Illusions in VR (Mr. A. Lecuyer,
• Tactile Displays (Dr J. van Erp, NL);
• Virtual Cockpit Simulation for Pilot Training (Dipl.-Ing. K.-U. Doerr, GE);
• Ergonomic Investigations for VR (Dr C. Meyer, GE);
• UAV Operations Using VR (Dr Ing. L. van Breda, NL);
• Productive Application of VR (Dr A. Roessler, GE);
• The Dangerous Virtual Building, an Example of the Use of VR for Training in
Safety Procedures (Dr M. Lozano, SP);
• Vizualization of Geographic Data in VR (Dipl.-Ing. T. Alexander, GE).
Saturday, 15 April 2000
Trond Myhrer (NO) & Steve Goldberg (USA)
Missing VR capability and future research
4 speakers:
• Influence on the representation of Spatial Information Acquired in Virtual
Environments (Prof. Dr E. Heineken, NE);
Development of Virtual Auditory Interface (LCDR Dr R. D. Shilling, USA);
Educational Conditions for Successful Training with Virtual Reality
Technologies (Dr A. von Baeyer, GE);
• Entertainment Technology and VR (Dr M. R. Macedonia, USA).
Panel discussion (HFM021 panel, audience participation).
1.4 Attendees
Attendees (total 43) of the workshop had various nationalities and backgrounds:
Total #
United Kingdom
civil res. inst.
2.1 Introduction
Functional requirements
In his keynote lecture Prof. Roy Kalawsky (Advanced Virtual Reality Centre,
Loughborough University) provided a ‘steppingstone’ for the session on functional
requirements. He pointed out that the two most crucial characteristics of Virtual Reality
(VR) are the experience of ‘being’ in the simulated world (immersion) and the interaction
through sensors (acting). Therefore VR is an essentially ‘man-in-the-loop’ simulation. He
pointed out that VR, is not new. The first notions of VR date back to 1956 (Stanton). In
fact, a functional decomposition of existing VR systems shows a continuum of levels of
immersion starting with non-immersive systems such (e.g. wearable or desktop displays,
‘joystick driven’, with low-level interactions) to fully immersive systems (e.g. headslaved displays, natural interactions, haptic feedback, etc.). As a simulation tool, the
perceived military benefits of VR are mostly related to system effectiveness and not to
weapon effectiveness. (not sure what this last sentence means)
The most important message brought by Kalawsky is that the human factor should be
central in application development:
functional requirements (image quality, display scene motion, content development)
must be driven by the end application (based on task analysis);
• military requirements must be performance driven in which human capability is a
major factor; and
• applications must be evaluated thoroughly with respect to human task performance.
Note that human task performance is governed by the environment itself, personal
capabilities, individual motivation as well as the situation under which the task is carried
An approach involving this sort of task analysis was illustrated by Reichert (Training
system Stinger simulator). Interestingly, TNO developed and spoke about a similar
application of VR to training STINGER operators at the previous RSG-28 workshop in
Orlando. Based on defining of training goals and tasks the functional requirements of the
simulator can be specified (scenario’s, interactions, minimum visual resolution, etc.).
Although Stinger simulators have been built and used for training, systematic
measurements of user performance and transfer of training have not yet been carried out.
Sastry presented a study on the use of Virtual Environments for helicopter deck landing
training in which training transfer was explicitly measured. Preliminary conclusions show
that immersive VE can be used to train visual motor skills. For training procedural
knowledge, simpler training devices can be used, although it may be more cost-effective
to have a single system for both types of training. Todeschini (Probing in VE) also
measured transfer but in the context of VR training for mine clearance tasks. He
demonstrated the importance of force-feedback in such applications.
Underlying tasks such as Stinger launching, flight deck operations and mine probing are
more basic human skills and abilities. Military personnel need to be able to navigate,
orient themselves and interact with the Virtual Environment. The second day of the
workshop brought together human factors researchers working in these areas. Witmer
presented research on how well people can learn their way around in a virtual world.
More specifically: how can we support the learning of routes and configurations VR by
adding visual and aural cues in VR? This research yielded concrete guidelines for
designing a VR in which task performance depends substantially on spatial orientation
and navigation (see Witmer: Acquiring real world skills in VR). Bles (Appropriate use of
VE to minimise motion sickness) presented work on modelling the functioning of human
equilibrium sensors (vestibular system) and showed which types of motion do and do not
cause motion sickness.
Another basic functional requirement, particularly in multi-user applications, is the role
of (non-verbal) human communication (human representations and behaviour). One
extremely promising development (mentioned by Kalawsky) is the use of avatars
(synthetic visual human representations) in Virtual Environments. Avatars can be driven
by instrumented humans immersed in the VE or computer generated with Artificial
Intelligence programs driving their behaviour. Loftin (Human Computer Interactions in
shared VE) presented working which software agents drive human representations
(avatars) in distributed shared VRs aimed at training soldiers in peace-keeping
operations. A major benefits of avatar-populated VRs is the reduction in the need for
having all the players in a scenario be represented by human being(not everyone needs to
be in the loop) which gives the flexibility of just-in-time training.
Important research questions are the required fidelity of the physical appearance of
avatars and how to model, generate and validate useful behaviour? Can avatars really be
surrogate team-members or coaches? When should they be reactive, when pro-active?
When should they behave rule-based, when stochastic?
Besides human factors functional requirements VR design is driven by operational
requirements (e.g. mobility, weight, flexibility) and economic requirements (e.g. cost and
return on investment). These issues were addressed by Johanson who demonstrated the
potential of a mission debriefing system for the Danish Airforce.
Designing VRs that integrate task analysis, functional requirements and technology
concessions can be an iterative process that is time consuming. New approaches are
needed. Patrey (Performance measurements in VR) presented an alternative approach to
traditional cognitive task analysis (rapid interactive design) in order to speed up
application developments (see also Loftin). This raised the question of how to interpret
performance measurements to adjust system parameters without explicitly modelling the
task: should we take a ‘neural net’ approach? (I don’t know what this sentence means)
Available techniques
The state of the art of VR technology was presented by Burdea in his keynote
presentation (Available VR now and in the near future). He gave an excellent overview
of developments in:
• computing engines (e.g. Intergraph Pentium III based system nowadays match the
computing power of SGI Infinite Reality systems);
• tracking devices (e.g. inertial/ultrasonic trackers);
• personal displays (e.g. light weight, high resolution HMO's);
• large volume displays (e.g. CAVE); and
• haptic displays (e.g. haptic gloves, haptic floors).
Burdea observed that:
• VR technologies are getting cheaper (displays, sensors, engines). This brings VR
research within reach of research organisations with low budgets for capital
• consequently, we see a stronger involvement of experimental psychology in
identifying the limitations of human performance in VR applications and in validation
The lower costs of VR technology has l allowed for the development of VR applications,
for example Doerr’s slow-cost virtual cockpit. Also, 3D worktable technologies are
allowing battlefield information to be presented realistically(Alexander). This is an
example of how command and control tasks can be supported with VR-tools.
By making use of human information processing capabilities (e.g. illusions in multimodal
perception), perceptions can be created without the need of high-tech display devices. For
example, LeCuyer (Simulating haptic Information with haptic illusions in VE) showed
that the perceived haptic amplitude of a spring is substantially affected by the visual
representation of the spring and vice versa. It should be noted that a qualitative use of
such illusions is feasible but a quantitative use requires individual calibrations. In fact,
Roessler (Productive application of VR) stated that a 6D (six degrees of freedom) user
interface was even closer to reality without forces than with (for the application
VR can be used to transfer information through sensory modalities that are not normally
used for that purpose. An example was shown by Van Erp (Tactile Displays) who used
the skin to sense spatial direction (which is usually sensed by our eyes or ears). By using
arrays of tactile micro-vibrators on the skin the position or direction of objects in the 3D
space around the user can be presented without putting a load on the visual or auditory
modalities. Using such VR techniques, the presentation of information can be prioritised
and re-routed depending on the situation in which tasks have to be performed (time
pressure, workload, context).
VR technologies can save costs. For example, remotely flying an unmanned aerial
vehicle (UAV) requires high-band width (and thus high-cost) video connections. Low
bandwidth connections generally yield a limited field of view, low update frequencies
and latencies. Together these effects substantially decrease operator performance
(overshoots, missing targets). Van Veen (replacing Van Breda) presented a series of
studies showing that UAVs can be successfully flown even at a low bandwidth by using
VR as an interface technology. This is done by embedding the camera image in a virtual
world (augmented reality) in which the visual feedback following an operator’s action
(e.g. rotating the camera) is anticipated and thus overshoots are reduced. Therefore,
search and fly performance are increased. Van Veen showed that using VR as an
interface can be equivalent to a band width increase by a factor of 400.
Auditory displays have not received the attention of visual and haptic displays in the VR
community. This is surprising given the fact that 3D sound displays are highly developed,
low cost and highly effective. Shilling (development of Virtual Auditory Interface),
showed the application of 3D sound in cockpits can reduce the time to complete an attack
in some situations by almost 40%! It is also surprising that most VR modelling tools
ignore 3D audio (Kalawsky). Also not available are models to represent the multiple
modalities of human information processing.
To conclude we can say that current research shows a need for:
• natural interaction devices in VE (usable, intuitive metaphors, multimodal);
• alternatives for cognitive task analysis in designing VE applications (rapid application
• team interaction in VE (models of human and object behaviour and social processes);
• design guidelines.
2.2 Bottlenecks and opportunities
During the workshop discussions forty participants from military organisations, academia
and industry put forward their opinions on the biggest bottlenecks and opportunities in
the development of military VR applications. Listed below are the applications
mentioned by workshop participants that represented the greatest opportunities for VR
technology(the number of times an item was mentioned is given between brackets):
• (17) training;
• (6) planning, mission rehearsal and debriefing;
• (3) integration of simulation and operation (real missions, augmented with VR);
• (1) remotely operated systems;
• (1) product design.
The most important bottlenecks mentioned are:
• (15) Most VR developments are not user driven: insufficient involvement of the
users, insufficient co-operation between designers, users and human factors people,
lack of natural interfaces, not enough attention for motion sickness and display
• (6) Lack of standardisation. This makes it hard to integrate systems and software
• (5) Not enough budget;
• (4) Not enough knowledge/imagination: behavioural models for people as well as
objects, scenario generation methods, no ‘out of the box’ ideas (people use VR to do
the same things they always did).
Obviously the sample of attendees of this workshop perceive training as the greatest
opportunity for VR applications which is reflected by the number of presentations on this
subject (Sastry: VR training of flight deck operations; Johanson: low-cost mission
debriefing system; Todeschini: VR training of mine-clearance). VR for training can
reduce cost and risk of casualties and improve flexibility and performance monitoring. At
the same time a number of factors seem to frustrate successful applications in this field
(lack of attention for human factors, lack of standardisation, lack of money, and a lack of
knowledge on how to model human and object behaviour.
The consequences of a lack of attention to human factors were mentioned by Prof.
Kalawsky in his keynote presentation:
• poor user interfaces and crude input devices;
• poor multi-sensor integration (inconsistent visual, auditory and proprioceptive
• poor facilities for team interactions (poor visual human representations and
communication tools);
• a parameterisation of immersion (as an assessment metric) has been fruitless so far.
2.3 Recommended actions
The way to overcome the bottlenecks mentioned above could be:
• identification of a killer application in the field of training (focus);
• involve human factors experts in the development of this application;
• develop VR design guidelines (see Kalawsky).
demonstrate convincingly the value of VR to the military (budgets).
To this purpose the military should develop a vision on the use of VR technology and
more clearly specify their needs. Industry should work on standardisation and should
substantially bring human factors into their development process. Academia and research
institutes should co-ordinate and accelerate their long-term research efforts to focus on
natural interfaces (innovative metaphors) and on how to model human and object
behaviour. In the short term academia should focus on human factors metrics and metrics
for team performance (cognition, communication), and a standard evaluation
methodology (Kalawsky).
Specific suggestions made during the workshop which could contribute to solving the
bottlenecks are:
• Establish an open NATO specialist group to:
• identify killer applications;
• identify a target list of user requirements and technologies for investment by
the military;
• foster development of behaviourally realistic intelligent agents and models;
• bringing together interdisciplinary groups and create common vocabulary on
shared problems;
• create a research network and identify new funding sources;
• share software libraries and create a central depository of devices and
modules; and
• open the non-classified publication of results to other organisations.
In general, better co-ordination between military organisations, industry and academia is
necessary in order to identify gaps in current knowledge and co-ordinate research.
The enthusiasm of the workshop attendees and the evident willingness to share ideas and
to discuss their findings provide a promising base for such co-operation. At the end of the
workshop the attendees had formulated a unanimous request for follow-up meetings to
work on, exchange and monitor progress on the above mentioned points. This could be
implemented in the form of an annual workshop on military applications as a satellite of a
major conference on Virtual Reality (e.g. VR2000, organised by Burdea).
What is essential for Virtual Reality Systems to meet Military Human Performance
Answers to this question centre on the three focuses of the workshop -- functional
requirements, state of the art, and future directions. Day 1 of the workshop detailed the
military requirements from which we derive performance goals. Prof. Kalawsky told us
that the environment, personal capabilities, individual motivation and the overall situation
govern human performance. Thus, the first partial answer is Military Human
Performance Goals include interacting within the training environment in the same way
we will interact in the real environment -- train like we fight. Day 2, state of the art,
spoke to the techniques and technology available in the marketplace, both commercial
and military. Speakers set the baseline for the VR systems that exist today. Participants
discussed in small groups the issue of what might be the bottlenecks and roadblocks to
maturing the technology, so that the military potential would be fulfilled as well or better
than industry applications. In the highly competitive world of entertainment and
automobiles, non-productive techniques don't last.
Simply stated, the second partial answer is that baseline applications are solid in the
automotive industry and entertainment industry, and military applications are beginning
to emerge and be evaluated. Day 3 looked to the future of VR. Work in considering VR
for teaching mental representations of knowledge, e.g. spatial knowledge, for enhancing
the sense of presence (3D audio effects), and for educational intervention techniques,
such as enhancing quality, quantity and retention of skills. The third partial answer to the
workshop title question, then, is the successful military application of VR depends first
upon multi-disciplinary implementation teams of scientists, engineers, practitioners, and
users, and secondly upon continued advancement of technology toward increased fidelity
to the real world. So, in the ensuing years following Sutherland's 1970 "Scientific
American" article that introduced the phrase virtual reality, we have seen literally
thousands of projects emerge and we are beginning to see return on investment.
However, we need continued investment and synergy for military goals to be met.
3.1 Future Work
Future work is divided into discussions of near and far term work. The intent here is to
give the policy makers of NATO an idea of what is emerging shortly versus what will
need continued investment. In the near term we can expect emergence of the following
Human performance measures derived in VR more quickly than from the real world;
Solutions to side effects that allow prolonged exposures to VR;
Usability guidelines for VR at the level of detail that we now have for GUI;
New metaphors linking past knowledge to new concepts being taught;
Behavioural models of human stress, emotion, fatigue, anxiety and other human
Intelligent tutors, agents, and behavioural models that enhance the cognitive
challenges of training;
Wearable computers that mix reality and virtual reality to produce superior
Networking for collaborative work from design to implementation to decision
Visualisation techniques that consolidate vast data into comprehensible information;
Smaller, faster, cheaper technology from industry, and NOT necessarily meeting
military needs.
In the longer term, the military needs a much closer synergy with academia and industry.
The trends of reduced personnel, reduced budgets, more accountability, increased
demand for return on investments and the expanding military role in operations-otherthan-war, will continue to strain the resources and limit the financial influence of the
military upon VR training technology development. In effect, the longer term strategy
does not yet exist that would give NATO members the science fiction sense-of-presence
of the holodeck or that of the movie, "The Matrix," where the effect was so real that
humans couldn't tell the difference between what was reality and what was not. That
longer-term strategy should exploit the near-term emerging technologies and attempt to
influence the direction of longer-term investments by industry and academia.
3.2 Future Meetings
The simplest answer to the question addressed by the workshop is continued involvement
by NATO members in the application of VR technologies to meeting military
requirements. Since VR is an integration of technologies to include modelling,
simulation, graphics, haptics and audio, and human factors considerations a
multidisciplinary approach is needed. Likewise, NATO will want a central focus of
military applications of this very critical training technology. Rapidly reconfigurable, low
cost, highly effective training environments don't exist. Pick any two of those criteria and
the third becomes unachievable. Yet the promise of VR is the possibility of all three for
military training. Such a potential seems well worth continued investment, influence and
involvement by NATO member countries.
In addition to M&S, educational applications may also provide an avenue of approach.
Many countries are investing in Internet capabilities for their citizens. The network will
soon be as comprehensive as the telephone and television. VR immersion technologies
and distance learning principles coupled with broadband Internet distribution would
eliminate the need for military capital investment and allow cost effective delivery of
training materials. Thus, collaboration of military agencies and civilian educational
agencies may be a synergy for combined resources, and long term technology
development strategy that would provide the critical mass necessary to influence industry
and academia toward training needs. NATO member nations could factor this strategy
into some of the thinking about operations-other-than-war.
Medical applications present a further avenue. The Human Computer Interface is
currently unacceptable for complex systems. Keyboard, joystick and mouse instruments
will give way to EEG, voice, haptic and eye interfaces as technology moves toward
human-centred design and network-centric warfare. Already, medical applications of VR
appear viable for training surgery and diagnostic procedures. Similarities of human
functions need further exploration. For example, France reported mine detection training
to be very similar to training for medical personnel to insert a needle. Continued analysis
for functional similarities between and across disciplines such as medical applications
would give additional leverage to military operations training techniques. The work in
metaphor development for VR is one step in this direction. Also, continued understanding
of internal human communication mechanisms may provide better Human Computer
Interfaces than currently exist.
One conclusion is clear. There is no obvious strategy, no clear consensus and no simple
combination of techniques to achieve military performance goals using VR. Three
possible strategies are presented, above.
Appendix A: Distribution
RTA Director for approval of publication in proceedings;
HFM-021 members;
Workshop Attendees.
Keynote Address: What is Essential for Virtual Reality
Systems to Meet Military Human Performance Goals?
Roy S. Kalawsky
Advanced VR Research Centre1
Loughborough University
Loughborough, Leics, LE11 3TU
The origins of virtual reality can be traced back to the
late 1970s and early 1980s with the pioneering research
in military crewstation design involving electronic
cockpits during. Many of the enabling technologies have
even earlier origins than this. More recent developments
in the commercial world have resulted in remarkable
improvements to some of the limitations of the early
generation systems. As the cost of the technology falls
and the computational performance increases there is a
growing need to ensure that a VR system is optimised
for both the user and the tasks to be carried out. Unless
the complexities of the associated user interface are
understood and carefully controlled there is a high risk
that future VR systems will be extremely difficult to use
and may be completely ineffective. Sadly, the thrust of
most research groups is focussed towards improving the
technology without attention to human factors. It is
tempting to try and relate the user’s performance in the
real world with that achieved in a virtual environment.
However, before this can be done it is important to
establish whether or not it is valid to make such
comparisons. This paper focuses on the need to develop
a reliable methodology to address the complex human
factors issues.
Perceived Military Benefits
The development of military based applications of
virtual reality is driven by the following perceived
• Improved cost-effectiveness — through process
integration across the life cycle
• Improved quality of decision making
• Enable teamwork within MOD and with other
• Better understanding of defence issues e.g. Human
aspects of warfare
• Focus on system effectiveness rather than weapon
VR is a human centred interface
Even though VR has been evolving for many years we
still do not have a reliable or robust definition for VR.
The early definitions are themselves becoming outdated
as new interaction techniques are developed. To provide
a common reference for the term VR the NATO
Research Study Group (HFM-021/RSG-28 produced the
two definitions below:
Virtual reality is the experience of being in a synthetic
environment and the perceiving and interacting through
sensors and effectors, actively and passively, with it and the
objects in it, as if they were real.
Virtual reality technology allows the user to perceive and
experience sensory contact and interact dynamically with
such contact in any or all modalities.
Overall, these definitions are appropriate for completely
synthetic environments but do not fully address the
definition of an augmented reality system involving both
synthetic and real environments. The synthetic environment is used to augment or ‘fill-in’ information in the
real environment. From a military perspective the
augmented reality system is a very important class of VR
system because the technology allows additional information (such as tactical data) to be overlaid onto the real
environment. A good example of the use of a synthetic
environment is a computer generated terrain displays
that is overlaid onto the real world through head up or
head mounted displays.
The reason why it is important to amend the original
NATO definition of VR to include augmented reality is
to acknowledge the different human factors issues an AR
system provides. The following definition should be
appended to the NATO definition:
Virtual reality can be used to augment the real world and
compensate for missing sensory information or to enhance
the real world in a way that does not normally exist.
How Best to Describe a VR System: A User Centred
AR Taxonomy
The basis for a human factors review of a complex user
interface is a functional description of the important
interface characteristics. In order to develop a functional
description or taxonomy for a VR system it is important
to define the scope of the system being investigated.
Rather than take a technological perspective, it is much
more useful to take a user centred view. This has the
advantage of ensuring that human factors issues are
properly represented. This approach has already been
used with success (Kalawsky, 1996), and Figure 1,
For correspondence with author: [email protected], tel. +44 (0)1509 223 047, fax +44 (0)1509 223 940.
See also: http://sgi-hursk.lboro.ac.uk/~avrrc/index.html.
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
below, outlines the sensory modalities and system
interfaces of a generic AR system.
External environment
Figure 1: Top Level Functional
Decomposition/Taxonomy of any Generic HumanComputer Interface
The functional decomposition of VR, illustrated above,
has many uses:
• It can be used to describe pictorially any VR system
and allows easy comparison with other systems;
• The diagram can be populated with current
technology or future technology capabilities
• It is possible to used linked cells in the
decomposition to describe associated human factors
issues (either requirements or known problems).
Deconstructing the Framework
When dealing with the components of the taxonomy in
more detail it becomes apparent that not only are various
technologies catered for but also inherent in the diagram
are descriptions of the human factors issues and
underlying processes of human factors integration.
Direct human-machine interface: This refers to the user
and the functional interface/devices used to experience
and control a VR environment.
User: The user is defined in terms of sensory/perceptual
processes (e.g. visual, auditory, kinaesthetic, tactile and
olfactory) as well as actions that can be initiated by
(voice, hands, head and eyes). For completeness, the
olfactory sense is included and although current VR
systems do not exploit this sense, research in this area is
starting to emerge.
Output interface: This refers to techniques that can be
used to provide information to the human perceptual
Input interface: This refers to the means by which
human initiated actions can be converted into
appropriate information for use in the environment.
Information processing: This is where data is processed
for delivery by the output interfaces. Data from the input
interface are processed and used to control the
environment. This section also has links to the
application environment that governs what is actually
undertaken by the overall VR system.
Application environment: This is the simulation
software that dictates what the VR system will do in
accordance with external input from the environment or
whatever is initiated by the user. There is a close
relationship between this and the information processing
section. Example application environments include
training systems, flight simulations, molecular
modelling, assembly plants etc., and it is feasible for the
application environment to be networked to other local
or remote applications.
External environment: This represents the real
(physical) world that may be linked to the VR system.
For example, in a medical application it is feasible to
overlay a virtual image onto a patient via an optical
system. To achieve accurate registration of the real and
virtual environments it is important to provide a link
between the display technology, application environment
and information processing sections.
The functional decomposition can be broken down into
lower levels of detail as required and partitioned as
shown in Table 1.
Functional Category
A. Information Processing
Cognitive Agents
Data Management
Control-display Coordination
Data storage and Recording
Image Generation
Tactile Stimulus Generation
Kinaesthetic Stimulus
Auditory Signal Generation
Speech Processing
Switch Processing
Virtual Hand Controller
Head Sensor Processing
Eye Sensor Processing
Physiological Sensor Processing
B. Direct Human-Machine Interface
Image Display
Tactile Feedback
Kinaesthetic Feedback
Audio Production
Speech Transduction
Hand Operated Controls
Head Sensing
Eye Sensing
Physiological Sensing
External Environment Viewing
Visual Defect Correction
Table 1: Functional Category Pointer to Descriptive
Figure 2 shows how the functional decomposition is
applied to an augmented synthetic environment
(augmented reality) system. Each cell in the functional
decomposition corresponds to a descriptive pointer
where more detailed information is stored. A description
of the sort of data that is stored in the table can be found
in (Kalawsky, 1996) and details factors such as, current
technical specifications, future technical requirements
and human performance implications.
External environment
Cognitive agents
Visual defect
Data management
Switch & inceptor
Virtual hand
Helmet sensor
processing A11
Hand operated
Auditory signal
generation A7
Head & helmet sensing
Eye sensor
processing A12
Physiological sensor
Control/display co-ordination
Tactile feedback
Tactile stimulus
generation A6
Eye sensing
Physiological sensing
Figure 2: Detailed Functional Decomposition of an AR
The numbers used in the diagram act as pointers into a
series of descriptive tables (Table 2) that are used to
describe the technical specification of the enabling
technologies (current, predicted, or even novel future
Potential Use
B. Direct Human-Machine Interface
B1. Visual
Desk top
High res., non
immersive display Large screen
CRT, Plasma
Head coupled 360° field of regard CRT, LCD
Full immersion, high colour shutter,
res., low lag display
B9. External
Integration of
Overlay of virtual
display onto real
Augmentation of
real environment
Chroma key
No correction for
curved screens
Single user mode,
Low - medium res,
Light transmission
Field of view
between real and
Table 2: Example Extract from Low Level Cross
Referenced Data
Classes of VR System
From a technical perspective it is convenient to
categorise a VR system according to the degree of
immersion it provides. In this context, immersion refers
to the extent that the user is enveloped in a virtual
environment and is related to the technology employed.
Three degrees of immersion have been defined as:
Fully Immersive
• Non-immersive
Fully Immersive VR Systems
Fully immersive VR systems are characterised by being
able to completely envelop the user with a synthetic
environment wherever they are. It is tempting to think
only in terms of the visual channel but other modalities
such as auditory perception are equally valid. However,
in the majority of applications, the visual channel will be
the most dominant and the auditory channel will be used
to augment the visual channel.
Head Mounted Displays
The development of head mounted displays can be
traced as far back as the early 1950s. They were the first
display technology to deliver a fully immersive
experience and since the early 1990s there have been
many different designs for the head mounted display.
These tend to fall into two categories — non see-through
and see-through. As these terms imply, the non seethrough head mounted display does not allow the user to
see any part of the real world. Conversely, the seethrough head mounted display makes it possible for the
real world to be overlaid with computer generated
graphics generated by the head-mounted display.
Although it may seem that the see-through and non seethrough head mounted displays are similar they actually
present very different human factors problems that must
be considered in the context of the application and
operating environment.
Non see-through head mounted displays
The non see-through head mounted display typically
comprises two display devices (typically CRT or LCD)
and a set of optics to magnify and position the image a
fixed distance from the user. The image can be presented
anything from a few meters to optical infinity (beyond
250 meters). The exact distance is usually a design
feature of the head mounted display and is fixed by the
display manufacturer. The image plane distance is
extremely important and is a function of the application.
There are a range of non see-through head mounted
displays available and these literally come in many
different configurations.
Technical Description
There are many different configurations for non seethrough head mounted displays and it would not be
practical to review them all in this paper. Refer to
(Kalawsky, 1993b) for a more detailed account. Figure 3
shows a simplified non see-through head mounted
display with a simple magnifying lens. The distance of
the virtual image from the user is governed by the
distance of the image source (typically a CRT or LCD)
from the focal point of the magnifier lens. If the image
source is located at the focal point then the virtual image
is located at optical infinity. In practice the head
mounted display manufacturers tend to position the
virtual image much closer than this, typically 1–20m
away. There does not seem to be a particular preference
so the image distance can vary from one design of head
mounted display to another. For most tasks, this does not
make too much difference except perhaps where it is
important to visualise large scale building structures at a
scale of 1:1. In this case, the virtual image should be
placed as far away as possible. However, in the case of a
see-through head mounted display, image distance is
extremely important.
1280×1024 might be adequate. However, if the required
horizontal field of view is in excess of 140° then this
would probably be inadequate. There are so many other
trade-offs that have to be considered that it is no wonder
an off the shelf head mounted display is unable to meet a
particular requirement. As a comparison, in the military
sector head mounted displays are specially designed for
each application. Unlike commercial applications where
one display is intended to fit all applications.
User Issues
There is no doubt, head mounted displays are unpopular
with potential end users. Apart from the above problems
comfort is a major factor. Current off the shelf head
mounted displays are still too bulky for many people and
after short periods of use people report discomfort in
areas of neck strain, eye strain, claustrophobia and
nausea. The general health and safety issues of head
mounted display are now being understood and the
enabling technology is being improved gradually.
One of the biggest drawbacks of a head mounted display
system is that it provides a single person experience
whereas the current trend in VR is for group or multiuser interaction. As soon as the user puts on the head
mounted display they are isolated from the real world.
Figure 3: Non see-through head mounted display
Despite the technical difficulties associated with head
mounted displays progress is being made with
development of higher resolution display sources.
Whether or not these developments make it possible to
reconsider the use of non see-through head mounted
displays remains to be seen.
Despite these concerns the applications where non seethrough head mounted displays can be considered
• Large scale architectural visualisation where large
screen systems would be impractical.
• Maintenance training
• Research involving phobias where it is important to
isolate the user from the real world.
Figure 4: Relationship between Image Source and
Virtual Image of a Non See-through HMD
Head mounted display technology still lags behind the
requirements of most applications. Notably, the display
resolution is still far too low to be of practical value. It
should be stressed that display resolution must not be
considered in isolation. An equally important and related
parameter is the field of view of the optical system. If an
application calls for a narrow horizontal field of view
(for example, 40°) then a display resolution of
See-through head mounted displays
The more exciting though technically more challenging
head mounted displays are the see-through systems.
Interestingly, the very first head mounted displays were
based on optical systems that overlaid display
information over the real world. These systems were
later developed to become an important system for
fighter pilots. A number of very sophisticated systems
have been developed. Commercially available seethrough head mounted displays are now available and
are based on cheaper versions of the military systems.
The term augmented reality (AR) is frequently used to
refer to see-through head mounted display systems.
Computer enhancement of the external environment
offers distinct advantages over virtual reality by not only
potentially avoiding the need for complex modelling of
people and the environment, but also by providing an
anchor in reality that should reduce the likelihood of
nausea being induced. Instead of replacing the real
environment with one that is completely artificial, a
number of early researchers (e.g. (Sutherland, 1965);
(Furness, 1986); (Knowlton, 1977); (Krueger, 1985)
(Kalawsky, 1992, Kalawsky, 1993b)) have used
computers to augment the real environment. Augmented
reality systems offer the potential to allow user the to
actively carry out tasks involving real world objects
rather than being confined to an artificial environment
such as is the case for virtual reality based systems.
Figure 5 shows a modern commercially available seethrough head mounted display.
Figure 6: Basic Optical System for See-through HMD
The majority of head mounted displays for commercial
applications have been predominantly non see-through.
The reason for this may be the great difficulty that is
experienced when an attempt is made to register the
virtual display with objects in the real world. Any
misregistration that arises from calibration errors, lags in
the graphics or head tracking system etc. is immediately
apparent to the user (Kalawsky 1992a; Kalawsky 1998).
For the tasks suggested for see-through systems
(maintenance, design etc.) the misregistration has proved
to be very problematical.
Figure 5: Sony Glasstron Augmented Reality Head
Mounted Display
Technical Description
There are two main ways in which the real world can be
augmented by a graphical overlay:
a. A see-through head-mounted display can be
employed enabling the user to see the real
environment through part-silvered mirrors that also
reflect a visually superimposed graphic image into
the user’s eyes. The optical system relies on a
providing an integration of the real world with
information generated by an electronic display such
as a cathode ray tube (CRT). The external
environment is generally viewed through the
combiner plate and the image from the CRT is
reflected by the combiner plate. Refer to Figure 6.
b. A conventional VR head mounted display can be
used to provide a non-see-through augmented reality
display in which the user sees a video image of
reality combined with luminance or chroma-keyed
graphics (Kalawsky, 1991). An AR system based on
electronic overlay relies on a video mixing system
taking video from a television camera viewing the
real world scene and superimposing it with a video
signal from a computer graphics system.
User Issues
Although AR concepts have been around since the
1950’s the technology and its application is still in its
infancy. This in the main, has been due to technological
limitations of synthesising real and virtual images in the
same visual field, and fundamental problems of image
registration and collimation. In recent years, AR systems
have become more sophisticated and offer particular
advantages over VR concerning some of the human
factors issues that arise. For example, in the case of AR,
orientation cues are still available to the user from the
visual scene in the real world. Users are therefore
unlikely to experience the feelings of vertigo and
sickness that can be brought about by traditional VR
systems (Caudell, 1994). However, AR configurations
produce unique issues of their own.
Research into the human factors issues surrounding the
use of AR systems is very limited and few formal
guidelines exist for any application of AR technology.
Irrespective of which technique is used to provide the
electronic display overlay there are several technological
factors that must be considered. These include:
• Image plane position of the virtual image
• Transparency (or rather the transmissivity/reflectivity) of the combiner assembly.
• Registration accuracy of the electronic image with
respect to the external environment.
Each of these factors will have an influence on how and
what information is displayed to the user.
Image plane position: All virtual display devices
produce an image at a particular position from the eye.
The position of the virtual image can be as far away as
optical infinity and is controlled by the position of the
image source with respect to the collimating lens. For
example, if the image source is located on the focal point
of the collimating lens the virtual image is at infinity.
Virtual image position is usually set at infinity for
aircraft applications but is inappropriate for other
applications. Commercial off the shelf head mounted
display systems usually fix the virtual image position at
some arbitrary distance (e.g. 3 m). For most game
applications this distance has not been proven to be
critical. However, when these displays are used in
conjunction with information derived from the real
world (i.e. operated in see-through mode), virtual image
position is very important. Unless the virtual image is
collimated to be coincident with the information in the
real world a misregistration occurs. The net effect on the
user is the need to re-accommodate when attention shifts
between the information displayed in the real world and
that displayed on the head mounted display. Particular
care must be taken when using see through HMDs if
there is an accommodation/convergence mismatch with
the external environment. When operated with such
defects it is quite easy for serious eye strain to occur.
The long term exposure to such eye strain is not fully
understood. Not all see through HMDs suffer from this
effect. However, due to possible commercial and legal
implications it is not possible to identify the problematical see-through HMDs in this report.
Transparency (or rather the transmissivity/reflectivity)
of the combiner assembly: The optical design of an AR
display will determine what percentage of the real world
is transmitted through the display to the user and what
percentage of light from the image source is overlaid
onto the real world. The nature of the semi-reflecting
surface of the optical combiner (beam-splitter) also has
an effect. Some devices work by employing a notch filter
to maximise the percentage of light overlaid onto the real
world. Unfortunately, this has the effect of removing
certain spectral components from the real world.
Obviously, the impact of this depends upon the
Registration accuracy of the electronic image with
respect to the external environment: Registration
accuracy is very important in head tracked augmented
reality display systems because mismatches between the
information presented in a virtual overlay compared with
the real world may affect user performance. There are
two types of misregistration error. The first is caused as
a result of a static misalignment of the virtual image with
the real world. If this error is present, it can usually be
calibrated out. However, the second type of misregistration error is a temporal misalignment caused by delays in
the computer and tracking system. It is possible to create
an AR system with almost zero static misregistration
errors but as soon as movement occurs computational
delays introduce misalignment.
Advanced embedded training systems offer the potential
to train operators in their real working environments
rather than spending time being trained elsewhere. In the
future, such systems may provide on-line feedback to
operators, perhaps tailored to different levels of operator
expertise, offering dynamic and flexible alternatives to
conventional training facilities (Zachary, 1997).
Semi-Immersive VR Systems
Semi-immersive VR systems represent a very exciting
class of system because they overcome many serious
problems associated with head mounted displays.
Distinct advantages include:- higher-resolution displays,
multi-participant experiences, wide angle display. A
semi-immersive display does not provide a fully
enveloping display image. Depending on the display
technology used a field of regard of up to 270° can be
obtained. Field of regard refers to the extent of the
displayed image expressed in angular terms. The term
field of view is sometimes used to represent the same
thing. However, to be strictly correct, ‘field of regard’
refers to the instantaneous field of view as perceived by
the observer with fixed head position. The field of regard
refers to the total display field of view that can be seen
by moving the head around. The field of regard is
therefore potentially larger than the field of view.
Flat Screen Systems
Though not normally considered by some to represent a
semi-immersive display system, it is feasible to refer to
flat projection screen based systems as a semi-immersive
display, provided the field of view is greater than 90°
horizontally, Figure 7.
Figure 7: Wide Field of View Flat Screen Projection
System (Rear Projection)
Technical Description
The flat screen system can be based on a single or multiprojection display system. The actual arrangement of
projectors depends upon the required resolution of the
whole system (in horizontal and vertical extent). A
single projector can achieve a maximum resolution of
about 1600×1200 pixels. (Please note, there are
specialised higher resolution projection systems
available but these are likely to be expensive. The more
typical resolution is either 1280×1024 or 1024×768.
Quite acceptable large screen displays can be produced
with these resolutions from either CRT or LCD
technologies. If higher resolutions or fields of view are
required then it is important to increase the number of
projectors. A typical arrangement is shown in Figure 8.
Rear or front projection can be used. The exact choice
being dependent on the way users interact with
information on the screen. The main problem with front
projection is if users need to get close to the screen they
can cast a shadow, which obscures displayed
information. However, a rear-projected display does not
suffer from this problem. Due to the composition of a
rear projection system the screen material tends to
diffuse the light more than a front projection system and
this can lead to an image with poorer contrast.
because the projector has to compensate for large
keystone errors. This is illustrated in Figure 9.
User Issues
The flat screen display systems present fewer user issues
provided the user does not rely on display peripheral
cues too much. If a single projector is used then the need
for frequent and sometimes difficult alignment between
projectors is avoided. If a CRT based projector is used
instead of a LCD based display then the user must be
prepared to re-converge the three CRTs occasionally to
maintain the system at optimum performance.
Image 1
Image 2
Image 3
Image 2
Image 1
Image 3
Figure 9: Arrangement of Projectors to achieve
Lower Levels of Distortion
Image 1
Image 2
Image 3
Figure 8: Arrangement of a Flat Screen Multiple
Projection Display
Whichever projection method is used it is very difficult
to match the display output from different projectors
precisely because of the errors (optical distortions)
present in all projection lenses. The magnitude of the
error increases the further you move away from the optic
axis of the lens system. In order to combine two images
together it is necessary to overlap the image of one
projector with another projector by a few degrees.
Special electronics units (known as an edge-blenders)
are used to match the image edges together. The more
sophisticated edge blenders enable an accurate colour
balance to be achieved between the overlapping
projected images. In the past, there have been attempts to
perform the edge blending in the graphics system but
this adds to the computational complexity of the system.
Unfortunately, this affects overall system performance
and edge blend deficiencies become very noticeable in
dynamic display imagery.
A further issue is the physical placement of the
projectors. In other semi-immersive systems it is
desirable to position the projectors close to each other
but in very large flat screen systems this is not feasible
Figure 10: Arrangement of Projectors for Flat Screen
Over time, the three CRTs will slowly drift out of
alignment and the edges of objects displayed on the
screen will have a coloured ghost like image drawn
around them in one or more of the primary colours. This
indicates that the CRTs are out of alignment. Obviously,
in a multiple projector system the user must be prepared
to re-converge each projector and to align each projector
with respect to each other. Changes in the thermal
environment in the laboratory frequently account for this
misalignment. It is worthwhile considering the use of a
temperature controlled environment for multiple
projector systems as this will certainly help reduce the
number of times projector alignment is carried out.
The flat screen system can be used for a wide range of
applications. The flat screen semi-immersive display is
without doubt a very cost effective way of creating a
compelling display environment. In its simplest form, a
single projection system can be used requiring only a
single graphics system. As more and more projectors are
used the complexity of the graphics system and the
requirement for an edge blending system increases the
overall cost.
Immersive Workstations
The immersive workstation is a term used to cover a
number of small flat screen projection systems that
provide a powerful visualisation capability. These
systems are very simple in concept and can be bought for
relatively low cost.
Viewing Direction
Rear Projection Screen
Fold Mirror
Figure 12: Schematic Diagram of an Immersive
User Issues
Immersive workstations are very useful tools for
visualising 3D objects in stereo mode. Users should be
aware that they offer a fairly restrictive viewing/
operating area due to the nature of the stereo display
system. It is not possible to provide a correctly computed
view for more than one person in head tracked mode. If
other people wear the stereo shutter glasses, they will see
a stereo image provided they are in front of the display.
However, the image will not be geometrically correct for
their viewing perspective. This may not be an issue for
many applications.
Figure 11: Immersive Workstation
Technical Description
The configuration of the immersive workstations is very
simple. It relies on a high brightness projector and a rear
projection screen. The screen is orientated at an angle
rather than being placed vertically in front of the user,
Figure 12. Some immersive workstations allow the
whole table to be rotated through 90° from horizontal to
a vertical position. Figure 12 shows a very simple
diagram of a typical Immersive Workstation. The fold
mirror is used to simply keep the size of the unit to a
more manageable size. Without the fold mirror the
required distance between projector and screen to create
a large image would be impractical for many
The stereo display is produced in a frame sequential
manner whereby alternate left and right eye images are
presented by the projector. In order to see a stable stereo
image the user must wear special glasses that shutter the
left and right eyes in synchronism with the projection of
left and right images. A small infra-red transmitter is
used to send a signal to the stereo shutter glasses so that
the left/right eye shutter can be correctly synchronised.
The immersive workstation is a very versatile device and
can be very effective for small group interaction (1–4
people). However, if head tracking is employed then the
person linked to the head tracking system will be the
only person able to see geometrically correct stereo
Figure 13: Silicon Graphics Reality Centre – Theale
Reality Centre Systems
A further development of the multiple projector flat
screen system has lead to the evolution of a curved
screen system, Figure 13. This is extremely similar to the
flight simulator used in the aerospace industry albeit
without the cockpit. Instead of a cockpit, a group of
users are situated at the focal point of the curved screen.
The main characteristic of the curved screen system is
that the user perceives a greater field of regard and the
edge distortion effects noticeable in a flat screen system
are almost eliminated. Single or multiple projectors are
used, the number being a function of the required display
resolution in horizontal and vertical extent.
Silicon Graphics coined the term Reality Centre TM to
cover this type of system.
Technical Description
The basic principle of a Reality Centre lies with the
arrangement of the projection system which is situated at
the focal point of a curved screen (Figure 14) whose
horizontal extent is anything from 90–180° horizontally.
To achieve the wider fields of regard it is usually
necessary to employ multiple projection systems and
overlap their screen edges. A video edge blender is then
used to blend the edges of different projectors together in
a way that make the overall image appear as a single
uniform image.
User Issues
The curved screen Reality Centre is a very convenient
tool for many applications. The curved screen currently
rules out non CRT projectors (e.g. LCD) since it is not
possible to incorporate distortion into the image to
compensate for the curved screen. As higher resolution
projectors become available it will be possible to replace
multiple projector configurations with a single projection
system. This will greatly facilitate system maintenance
and remove the need for edge blending systems. Setting
up and maintaining a single projector can be quite time
consuming if continuous peak performance is required.
If multiple projectors are used it is necessary to achieve
alignment of each projector against a projected reference
image. Over time, the CRTs used in the projectors will
age at different rates and between channels. The better
edge blending technology will permit some degree of
colour/matching between adjacent projectors as each
CRT ages. However, if the system is used frequently
then it might be advisable to rotate the projectors around
or be prepared to swap out the CRTs on a more regular
Edge Blend
In order to reduce the amount of re-calibration it is
advisable to maintain the temperature of the room where
the Reality Centre screen and projection is installed.
Temperature drift is one of the main causes of the image
going out of alignment.
Reality Centres can be used in all sorts of application
though the educational value has yet to be determined.
One of the particular strengths of a Reality Centre along
with flat screen and Vision Dome systems is that they
are ideal for groups of people. The cost of owning such a
facility can be very high but alternative ownership
schemes such as leasing might prove to be extremely
attractive. A number of organisations have established
Reality Centres as a commercial centre where they hope
to sell time on the facility to external organisations.
A very interesting development has taken place with the
flat screen projection system. Instead of employing a
single flat screen, several screens are used at right angles
to each other Figure 15. By arranging for each screen to
be orthogonal with respect to each other, it is possible to
create a room called a CAVE, whose walls are formed
from rear projector screens. The CAVETM is a multiperson, room-sized, high-resolution, 3D video and audio
environment. The CAVE was developed at EVL
http://www.evl.uic.edu) and is available commercially
through Pyramid Systems Inc. There have been many
earlier examples of such screen-based systems involving
rear-projected displays. Many of these have originated in
the Aerospace industry. In 1987 the author saw a very
early example at Wright-Patterson Airforce base. From
1990 British Aerospace used a multi-faceted display
system for its cockpit research programme. These early
generation systems were not known as CAVES but had
all the properties of today’s CAVE systems.
Edge Blend
Figure 15: CAVE Display
High Resolution
Figure 14: Diagrammatic Representation of Reality
Centre Screen System
Technical Description
There are many variations of the CAVE concept but they
all rely on the principle of the user being surrounded by
three or more orthogonally arranged rear projection
screens. Figure 16 shows a top view of a three-sided
CAVE system. The three projectors are situated outside
the inner projection viewing area. In practice, the size of
the image on the walls of the CAVE is a function of the
lens used and the projection throw distance. It is possible
to extend the number of sides to a CAVE up to the
maximum (i.e. six) by carefully positioning additionally
projectors around the CAVE.
It is desirable to use as large a CAVE as possible. This in
turn requires a large space in which to site the CAVE
and associated projection equipment. The space
requirements of a six sided CAVE should not be underestimated. In order to reduce the amount of space
required to realise a CAVE, mirrors can be used to
increase the effective throw distance as shown in
Figure 17.
Obviously, the cost of a CAVE system increases as a
function of the number of sides is employed. Each side
will require its own dedicated graphics channel and
projection system.
Image 2
Image 1
Image 3
Figure 16: Three Sided CAVE Configuration
Image 2
Fold Mirror
Image 1
Fold Mirror
Image 3
Figure 17: Arrangement of Fold Mirrors to Increase
Effective Throw Distance and Reduce CAVE Space
Requirements (Note Front View)
The blending of edges in a CAVE becomes very
challenging because of the abrupt angular changes that
occur at the junction of the sides of a CAVE. Dedicated
edge blending technology exists that will match the
geometry of the corresponding points of one CAVE side
with another. The edge blending will always be a
compromise because of the viewing geometry with
respect to the angle of the sides of the CAVE.
A requirement of all projection display systems is the
reduction of veiling glare caused by reflection of light
from surrounding areas from the projection surface. In a
CAVE the light from one side will inevitably be
reflected from the opposite side and cause a significant
reduction in contrast. Many CAVE users frequently
complain about the poor contrast displays.
User Issues
A six-sided CAVE can provide a total immersive
experience for one user. The display presented on each
wall of the CAVE can be a stereo image in which case
the user perceives a display with depth. Care must be
taken to carefully calibrate a CAVE system otherwise
the user can experience nausea effects. Due to the nature
of a head-tracked CAVE, if other users are present, they
will obtain a distorted image because they will not be at
the same viewpoint of the person being tracked. The
technology does not exist yet whereby multiple users can
be tracked in the CAVE so that each gets a correct view.
One of the main problems present in head mounted
display systems is the accommodation/vergence effect.
The accommodation/vergence system is tightly coupled
in the human perceptual system. When the user fixates
on an object the accommodation response is partly
driven by the eye vergence. Any errors in the object’s
distance perceived by the vergence system and the
accommodation response will cause eye strain for the
user. The head mounted display produces an image at a
fixed focal plane. A CAVE system can similarly present
accommodation problems because for a given viewing
position the observer’s eyes have to re-accommodate if
the object under view is displayed on two or more
display surfaces. Even if the CAVE system can be
calibrated, it is not possible to compensate for the
different accommodation required as an object is viewed.
It is possible that people may develop nausea or motion
sickness that was symptomatic of head mounted display
The veiling glare problem briefly described above is
often the source of complaints of poor contrast by users
of the CAVE system. If the background scene can be
kept quite dark (not always possible with some virtual
environments) then the veiling glare can be kept to a
If the user in a CAVE rotates or rolls the image, it is
possible to induce sufficient visual cues to interfere with
the user’s balance so that they fall over. This is
particularly the case for a six-sided CAVE where it is
possible to loose sense of true horizontal. This
phenomenon is exploited in some fairground ride
systems and people do actually fall over. Some people
are more susceptible to these effects than others.
A CAVE system is very expensive so this limits their use
to applications where the cost can be justified. The
following application domains are evaluating CAVES:
• Automotive
• Architectural
• Art
• Oil and gas sector
Vision Dome Systems
The Vision Dome concept is not a new idea. It is based
on the astronomical planetarium which instead of
projecting a film-based image onto a spherical surface,
the film projector is replaced with a CRT based
projector. The Vision Dome theoretically can present a
full 360° field of regard image but in practice the full
field of regard is seldom used. Figure 18 shows the
Loughborough University Vision Dome.
Figure 19: 5m Vision Dome Schematic
The single projector means that a single graphics pipe is
required. However, the spherical nature of the screen
requires real-time distortion correction to be applied to
each image before it can be displayed in the vision
dome. Fortunately, many high performance graphics
systems can cope with the increased computational load.
Portable versions of the Vision Dome exist and employ
similar air pressure maintained screen systems. Apart
from being smaller they do not require the large external
support structure. This means that erection of the
assembly takes a matter of a few hours instead of several
Figure 18: BT/Loughborough University Vision Dome
Technical Description
The Vision Dome comprises a hemispherical projection
screen that is driven by a single projector located at the
focal point of the screen. Figure 19 shows a typical
Vision Dome installation.
The necessity for a single projection system at the focal
point of the hemispherical screen places a requirement
for a very high-resolution projection system. The optical
system has been designed to cover the whole
hemispherical surface and the projection system remains
as the limiting factor.
The screen used in the Loughborough University 5m
Vision Dome is quite interesting in that it is maintained
by air pressure. An internal aluminium structure
comprises two layers (one layer is the screen). Air is
drawn out from between the two layers and this forces
the screen to take on a perfect hemispherical shape.
User Issues
The Vision Dome does not present an accommodation
conflict as in a CAVE system to the user because the
image plane is maintained at a consistent distance from
the user across the whole field of regard.
One of the definite advantages of the Vision Dome
compared with a CAVE is the use of a single projection
system. Therefore, the need for matching different
projectors and display surfaces is eliminated. However,
there is a need to employ a special graphics library that
replaces standard graphics calls with modified versions
that take into account the required distortion correction
for the spherical dome surface. Modified graphics
libraries are available for most platforms including NT
based PCs. Since only one projector is required this
means that the projector must be extremely high
resolution to cover the whole field of view.
Unfortunately, requires a more expensive projection
system but means that a single pipe graphics system can
be used with a significant reduction in cost.
Effective 3D user input devices are still required in
common with all other VR systems.
Non-Immersive VR Systems
Desk-top Systems
It is considered unnecessary to review in any detail the
use of non-immersive desktop VR system since these are
based on conventional display monitors. It is possible to
drive many display monitors in frame sequential stereo
mode and achieve the benefits of an immersive
User Issues
Without a doubt, a desktop CRT monitor will give the
best image in terms of the following:
• Resolution
• Contrast
• Clarity
• Colour gamut
For some very critical users, there is simply no
alternative technology. Even modern LCD monitors fail
to compare with the highest quality CRT display.
Portable VR — Wearable Computing
VR technology is normally associated with large fixed
installations but advances in wearable computing
technology have made it possible to produce mobile VR
systems. There are a number of military initiatives
around the world looking at how the future solider would
be deployed with technology. One such programme is
known as Dismounted Infantryman. Loughborough
University are addressing some of the complex human
factors issues of wearable computers, Figure 20 shows
the second generation system.
Figure 20: Loughborough University In Field Computer Mark 2
The main operational feature of wearable computing
systems are:
• Hands free interaction
• Contextually aware
• Always on and assisting the operator
The use of high powered portable computing devices
presents a whole new series of human factors issues.
These will not be discussed in this paper because of
space constraints.
Human Factors Issues of VR Systems
The functional requirement of a VR system is driven by
the end application and must take into account human
capability (physical & cognitive).
Sensory Conflict!: Real World versus Virtual
In order to deal with the complex area of human
performance and effectiveness (the goal of human
factors) it is important to note the distinction between
real and synthetic environments. The most appropriate
way of doing this is from a user’s perspective. The user
will generally experience sensory conflict in the
synthetic environment, even though they may not be
immediately aware of the effects. It is easy to recognise
the effect of sensory conflict when ones compares a real
world experience such as a roller coaster ride with a
video of the same experience. The sensory inputs to the
rider on the roller coaster will be through, vision, sound,
smell and proprioception. The rider will also experience
a wide range of rich sensory cues such as air rushing
over the face, the sound of the roller coaster, other riders
screaming and shouting, sense of vibrations, extreme
inertial forces, accelerating, turning and descending and
the intense emotions of fear and excitement. In contrast
to this, a video of a roller coaster ride provides typically
two sensory inputs, vision and sound. Not only are the
number of sensory inputs limited they also tend to have a
lower fidelity than in the real world. Additional effects
are also present such as temporal lags introduced by the
video system. All these inherent features reduce visual
fidelity of the experience. It is also possible that some
sensory cues may even be contradictory.
The ‘Perceptual Sense of Being’ in a Virtual
Environment — Presence
An important differentiating characteristic of VR
systems compared with other human-computer interfaces
are their ability to create a sense of ‘being-in’ the
computer generated environment. Other forms of media
such as film and TV are also known to induce a sense of
‘being-in’ the environment. Some VR practitioners have
tended to use the term presence to describe this effect
(Sheridan, 1992), (Heeter, 1992), (Kalawsky, 1993a),
(Zelzter, 1994), (Hendrix and Barfield, 1996). This
means that people who are engaged in the virtual
environment feel as though they are actually part of the
virtual environment.
In order to understand what it means to be present in a
virtual environment it is necessary to understand what
characteristics of the real world enable us achieve a
sense of presence. A good example of real world
experience is a roller coaster ride. The sensory inputs to
the rider on the roller coaster will be through vision,
sound, smell and proprioception. A roller coaster rider
will experience air rushing over their face as well as the
sound of the roller coaster and other riders screaming
and shouting. The sense of vibrations and extreme
inertial forces when accelerating, turning and
descending, etc will be very real. It is obvious that most
people will also experience intense emotions involving
fear and excitement. A video of a roller coaster ride there
are only two sensory inputs, vision and sound. Both of
these sensory inputs will be much less real than the real
world. Stereoscopic depth cues will be absent from
visual and auditory information. Temporal lags
introduced by the video system will further reduce the
visual fidelity of the experience. Some sensory cues may
even be contradictory. The body will feel comfortable in
a normal seating posture but the visual cues (with
reference to the horizon) will be indicating that the body
is anything but stable. If the roller coaster is experienced
in an IMAX cinema the reaction of others will have an
effect. Some people report that they can suppress the
sense of presence by weakening or strengthening their
awareness. Upon receipt of sensory information, some
people can fill in gaps to create a better or enhanced
sense of what is happening. For example, people who
have previously experienced a roller coaster ride would
experience a different state of awareness than someone
who had never experienced the ride. This implies that
previous experience may affect the sense of presence.
Intersensory Interactions
Traditionally, sensory modalities have been investigated
in isolation from another. It has been suggested by
(Sherrington, 1920) that all parts of the nervous system
are connected together and no part is capable of reaction
without affecting or being affected by other parts. This
means that examination of part of the system will
inevitably lead to an incomplete understanding of the
perceptual experience. Intersensory interaction relates to
the perception of an event when measured in terms of
one sensory modality which is changed in some way by
the concurrent stimulation of one or more other sensory
modalities. Given the nature of the human sensory
system there is great diversity in the intersensory
interactions that can be experienced and this adds to the
difficulty in understanding what is happening.
Spatial Location
There are at least four sensory modalities that are
capable of providing spatial information to the human
being. These are visual, auditory, tactile and
proprioception. The visual sensory modality is the most
spatially acute of the spatial modalities with a resolution
acuity of about 1 min of arc. In contrast, the ability to
spatialise a 1kHz tone placed in front of the participant’s
head at varying angular distances from the median plane
of the head gives a minimum angle of about 1°. Tactile
acuity is a very difficult thing to define because it
depends on what part of the body is being stimulated.
The tongue has a two-point threshold of about 1mm.
There are four sensory modalities that support the
perception of orientation: visual, tactile, proprioception
and vestibular sense. Proprioception is a very powerful
mechanism for conveying a sense of body orientation
though it has been shown that with time the body can
adapt to unusual positions and this can lead to false
orientation cues being perceived. The visual and
vestibular senses are extremely accurate in conveying a
sense of orientation of gravitational direction. This is one
of the reasons why the perceptual system can make
serious errors in orientation judgement if one of these
two sensory modalities is missing or conflicting with the
other. Misperception of the body and the gravitational
direction vector can cause a shift in auditory localisation
cues (Graybiel and Niven, 1951). This phenomenon is
known as the audiogravic illusion.
Egocentric localisation
Egocentric localisation is the ability of the human to
perceive the direction and distance of objects relative to
the observer. Egocentric localisation is achieved by the
visual, auditory, tactile and proprioception sensory
modalities. It is usual for several of these modalities to
act together to give an accurate sense of localisation.
In the real world it is common for several of sensory
modalities to receive simultaneous stimulation in a way
that reinforces a common multi-modal perception.
However, in the virtual environment system it is possible
that one or more of the sensory modalities will receive
incorrect stimulation due one of the sensory channels not
being provided. This phenomenon is sometimes referred
to as intersensory bias.
Whilst we can examine sensory interaction and relate
this to specific human capability, the term presence has
defied all attempts to define it in a quantifiable manner.
There is clearly a coupling between the senses and the
phenomenon of presence (Gilkey, 1995). Gilkey has
examined the level of presence experienced by suddenly
deafened adults. These deaf adults frequently complain
of a sense of unconnectedness with their surroundings,
which supports the view that auditory cues are important
for establishing a sense of presence. It is unfortunate that
many people make the mistake of assuming that the most
important cue in a virtual environment is the visual
modality. Even if we concentrated entirely on the visual
channel there would still be sufficient auditory cues
around in the real world (including self generated
auditory noise such as breathing) to limit the sense of
sensory depravation that is reported by suddenly deaf
people. Interestingly it has been reported by
(Gillingham, 1992) that acoustic isolation and lack of
auditory cues may account for spatial disorientation.
The term immersion is also sometimes used erroneously
to describe the experience of presence. The term
immersion in fact refers to the extent of peripheral
display imagery. If the display presents a full 360°
information space then we are dealing with a fully
immersive system. However, if the extent of the display
is less than this we have a semi-immersive system. The
term non-immersive is usually reserved for desk-top VR
systems. To avoid confusion it is best to associate
immersion with the technology characteristics of the
display. Unfortunately, these terms are not
interchangeable and refer to quite different things.
Presence is essentially a cognitive or perceptual
parameter whilst immersion essentially refers to the
physical extent of the sensory information and is a
function of the enabling technology.
Perceptual Conflicts — Phantom Illusions
Gibson (Gibson, 1986) has mentioned the notion of coperception of one’s own movement, in other words
awareness of locomotion. The visual system “is
kinaesthetic in that it registers movements of the body
just as the muscle-joint-skin system and the inner ear”.
In the real world, the visual system perceives
information about the environment and one’s own self in
that environment. Our whole perceptual system behaves
in this manner and processes many reinforcing cues from
the environment. It is better to think of these cues as
reinforcing since they are all contributory rather than
some current views that suggest these cues provide a
degree of redundancy. Visual kinesis is a powerful
perceptual process as evidenced by a wide-angle
panoramic projection screen. It is quite easy to produce
very convincing and compelling visual cues that give the
participant a sense of self-locomotion. The visual
experience can appear to untrained people as a very
vivid illusion of reality even though the participant is
anchored to the floor. A similar illusion can occur whilst
sitting on a stationary train in a station and an adjacent
train pulls away. Sometimes, you become convinced that
you are moving and the other train is stationary. It comes
as a surprise when you discover that you are in fact
stationary. What is very interesting with these
experiments is the way visual cues can override cues
from the vestibular system. Although it is tempting to
isolate a particular sensory modality when trying to
explain perceptual phenomena it is problematical.
Someone who is completely blind would argue that they
can ‘see’ the environment through auditory, haptic and
kinaesthetic cues. Indeed, when deprived of the visual
channel you soon become aware how extremely
important the other modalities are. By allowing, the
person to move their head and move within the
environment, proprioception fills in much of the
information that would normally be provided by the
visual channel. The presence of all sensory modalities
removes some of the ambiguities that can occur with a
reduced set of sensory inputs.
Great care must be taken not to infer that everyone
behaves in the same manner. Some people are far more
sensitive and can compensate for conflicting sensory
cues than others.
In the majority of experiments conducted in presence,
the experimenters do not address the issue of sensory
conflict. It is quite possible that our real-world
experiences which are based on a full set of sensory cues
do not readily map onto our sense of presence in a
sensory deprived computer generated environment. Not
only is our sensory system deprived of certain perceptual
cues there may be sensory conflicts, which arise from
issues such as lags or temporal anomalies in our system.
Our experiences or priori knowledge of real world
systems can greatly influence our internal representation
of a sense of being present in an environment. For
example, test pilots are used to dealing with tasks in a
fixed based (no motion cue) simulator and transfer the
experience to the real world. However, it has been
established that most combat pilots perform better in
simulated missions compared with real battle situations.
Obviously, risks are much easier to take in a simulator
than in the real world. As a converse argument combat
pilots sometimes make different decisions when under
combat stress due to a different level of adrenaline.
Unless people are carefully trained (and it is very
difficult to determine if this can actually be done) then
there is a great danger that the subjective evaluation
techniques may not be sensitive to the same parameters
for each of the experimental participants.
A computer-generated environment can affect the
participants experiences in a very profound way by
allowing events or situations to be experienced that
cannot be achieved in the real world. For instance, it is
easy to transport someone to a different temporal domain
where events can be slowed down or speeded up
compared to real time. In these situations, it is not
practical to try and map this onto a real world
experience. Consequently, researchers should be very
careful when using terms such as low and high presence.
A crude but repeatable measure for presence would be to
count the number of sensory inputs that are missing from
the virtual environment compared with the real
environment (Sheridan, 1992). Unfortunately, even this
approach is flawed because each sensory modality does
not contribute equally to the sense of presence. It is also
likely that individual contributions will change over a
period. For example, it is well known that people can
become desensitised to certain stimuli.
Real World Versus Virtual Environment
There is considerable merit in being able to compare
performance in the real world against performance in a
virtual environment, especially if the virtual environment
is mimicking the real world in some way. This means
that metrics developed for the real world case can be
deployed in the virtual environment. However, this
presumes human performance is the same in real and
virtual environments. This factor is very important for
training applications where a virtual environment is used
to train a particular skill and the skill has to be
transferred into the real world. Skill transfer is a very
important factor but equally human behaviour and
performance in the virtual environment is very
important. Conflicting sensory cues could actually
modify the user’s performance in the virtual
environment in a detrimental or beneficial way. In some
cases the ability to present only a subset of real world
attributes might actually improve the training process.
Pilot training is a good example where basic procedural
tasks can be taught without the trainee having to worry
about flying the aircraft at the same time. These training
systems are known as part task trainers. To avoid many
training transfer issues the part task trainer is made as
real and representative as possible.
It is possible to extend this idea by investigating the
quality of a virtual environment in terms of the tasks to
be undertaken. If we gather data on the user’s
performance in a real environment and the user’s
performance in a virtual environment then we have some
measure of the quality of the virtual interface.
The Quest for Understanding Presence
There has been insufficient research into the causes of
presence to be able to discuss them definitely and
accurately. “There is no scientific body of data and /or
theory delineating the factors that underlie the
phenomenon” (Held and Durlach, 1992). Despite this,
there is a growing quantity of research that is attempting
to derive a single dimension for presence. This research
is based on subjective rating techniques. Zelzter
proposed a description of virtual reality in his AIP cube
(Zelzter, 1994) which sets out to define the components
of a synthetic environment in terms of a co-ordinate
system giving a measure of the quality of the system
across three interacting parameters: autonomy,
interaction and presence. Zelzter’s cube illustrates the
three different axes — defining a co-ordinate system
which can be used as a qualitative measure of virtual
environments. Autonomy is defined as the ability of the
environment to act and react to simulated events.
Interaction is the fidelity with which the environment
deals with interactions between its participants both
human and synthetic. Presence provides a rough
(dimensionless) measure of the number and fidelity of
available input and output channels. Zelzter associated
the position of an application within the cube with task
performance. He indicated that while clearly, an
evaluation of a virtual reality system in these terms is
highly task dependent, every design solution for a virtual
environment can be characterised within these bounds.
At first sight, Zelzter’s AIP cube looks to be a good way
of describing a particular virtual reality system according
to where it fits in the cube. However, it is very difficult
to characterise a virtual reality system in this way
because of the lack of a clear definition for each axis. In
particular, the term presence is very difficult to specify
in a simple way that would fit the AIP cube. It is
tempting to try to classify attributes of a virtual reality
system in this way, and then devise a measurement
process but there is a serious danger that the real
performance controlling factors of a virtual reality
system will not be addressed. Moreover, it is not easy to
justify the use of a dimensionless performance parameter
if it cannot be measured objectively or subjectively
against clearly defined metrics.
To begin to understand where and how to evaluate the
user’s performance when using a virtual environment it
is necessary to look further into the unique properties of
the system. Traditional empirical human factors based
evaluations such as measuring the display resolution of
the system are useful but do not necessarily relate too
well in terms of overall user performance. For example,
it has been shown that performance in a virtual
environment is affected if one of the input modalities is
removed (Pausch, Shackleford and Proffitt, 1993). This
suggests that if we undertake empirical based
evaluations we will not be able to draw too many
conclusions regarding an integrated interface. In this
context, we need to consider the virtual environment
system as an entity and thus treat the system as an
integrated interface.
Traditional human factors evaluation techniques do not
take into account attributes such as presence and greater
interactivity. There have been numerous attempts to
produce a single metric representing the degree of
presence for a virtual environment and then relate this to
some measure of human performance (Slater, Usoh M.
and Steed A., 1994). Research which investigated the
sense of presence (as yet undefined) within virtual
environments as a function of visual display parameters
(Hendrix and Barfield, 1996). The research indicated
that people reported higher levels of presence when head
tracking was used and stereoscopic visual cues were
employed. An increase in field of view also resulted in a
reported increase in the level of presence. In reality,
these finding are no surprise since we routinely use the
cues in the real-world.
There have been many attempts to define a
straightforward definition for presence, largely without
success. Some VR practitioners try to define different
classes of presence such as ego-presence and objectpresence (Hendrix and Barfield, 1996). Indeed, it is
tempting to try and derive a simple measure for the
amount of presence a particular system is able to provide
and then relate this to user task performance.
Unfortunately, this approach is flawed because presence
is a multi-dimensional parameter that is arguably an
umbrella term for many inter-related perceptual and
psychological factors. However, it is clear is that
presence is a cognitive factor that must be treated
differently than other perceptual aspects of a humancomputer interface such as brightness or contrast of an
image. If presence can correlate usefully with
performance and provide the means to achieve effective
communication and control in interface design, (Ellis,
Issues of Evaluation
A virtual interface is radically different compared to
conventional computer interfaces and as such needs
quite a different approach to performance evaluation.
The user’s performance is governed by the environment,
personal capabilities, individual motivation, the tasks to
be performed and the situation under which those tasks
are to be carried out. For instance, if the user is
performing two tasks simultaneously then performance
on a single task might not be the same as when only one
task is being undertaken. Whilst it is possible to perform
empirical experiments to predict human performance
these do not tend to deal with a complex situation where
numerous activities have to be undertaken concurrently.
An empirical understanding of human performance is
important but what is probably more important is an
understanding of the overall user performance. This
seems strange when a virtual environment system has the
potential for so much variability in the design of the
It is easy to overlook that we are dealing with a multisensory interface that can provide auditory, kinaesthetic
and visual displays. One point to bear in mind during the
evaluation process is that the user will experience fewer
sensory cues in a virtual environment than in the real
world. This inevitably means that our knowledge, which
relates to the real world may only partially fit the case
for virtual environments. Indeed, we only need to think
of cues such as motion perception to begin to understand
the complexity of the problem.
The user of a virtual reality system will generally act
inside the environment rather than outside as with other
computer based systems. In many ways the flight
simulator (one form of virtual environment) has similar
attributes. Consequently, a number of interesting human
factors challenges will result. For example, if a user is
performing a task in a virtual environment it is quite
possible that an experimental evaluator (on the outside)
will interfere with the performance of the user.
Fortunately, this type of problem can be overcome by
careful experimental design. In situations where highly
realistic virtual environments are being used, the lack of
certain real or redundant sensory cues may have a
detrimental effect on the user’s performance and
subjective experience. An equally important issue is one
of perceptual conflict where for example, dominant
visual cues may conflict with whole body kinaesthetic
cues. This has been known to be the cause of many
accidents in the aerospace sector where pilots have
tended to believe their own proprioceptive senses rather
than aircraft instrumentation. In some instances it will
not be possible to avoid such complexities, as the
enabling technology will be limited. However, this is
where an understanding of empirical human performance
becomes important. One way of avoiding the difficulties
of human performance evaluation is the development of
an evaluation framework. The framework could help
formalise the whole process and ensure that a consistent
approach is taken.
User Interaction Devices
A review of the human factors issues of VR systems
would not be complete without a discussion on the user
interface. Most VR system users would agree that the
user interface is poor and the input devices are relatively
crude. Even though there are a variety of different input
devices ranging from 3D joysticks to glove like devices
none of these are particularly intuitive. This partly comes
from the need for an effective 3D interface device and
user metaphor. Some tasks will require force feedback
and this area is seriously lacking in terms of the maturity
of the enabling technology.
If the VR system is to be used to support group
interaction then the situation is even more serious
because no group interaction devices exist. All the
interface devices are severely restricted because of the
need for cables and wiring. In the future wireless
interface devices will be required.
The VR System as a Collaborative Tool
Arguably one of the best uses for VR is collaborative
working at remote locations. It is now perfectly feasible
to link two or more VR systems together over computer
networks and establish a common virtual environment
between the remote users. With such a system it possible
for each remote user to interact with the data set and
makes changes that are then reflected to all other
collaborating users. The justification for collaborative
working arises from the complexities of today’s projects,
which tend to be multidisciplinary and involve teams of
people who could work for different organisations.
These organisations could be international and one clear
benefit of the collaborative link up would be the cost and
timesaving compared with travelling to a common
destination. The collaborative VR system would not
solve the time zone differences but would save
considerably on project costs. The collaborative VR
system is very different to a video conferencing system
because it allows the users to interact with the data being
discussed or reviewed.
There is an important issue of scale regarding the
collaborative VR system. On one hand there is the
potential for collaboration involving hundreds of users
whilst on the other it is possible to restrict collaboration
to just two–five people. It is very clear to see how chaos
or confusion could set in if a large number of people are
collaborating together. Figure 21 shows what tends to
occur on the user’s display. There are many
commercially available products that support this type of
interaction over the internet. The usefulness of these
systems has yet to be proven.
One of the first social issues to be addressed is whether a
human form actually needed? However, human forms
communicate certain social expectations. In a team,
working environment is an avatar misleading for a
group/team? One area where an avatar comes into its
own is if an autonomous intelligent agent is implemented
in the virtual environment. Without some form of
intelligent control, the avatar will be crudely driven by
simple user gestures. Obviously, full body suits could be
produced but these are too immature at the moment and
more importantly would anyone want to instrument
themselves up in this way.
Figure 21: Potential Chaos with Large Numbers of
Collaborating Users
There are other available products that enable much
more effective interaction (though this is still limited).
Figure 22 shows the sort of environment that is provided
by Parametric Technologies Corp. (formerly Division)
Human Factors Evaluations in Virtual Environments
Issues of Evaluation
It is important to recognise that a virtual interface is
radically different compared to a conventional computer
interface. This implies that there are likely to be different
human factors issues that need addressing. Moreover, it
is unsafe to assume that evaluation methods that work
for real-world situations may not necessarily work for
synthetic environments.
User’s performance is governed by the following:
• Environment
• Personal capabilities
• Individual motivation
• Tasks to be performed
• Situation under which tasks are to be carried out
A key problem with many human factors based
evaluations is that they are often. If the user is
performing two tasks simultaneously then performance
on a single task might not be the same as if only one task
was being undertaken.
Better to derive a functional or parametric form
representing presence.
Figure 22: Collaboration using dVMockup
There are many human factors issues that arise from the
use of collaborative VR systems. In any collaborative
link up we need to know who we are interacting with. It
is not necessarily a good idea to have a cartoon like
character representing what we are doing in the virtual
environment. Obviously, a question to be addressed is
what does the avatar communicate and how should it be
represented. This raises social implications such as Who am I actually dealing with? As shown in Figure 22
it is possible to introduce a perceptual expectation
mismatch because users can seem to float up in space
and assume all sorts of unusual attitudes. This would not
occur in the real world and whether this actually helps in
the collaboration process has yet to be determined. Even
so it is still reasonable to assume laws of physics hold
true since this facilitates our interaction in the
Rather than go into specific evaluation issues here the
interested reader is recommended to obtain the following
paper (Kalawsky, Bee and Nee, 1999).
VR systems have technological limitations which are
slowly being overcome. However, our understanding of
the human factors issues is seriously lacking. It is not
simply a question of understanding how we perform in
the real world and simply mapping this onto the virtual
environment. As this paper has reported, we have to
understand human performance in the context of sensory
conflict and misrepresentation. From research undertaken it has been established that our behaviour and
performance does not necessarily relate to that which
would be achieved in the real world under the same task
situations. This need not be a major problem because as
we begin to understand human interactions we may be
able to exploit more fully the unique properties of the
virtual environment. Moreover, as our knowledge
increases we should be able to impose more definitive
requirements on the development of the enabling
technologies and so ensure that they are more suitable
for the task in hand.
Caudell, T.P., (1994), Introduction to Augmented
Reality, Proceedings of SPIE Proceedings
Telemanipulator and Telepresence Technologies.
Ellis, S.R., (1996), Presence of Mind: A Reaction to
Thomas Sheridan's "Further Musings on the
Psychophysics of Presence", Presence, vol. 5,
Furness, T.A., III., (1986), The Super Cockpit and its
human factors challenges, Proceedings of
Proceedings of the 30th Annual Meeting of the
Human Factors Society; pp.48–52.
Gibson, J.J., (1986),The Ecological Approach to Visual
Perception. Hillsdale, New Jersey: Lawrence
Erlbaum Associates.
Gilkey, R.H., (1995), The Sense of Presence for the
Suddenly Deafened Adult, Presence, vol. 4, pp. 357363.
Gillingham, K.K., (1992), The spatial disorientation
problem in the United States Air Force, Journal of
Vestibular Research, vol. 2, pp. 297-306.
Graybiel, A. and Niven, J.I., (1951), The effect of a
change in direction of resultant force on sound
localisation: The audiogravic illusion, Journal of
Experimental Psychology, vol. 42, pp. 227-230.
Heeter, C., (1992), Being There: The subjective
experience of presence, Presence, vol. 1, pp. 262271.
Held, R.M. and Durlach, N.I., (1992), Telepresence,
Presence, vol. 1.
Hendrix, C. and Barfield, W., (1996), Presence with
Virtual Environments as a Function of Visual
Display Parameters, Presence, vol. 5, pp. 274-289.
Kalawsky, R.S., 12 Oct, Integrated Real-Virtual World
Display System, Patent Appl.9121707.5,,
Kalawsky, R.S., 24 Apr, Integrated real-world virtual
world system, Patent Appl.US 07/959;919, USA,
Kalawsky, R.S., (1993a), The Science and Engineering
of Virtual Reality, Proceedings of Virtual Reality
International '93,, London.
Kalawsky, R.S., (1993b),The Science of Virtual Reality
and Virtual Environments: Addison-WesleyLongman.
Kalawsky, R.S., (1996), Exploiting Virtual Reality
Technological Issues, ISBN 1356-5370,.
Kalawsky, R.S., Bee, S.T., and Nee, S.P., (1999),
Human Factors Evaluation Techniques to Aid
Understanding of Virtual Interfaces, BT Technology
Journal, vol. 17, pp. 128-141.
Knowlton, K.C., (1977), Computer displays optically
superimposed on input devices, Bell System
Technical Journa, vol. 56, pp. 367-83.
Krueger, M., (1985) VIDEOPLACE: A report from the
Artificial Reality Laboratory, in Leonardo, vol. 18.
Pausch, R., Shackleford, M.A., and Proffitt, D., (1993),
A User Study Comparing Head-Mounted and
Stationary Displays, Proceedings of IEEE
Symposium on Research Frontiers In Virtual Reality.
Sheridan, T.B., (1992), Musings on Telepresence and
Virtual Presence, Presence, vol. 1, pp. 120-25.
Sherrington, C.S., (1920),Integrative action of the
nervous system. New Haven: Yale University Press.
Slater, M., Usoh M., and Steed A., (1994), Depth of
presence in virtual environments, Presence, vol. 3,
pp. 130-144.
Sutherland, I., (1965), The Ultimate Display,
Proceedings of IFIP Congress;pp:506-508.
Zachary, W., Ryder, J., Hicinbothom, J., & Bracken, K.,
(1997), The Use of Executable Cognitive Models in
Simulation-based Intelligent Embedded Training,
Proceedings of Proceedings of the Human Factors
and Ergonomics Society 41st Annual Meeting.
Zelzter, D., (1994), Autonomy, Interaction and Presence,
Presence, vol. 1, pp. 127-132.
A Virtual Environment for Naval Flight Deck Operations Training
Lt. Edward A. Trott
Dr. Venkat V.S.S. Sastry1
Mr. Joseph Steel
Applied Mathematics & Operational Defence Procurement Agency
Research Group
Abbey Wood
Cranfield University
RMCS Shrivenham
Wilts SN6 8LA, UK
The main aim of this paper is to develop a prototype
virtual environment for training Flight Deck Officers
with a view to study the types of interactions required in
such an environment. The application is ideally suited to
exploit techniques based on proprioception, in particular
the trainee’s arm signals.
1. Introduction
A virtual environment is a synthetic sensory experience
that communicates physical and abstract components to
a human operator or participant (Kalawsky, 1993).
Virtual Environments (VE) offer greater potential to
enhance the communication between the human and the
computer as they offer most intuitive and natural
interfaces. They have been exploited in diverse
applications ranging from medicine to training soldiers.
Their potential is far more evident in training
applications (Nemire, 1998) for the enriched interaction
styles such environments support. A typical VE
synthesises one or more sensory inputs to facilitate a
particular user’s task. Exploiting proprioception (sensory
awareness of parts of the body) enhances interaction in a
virtual environment (Mine, 1997). One form of
proprioception is the use of body-relative actions called
gestures to issue commands to alter the environment.
Current research work in this area includes two-handed
input (Hand, 1997) and gesture-based interaction (Mapes
& Moshell, 1995). The gestures involved in the present
application are quite unique, and thus provide an ideal
test bed for exploring 3D interactions in VE.
Training Flight Deck Officers (FDO) is an important
aspect of Naval operations and currently uses a range of
traditional teaching material augmented by instructorassisted scenario generation (AIR 230 Course). The
instructor directly controls the scenario presented to the
trainee. While this approach has some strength, we
believe that a virtual environment offers much
significant benefits and that the application readily lends
itself to exploit the natural interaction styles, such as arm
signals, that are inherent in the training of Flight Deck
Operations (Trott, 1999). At the same time, the
application raises several challenges. The main purpose
of this article is to present the results of our initial
prototype, with a view to enhance the model. The
development of the application is by no means complete,
and should be treated as an initial investigation.
The paper is organised as follows. A brief motivation is
presented in Section 2. The problem we are attempting to
address in this paper is described in Section 3, together
with typical training scenarios. The details of developing
the Virtual Environment are given in Section 4. Some of
the points highlighted in developing the prototype are
summarised in Section 5.
2. Rationale
Exploiting natural interaction metaphors offered by the
application can enhance the current set up for supporting
the training of Flight Deck Officers. The main
motivation for the present study can be summarised as
• To provide an enhanced training environment for the
• To allow interactions using natural metaphors that
will enhance the experience of a flight deck officer in
which he/she will be able to control the environment
in response to his/her actions. For example, ask the
helicopter to move to the next gate position in
response to an arm signal.
3. The Problem
Current Practice
Currently the British Royal Navy Flight Deck Officers
(FDO) are trained at RNAS Culdrose, Cornwall,
England. Their training makes extensive use of real
simulation, that is real people using real equipment. In
this case the real equipment is an actual helicopter,
although training is not performed on-board ship.
If the weather conditions restrict aircraft flights or
aircraft are unavailable, then the training makes use of a
virtual simulator. This consists of three large projection
screens that display images from three front projectors
driven by three networked PCs. The system shows the
view as seen from the landing deck of a frigate. The
trainee stands in front of the screens and directs the flight
of the simulated helicopter using the appropriate signals.
The class instructor who is sitting behind the trainee flies
For correspondence with author: [email protected]; tel. +44 (0) 1793 785315; fax +44 (0) 1793 784196.
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
the helicopter. The direction of view of the system is
fixed and cannot take into account the direction of view
of the trainee. The current system also has limited
graphics capability and environmental effects such as
reduced visibility, fog and variable sea-state are barely
Although the virtual simulator is available, final
examinations must be passed using the real equipment.
There are two main reasons for this: the virtual simulator
is unable to replicate the feel of being exposed to the
prevailing weather conditions nor the feel of the
proximity of the helicopter as it lands.
Projection Screens
Audi speakers
Audi speakers
Figure 2: A typical approach of an aircraft. Helicopter
begins approach relative angle 165° from ships head
(1); Aircraft reaches gate (2); Aircraft alongside flight
deck directly of the ‘bum’ line (3); Traverses across
flight deck maintaining its hover (4); aircraft descends
to flight deck (5).
Student area
Flight Deck
Captain’s Area
display screen
Captain’s display
Sound Mixer
Instructor’s Control Station
Figure 1: Flight Deck Operations Simulator
Some of the limitations cited above can be addressed by
developing a virtual environment, which offers far
greater potential. To explore these possibilities, a subset
of training scenarios are considered for the initial design,
which are explained in the next section.
Example Scenarios
Flight Deck Officer’s training consists of a number of
scenarios including Landing and Takeoff (varying angles
of approach), Rotors Running Refuel, Helicopter Inflight Refuel, Weapons Loading, Personnel Transfer, and
Helicopter Shut Down and Start Up using either a Sea
King/Lynx. All these scenarios offer a rich variety of 3D
interactions that enhance the learning and training
experience. For the purposes of this investigation, we
have selected two scenarios — Helicopter landing and
take off under varying environmental conditions.
The trainee FDO immersed in the virtual environment
makes an assessment of the wind speed and direction
(not currently implemented) and then signals the aircraft
when he is ready to receive it. It is assumed that the
aircraft is out of radar-controlled approach and is within
the visual range for FDO to take control. On receipt of
the signal, the aircraft moves to its next waypoint or
gateway. A typical approach of an aircraft to the ship’s
flight deck is shown in Figure 2.
4. Development of a Virtual Environment
The virtual environment consists of a visual model of the
flight deck and its associated dynamics, a visual model
of a helicopter (Sea King) and its associated dynamics,
and finally a visual representation of the flight deck
officer including body articulation (limited to hands).
For the training purposes, few environmental effects
such as fog and night time are also included. These are
discussed in detail in the following sections.
Flight Deck Officer
A simple model of a mannequin is used to represent the
flight deck officer. The body articulation is limited to
arms only. Currently there are nearly 56 distinct gestures
used by the FDO. Of these, 9 gestures (See Figures 3 and
4) which are directly relevant for launch and recovery
operations are chosen for the prototype. An existing
model of a man from the Division software library has
been modified to facilitate the emulation of arm signals.
The animation of the limbs is achieved using the
keyframes animation technique where each frame
describes a particular state of the object, for example its
position and orientation. Each hand signal is stored as a
key frame animation sequence and was stored in a
separate library.
The FDO is required to carry lighted wands at night in
order to make his signals visible to the pilot.
Accordingly, our virtual FDO has two lighted wands,
which come into effect during night
time training.
Helicopter Approach
A 3D model of the Sea King helicopter is modified to
provide realistic rotor disc motion and the addition of
navigation lights to the helicopter stub wings. This was
achieved by mounting a spotlight just in front of the
navigation light and setting an appropriate object
luminescence property in the object’s texture file.
The movement of the helicopter is governed by a series
of keyframe sequences in a required direction. The
helicopter movement is triggered by an event (such as
receive, approach, move away, left, right, up, down,
hold, wave off) raised by FDOs hand signal. In response
to this signal, the aircraft will move to the next ‘gate’
position. Once the aircraft has been successfully directed
over the deck, a final signal to descend to the deck is
given. When the collision between the helicopter and the
deck is detected, the helicopter object is parented with
the deck object, so that the helicopter moves in
accordance with the motion of the flight deck and that of
the ship.
Platform Dynamics
The platform consists of the deck, harpoon grid and a
model of the FDO standing on the deck. To keep the
frame rates to a minimum, a simple animation sequence
is created for the platform that emulates the motion of a
Environmental Effects
Environmental effects such as lighting conditions and
visibility effects can be easily incorporated into the
virtual environment. To effectively light the scene and
allow for a number of different lighting conditions five
light sources are used. Four of these light sources are
used for scenery lighting and the remaining light is used
to illuminate the deck. An appropriate texture is applied
to the sky to produce an impression of a marginally
cloudy day. Fog is emulated using the library function
dvFog which allows a colour and distance parameter
(beyond which the objects are invisible) to be specified.
Note that when fog is enabled, the sky is obscured; and
affects the intervisibility computations.
Virtual Command Interface
The prototype environment is developed using the
Division software dVS/dVISE. Due to the current
limitations, the arm signals of the trainee are simulated
using virtual menus (See Figure 5). A limited number of
training manoeuvres is implemented for helicopter
landing and takeoff under different environmental
conditions. The use of directional sound is also explored
with limited success. Note that for a fully functional
immersive environment, appropriate hardware, and
additional software for gesture analysis should replace
the virtual menus mentioned above.
5. Remarks
As the main focus in this study on developing a virtual
environment for training, no specific experiments were
conducted to evaluate the overall benefit of such a
system. However, the exercise has revealed several
important factors.
Top most in the list is the need for a software component
for gesture interpretation. The computational demands
for training purposes are moderate, as the objects in the
virtual environment remain fairly static. Selection of
items using Virtual menus is not intuitive, and could be
enhanced using additional visual cues.
The next stage of the work is an investigation into the
recognition of various arm signals using a single tracking
device in each hand. Simply knowing where each hand
and it’s orientation is insufficient. It is expected that
knowledge of how each hand has recently moved will be
required to determine the relevant signal. For example,
arm signals for FDO Up and FDO Down (See Figure 4)
trace the same path, but differ in start and finish
positions. We envisage that it will not be necessary to
have tracking devices at either the elbows or shoulders.
The use of a neural network to facilitate this task is
6. Conclusions
The prototype environment for FDO training has
highlighted some of the requirements that are essential
for a fully immersive tool. There is a clear need to track
the position of both head and two arms. While the
current tracking system is capable of tracking position
data up to four trackers, this is not currently
implemented, and will be pursued in a future
investigation. Use of directional audio cues have been
explored very briefly, but needs a detailed study.
AIR230 Course. Royal Naval School of Flight Deck
Operations. See also Windows User Guide.
Hand, C. (1997). A Survey of 3D Interaction
Techniques. Computer Graphics Forum, 16 (5), 269–
Kalawsky, R.S. (1993). From visually coupled systems
to virtual reality. An aerospace perspective.
Proceedings Computer Graphics, 91.
Nemire, K. (1998). Individual combatant simulator for
tactics training and mission rehearsal. SPIE
Proceedings, The Engineering Reality of Virtual
Reality, 3295, 435–441.
Mapes, D. & Moshell, J.A. (1995). A Two-Handed
Interface for Object Manipulation in Virtual
Environments. Presence: Teleoperators and Virtual
Environments, 4 (4), 403–416.
Mine, M. (1997). Exploiting Proprioception in Virtual
Environment Interaction. PhD Thesis, University of
North Carolina, Chapel Hill, Computer Science
Technical Report, TR97-014.
Trott, E.A. (1999). An Investigation into the use of
Virtual Reality for Naval Flight Deck Operations
Training. MSc Project report, Cranfield University,
Royal Military College of Science, Shrivenham.
FDO Ready to receive – you are cleared to land
FDO Approach
FDO Move left
FDO Move right
Figure 3: A subset of Flight Deck Officer’s hand signals
FDO Down
FDO Wave-Off
FDO Hold
Figure 4: A subset of Flight Deck Officer’s arm signals (ctd.)
Figure 5: Virtual menues used in the virtual environment
Mission Debriefing System
Major Birger I. Johansen1 and Bo Fredborg MSc EE
Senior Systems Consultant and Systems Engineer
Systematic Software Engineering A/S
Søren Frichs Vej 38K
8230 Åbyhøj
Systematic has developed a debriefing system for
aircraft crews to improve their skills based on
experiences from completed missions. The system is
developed on Commercial Off The Shelf (COTS)
software and on a PC. The panel should see this input as
a portable, low-cost Virtual Reality (VR) training system
for aircraft crews. The benefit of the portability is that
the system can be used anywhere the unit is deployed
and by any crewmember.
Flight hours are rather expensive and therefore the air
forces must maximise the benefits from spent flight
hours. This, combined with the fact that most air force
units need to operate from different deployments remote
from home bases, led the operational fighter squadrons
to express a need for a low-cost debriefing system.
The users were directly involved in the design and the
focus was set on functionality — not technology. This
approach has resulted in a system which gains accept
among users and therefore becomes an everyday training
tool. Driven by user requirements, the system is
developed to run on a Microsoft Windows 2000
platform, and the system can interface with other
systems. Furthermore, it has been essential to develop a
system, which could be rapidly implemented.
The debriefing system uses already existing information
from the aircraft. The aircraft is equipped with Global
Positioning System (GPS), three video cameras, and a
microphone system to record the pilot’s voice
communication. The video cameras record the pilot’s
view through his head-up display and the entire
instrument panel.
Prior to the debriefing session all information from the
aircraft (GPS-data, video- and audio recordings) is fed
into the debriefing system. The GPS-data is loaded into a
three dimensional (3D) model containing geographical
information, the video and audio recordings are
digitised, and all data are synchronised. On each
monitor, four visual sources can be displayed
concurrently, e.g. video recordings from three different
aircraft and the graphical 3D view of the area, including
aircraft. The selected visual sources are displayed along
with a selected audio recording. The 3D graphic makes it
possible to see and follow selected aircraft from different
perspectives on their mission. Furthermore, it is possible
to see them chase other aircraft and to track their route
by position, direction, and speed. The crew and other
mission participants can by themselves prepare and
execute the debriefing session.
Systematic has developed a portable, low-cost VR
training system for aircraft crews, which converts reality
to virtual reality, reflecting the reality. The chosen
approach, with heavy user involvement, has resulted in a
system, which is easy to use and will gain much better
acceptance. A system based on well-proven COTS
products reduces costs as well as risks. Finally, the
system gives added value to the flight hours spent.
It is our aim with this paper to disseminate
understanding for the possibilities given by new
commercial off the shelf (COTS) products — in this case
especially for low cost virtual reality training tools. We
find that today’s COTS software fulfils most of the
requirements that the military has to an everyday
debriefing system. By combining the COTS products
using Systematic’s competence in software integration, a
low-cost easy-to-operate operational training system, has
been developed.
Through this paper, we discuss functional requirements,
use of commercial state of the art technology, influence
on training and human performance requirements, and
describe the development process and functionality in
our debriefing system.
In connection with the training of combat pilots much
time is spent on manoeuvres in actual air combat
techniques. The Danish Air Force spends more than half
of the flight hours on such manoeuvres. Furthermore, the
remaining flight hours often contains elements of air
combat. It is therefore essential to get full benefit from
the training, especially as flight hours are extremely
costly. Nevertheless, subsequent debriefing and evaluation of a training session is often deficient or nonexisting.
The present project has endeavoured to remedy this
inadequacy by investigating the possibilities for building
an inexpensive, simple and user-friendly, but yet hightech, mission debriefing system, for “everyday use”. We
have used virtual reality (VR) and 3D techniques for
For correspondence with author: [email protected], tel. +45 89 43 21 53; with company: tel. +45 89 43 20 00, fax +45 89 43 20 20
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
constructing factual conditions for training in a Virtual
Environment (VE). The VE facilitates the debriefing of
pilots and thereby enhances the learning. Presently the
system is developed as a 1. generation version with basic
functionality financed by our company. We find though,
that the idea has much potential, and we will promote
our ideas broadly within NATO.
Security in the Euro-Atlantic area has substantially
improved during the 1990s, by comparison with the four
decades that preceded them. The threat of massive
military confrontation has gone, and co-operative
approaches to security have replaced former
confrontation. Nevertheless potential risks to security
from instability or tension still exist.
In these changed circumstances affecting Europe’s
security, NATO forces have been adapted to the new
strategic environment and have become smaller and
more flexible. Conventional forces have been
substantially reduced and in most cases so has their level
of readiness. They have also been made more mobile, to
enable them to react to a wider range of contingencies;
and they have been reorganised to ensure that they have
the flexibility to contribute to crisis management and to
enable them to be built up, if necessary, for the purposes
of defence. Increased emphasis has been given to the
role of multinational forces within NATO’s integrated
military structure. Many such measures have been
implemented. Others are being introduced as the process
of adaptation continues.
Airforces are characterised by their ability to operate
from far distance, geographically dispersed bases and
concentrate their efforts against the main targets. They
are also able to react very fast and to maintain a high
degree of readiness. These characteristics have made
airforces even more important to NATO’s new strategic
concept, Combined Joint Task Forces (CJTF). The main
issue in this concept regarding air forces is high
readiness, interoperability and the ability to operate away
from home bases with a minimum of preparations.
Furthermore each participating unit must be able to
perform a larger variety of roles, than before — e.g.
using heavy bombers for close air support. The
operational environment has become much more
dynamic — it is never possible to foresee which type of
operation that will turn up. This again puts higher
demands on continuous and flexible training.
Another consequence of the new operational
environment is the reduced military budgets. This means
that it is essential to gain as much as possible from the
applied training efforts. In real operations like Allied
Force in Kosovo last year, it is extremely important that
the pilots learn from each mission to make continuos
improvements. In this specific example, most of the
participating units operated far away from home bases.
Therefore they were not able to take advantage of their
usual static training equipment and simulators.
Project Objectives and Means
An obvious need for mobile training, rehearsal and
debriefing systems has evolved. Given the fast
development within virtual reality technology and low
cost flight simulators for PCs, we have seen a good
opportunity to use commercial technology and existing
sensors, video recordings, and tapes from the aircraft to
develop a debriefing system for air force pilots.
The overall objective was to create a low cost, easy-tooperate, and transportable debriefing system. With this
objective the intention was that each training session and
live mission should be followed up by a high-quality
debriefing activity, giving full benefit of the costly flying
time to the pilots.
The aim was to base the system on COTS products and
existing and electronically available data from the
aircraft. Furthermore, the aim was to combine the
collected data from the aircraft and thereby constitute a
3D Virtual Reality (VR) replay of the completed
missions and training sessions.
The project is financed by Systematic and the ingredients
used are Systematic’s skills and knowledge,
technologically as well as military, a range of COTS
products, and the requirements set by airforce pilots.
System requirements and functionality
This section describes the scenarios and missions that are
supported by the debriefing system. To stress out the
need for a debriefing system, we give a brief description
of the main categories of existing systems.
Air Combat Manoeuvring (ACM)
Air combat comprises all kinds of manoeuvres in the air
in a one-to-one, many-to-one or many-to-many situation.
ACM includes all the classical movement patterns such
as half loop, full loop, split S, break turn etc. Basically,
air combat is a question of gaining the right position in
relation to the opponent.
A training session consists of a number of scenarios,
ranging from 3 to 10 — depending on the number of
fighters involved. A scenario lasts from 5 to 10 minutes.
The starting point of a scenario is an initial position
where, for example, the different players have got radar
contact (approx. 30NM distance). Typically, the
situation then develops rapidly, depending on the actions
that take place during the session. After only a few
minutes, the situation typically becomes very complex
and the pilots often lose control of the situation. As an
example, a pilot who tries to escape will lose control of
what is going on, as he has no longer radar contact with
the other fighters.
When a training session is over, the pilots involved
should evaluate the session. This is typically a difficult
process, partly because the individual sessions develop
in a complex way where each pilot may have different
opinions on what actually happened, if they are able to
contribute to the situation at all. But also because the
individual scenarios become indistinguishable when the
pilots have returned to the air base. As a result hereof,
debriefing is deficient or non-existing. Consequently,
much value of the training is lost. This should be viewed
against the large resources spent on keeping the fighters
in the air.
Existing Solutions (ACM)
In order to enhance the debriefing possibilities, various
systems are available for the pilots for recreating the
individual scenarios that constituted the training.
Generally speaking two solutions exists: A low-cost and
an expensive solution.
Low-Cost Solution: Video
The F-16 fighters used by the Danish Air Force are
equipped with three standard video cameras, which
records the Head Up Display (HUD) and the two Multi
Function Displays (MFD). The pilots can use these
videos in a subsequent debriefing. Videos are excellent
for the initial scenario and evaluation of shootings. In a
debriefing situation, the pilots involved will endeavour
to recreate the individual scenarios in the training
session. If the pilot has lost control, however, videos are
of little use (the radar image may be of no value).
Furthermore, it is difficult and time-consuming to
synchronise multiple videotapes and ECM as well as kill
removal are not covered by video at all.
The Expensive Solution: Real-time ACM Instrumentation
Real-time ACMI covers the expensive and extensive
solution where the individual fighters that participate in
the session downlink information in real-time to a
control station on the ground. Via the control station, the
individual scenarios are monitored and stored for later
debriefing. The control station may even intervene
during the training session, either in order to influence
the situation in a certain direction or due to kill removal.
Real-time ACMI involves pod-mounted electronics
(GPS, MUX-BUS interface and data link) as well as
antenna coverage on the ground and all control facilities
on the ground. Consequently, the solution is quite costly
in terms of electronic equipment and staffing, and ACMI
will not become a natural part of every training session.
ACMI must be planned a long time in advance and will
only be used few times a year.
The first generation of the debriefing system is an
autonomous system and does not require any changes in
the cockpit or instrumentation of the aircraft. The system
is centred on a debriefing facility, based to the greatest
possible extent on COTS hardware and software.
The debriefing system uses already existing information
from the aircraft, the Global Positioning System (GPS)
data, the three videos (HUD and 2xMFD), and a
recording of pilots’ voice communication.
The HUD, MFD, voice recording, and GPS data of the
individual aircraft are loaded into a Personal Computer
(PC), which synchronises the data. From the
synchronised data the PC constructs a 2D/3D synthetic
world of “what happened”.
The three videotapes and the voice recording are used to
give a detailed image of the pilots’ actions, displaying
what happened inside the cockpits. The GPS data from
all aircraft are loaded into a 3D model of the battle cube.
The 3D model does, just like a Geographical Information
System (GIS), contain a 3D graphical model of the
landscape in the battle cube. This 3D model of the
landscape combined with the aircraft GPS data gives a
“Gods eye view” of the battle cube. The debriefing
system makes it possible to navigate around in the battle
cube. This makes it possible to view the scenery from
different perspectives.
All aircraft that can provide the information described
above can be included in the debriefing session.
Consequently, the system can be used not only by the
Royal Danish Air Force’s F-16 fighter pilots.
Furthermore, a debriefing system like this can be used
independently of the geographical location and extension
of the individual training sessions. Compared with the
real-time ACMI system, this provides an obvious
advantage; the real-time ACMI system is not mobile, but
limited to the location that is covered by the antenna
equipment of the ground station.
The latest techniques in Virtual Reality and 3D have
been investigated in connection with the construction of
the synthetic world. These areas undergo extensive
research and development within the experimenting field
of computer science, and are consequently considered to
contain some of the building blocks for the future
development within HCI (Human Computer Interaction).
The debriefing system includes leading edge
technologies within these fields. It is our aim to present a
system that will delight and motivate the pilots to carry
out high-quality debriefing.
Development of the debriefing system
Systematic’s mission debriefing system
Based on informal discussions with both pilots from Air
Station Ålborg and the Danish Air Materiel Command,
we have developed a first generation model to show the
This section is a brief description of our approach to the
project. Based on our interviews with potential users, a
retrieval of user requirements and a study of existing
COTS products and their facilities, we started the
development process. Knowing that we had to do with
new technology, it was essential for us to study and
develop small prototypes of the different functionalities
in the system. We decided to break the system into three
main subsystems, which were to be developed and tested
sequentially. The initial aims were:
• To see if we could develop 3D graphics using
“cheap” COTS technology and already available
• To test the different 3D graphical components/effects
that we wanted to make use of.
• To establish a 3D-terrain model, which was suitable
for debriefing purposes.
• To establish a user-friendly interface and the
framework from which the debriefing application
should be prepared and presented.
These were the results we got from our first
• Functionality to convert the database from the F-16
MLU simulator to “PC-format”.
• A prototype application showing a landscape of size
10 x 10 NM.
• Playback of flights. (Specifically two flights flying
different routes.)
• Possibility to see the flights in a follow-mode (seen
from one of the flights or in a “God’s-eye-view”).
• Portability between PC and SGI (holobench).
• Possibility to run the application on a PC with a
powerful graphics card.
In the following text we describe each of the initial
prototypes, it’s purpose, the method used to develop it
and the result/experiences gained.
Prototype 2
The next step was to develop an application that could
visualise a complete geographical database and to make
3D movement through the landscape.
Prototype 1
This part resulted in a 3D-terrain model with a
visualisation of a number of aircraft including their
tracks, so that one can get an overall view of the full
mission or extracted parts of a mission or flight.
• To create functionality to visualise a complete
geographical database covering a normal theatre of
• To get a 3D-terrain model and to show it on a PC
(We decided to get the necessary data from the
Danish F-16 simulator).
• To make a 3D visualisation of aircraft (including
their historical tracks).
• To create lively navigation and animation methods
(the aircraft should be able to manoeuvre and
navigate in a realistic way so that an aircraft would
bank naturally when turning and so forth).
• To enable the user to choose between different angles
of view (e.g. “God’s-eye-view”).
• To visualise other objects (e.g. Surface to Air Missile
sites with threat domes).
• Portability: To be able to port the system between the
normal PC platform and a more static SGI graphical
supercomputer with holobench.
In brief we have had a very open and innovative
approach where following main activities were carried
• Information search on the Internet to get components
and pieces of code, which could be useful.
• Selection of a portable visualisation core component.
(Optimizer™ from Silicon Graphics).
• Courses in the use of Optimizer™.
• Prototyping and test using visualisation methods and
• Get inspiration through the studies of existing ACMI
• Initial development on PC — later ported to SGI.
• Use experiences from prototype 1.
• To develop and implement efficient methods to get
and drop tiles of terrain in the visible area.
• To convert the F-16 MLU simulator database to PCformat”.
• A prototype 2 application with functionality, which
in principle (if terrain data is available) can show any
given terrain.
• Geographical data enabling the system to cover
Denmark and Southern Norway.
• This prototype was only developed for a PC.
Prototype 3
The third prototype is the set-up and administration tool,
developed on a Microsoft Outlook user interface.
• To obtain functionality to administrate flights and
missions. (A flight is an operation/flying session
performed by one aircraft and a mission is a
combination of concurrent flights).
• To be able to perform video playback.
• To synchronise video inputs and the 3D-terrain
• To present a graphical user interface (GUI) for
debriefing and administration in a Microsoft Outlook
• Standard components were to be used
− Standard Template Library (STL) from Silicon
− Microsoft Foundation Classes (MFC) from
− Windows media standard components for video
− Microsoft Access Database.
• Use of simple application development (Visual C++,
• Use of well-known components for the GUI
(Microsoft Outlook).
• A quad-view (four concurrent views on same screen)
with an intuitive timeline that permits playback,
review, pause etc.
• An intuitive, easy-to-learn GUI.
• Use of Windows standard functionality to
synchronise video and data.
Integration to a first generation model
Before integration of the three prototypes into the first
generation of debriefing system, we had to solve some
minor problems that occurred during test of prototypes:
• Geographical data are extensive and requires a
harddisk of at least 1GB for the database. We
improved our hardware to the necessary level.
• Movements through the 3D terrain require loading
and initialising of huge amounts of data. Therefore a
dual processor system and a very fast harddisk must
be used to give video and other resources enough
processing power.
• Using video and 3D-graphics in the same session
creates performance problems — Windows 2000
combined with multiple graphic cards solves the
Once these problems were solved we were able to load
the real, digitised video from F-16 aircraft and through
prototype 3 we could initiate, administrate and run the
debriefing system with the introduced functionality. The
result is promising and after some pre-tests with real
users and the necessary adjustments and improvements
in functionality a flexible system is ready to be
implemented with operational fighter squadrons.
Use of Systematic’s debriefing system
Using the Systematic debriefing system is a 3-step
• Digitisation of source data
• Mission/flight set-up
• Debriefing
These processes are described in the following.
Digitisation of source data
After completing a flight, data collected from the plane
must be converted to formats suitable for computer
processing. Analogue data must be digitised and stored
in appropriate formats.
• Video. Video recordings from the HUD and MED’s
must be digitised and converted to “mpg1” format.
• Discrete flight path information. Flight path
information consisting of at least position (time,
latitude, longitude, height) and optionally orientation
(heading, pitch, yaw).
• Event registrations. Identification of events that
occurred during the flight. These could be: Weaponrelease, radar lock-on etc.
• Environment. Stationary and moving objects which
give important input to the flight debriefing. This
could for example be location of a SAM site.
Mission/flight Set-up
The purpose of the mission/flight set-up phase is to
arrange the source data into logical units such as flights
and missions. For example, a flight is a container for all
data relating to a flight including; name of the pilot,
identification of the plane, the videos recorded from the
plane and the flight path data from the plane.
Data is arranged in a hierarchical structure:
Figure 1: Data structure
The different types of data/files should be read as
• Mission. A mission defines a collection of related
flights. A debriefing typically involves several
• Flight. A flight defines the pilot, the plane, a set
videos recorded from the plane and flight path
recording from the plane.
• Pilot. Defines the characteristics of a pilot
• Plane. Defines the characteristics of a plane/aircraft.
Aircraft type/model, visual representation
Video. Defines video recording from a plane (related
to a plane). Includes start and stop time for the video.
Data. Defines flight paths.
Computed “annotations”:
− Speed, Height
− Distance
− Radar coverage
Information layers (on/off toggles)
− Flights
A debriefing is concentrated around a mission.
The screen is divided into five sections as displayed in
Figure 2. Four of the sections are dedicated to displaying
video and/or the 3D synthetic environment. The
remaining section is dedicated to the timeline and
playback controls.
Combining commercial off the shelf (COTS)
technology with military requirements
Figure 2: Division of screen in debriefing mode
The user can make use of following functionality:
• 3D syntetic environment:
− God’s-eye view
− Follow mode
− Free movement
• Video-playback:
− On/off
− Sound
• Play-back control:
− Play
− Fast forward
− Reverse
− Slow-motion
− Single step
− Search (time, event)
• Pop-up time based annotations/Attachments on:
− Data
− Audio/Video
− Flight
− Mission
To reduce cost and improve the usability and learning
process, Through studies of a range of commercially
available products, we have experienced that today’s
COTS products basically cover all given requirements to
a debriefing system.
COTS Hardware
The PC market, driven by the requirements set by the
entertainment industries “need” to produce more and
more realistic games, produces high-performance
entertainment PCs are capable of delivering the high
performance in the areas essential to 3D graphics and
video applications. The essential areas are:
• Processing power — Fast processors are required to
handle movements through the 3D-terrain model.
Multiple processors are recommended.
• Main storage — Memory (RAM) is essential to store
the 3D-terrain in use.
• Mass storage — Harddisk space is needed to store
digitised videos and 3D synthetic terrain. Today
mainstream harddisks are both fast and has large
• 3D graphics — A 3D accelerated graphics adapter is
essential to produce 3D synthetic environments at
suitable resolution and frame rates. The
entertainment industry drives the need for 3D
graphics performance. Current and next generation
consumer 3D graphics systems are powerful enough
to drive the 3D synthetic environment.
COTS Software
We have found that most of the necessary software for
the debriefing system is available in different COTS
products, which can be acquired within a reasonable
price or directly downloaded from the Internet. By using
these products we also make it easier for the user to learn
to use the system. We decided early in the project to use
Windows 2000 instead of Windows NT. The reason for
this is that Windows 2000 can handle concurrent use of
video and 3D-graphics.
We have spent many hours searching for relevant
software products on the Internet and other places. We
have certainly gained benefit from these efforts.
Generally speaking there is COTS technology available
— especially from the entertainment industry — to
support and develop a range of high-tech, virtual reality
training systems. Our task has almost been reduced to
integration of already well-proven and tested blocks of
software code. However, it must be stressed out that the
main challenge was to make the individual products
work together.
could be investigated and evaluated. One of the
problems of the Navy in air combat is finding the
optimal defence process, and the debriefing system may
turn out to be useful.
The debriefing system has great extension possibilities.
As an example, the air force’s simulators could use the
debriefing system for evaluation of the simulation
training. By doing so, simulation and use of the
debriefing system will become an integrated part of the
general simulator training. Consequently, the possibility
for evaluating “what if” situations (situations where a
training scenario is evaluated against new actions) would
become a reality. An existing training scenario that has
been practised and debriefed in the debriefing system
could provide input to the simulator. The simulator could
then fly with the scenario, and what-if situations could
be simulated in order to evaluate the effect.
The interaction with other ground systems, such and C3
and Mission Planning Systems, are further areas to look
into. As an example, the debriefing system could be used
to build an Airspace Co-ordination Order (ACO): With a
“magic wand” the operator could guide and virtually
draw a route through the 3D landscape. An F-16 fighter
could then use the ACO generated. When the mission is
completed, the route planned and carried out could be
compared in the debriefing system.
Another opportunity would be to investigate the
debriefing facility in an interaction with other armed
forces. As an example, the Navy’s air combat system
The F-16 is equipped with a MUX-BUS interface.
Through this interface much more information, e.g.
weapon-release can be accessed. Recording these
information and successively replay during the
debriefing will give a much more detailed image of the
The opportunities described above are just some areas
where it may be possible to use the debriefing system.
When the system is in operation, other opportunities are
likely to appear, and technology will show us which.
We have developed a mission debriefing system that in
principle covers the basic requirement and to some
extend even exceeds these requirements. No dedicated
software has been developed for use in this first
generation of the system. The input to the debriefing
system is not made especially for this purpose, but
already available sources have been sufficient (a
digitisation of the flight videos has though been
necessary). Available COTS software and hardware has
shown its value for this purpose, which means that the
main task for us has been to integrate already available
products and input. As integration is one of our
company’s main business areas, we are able to do this
quite fast and therefore within an affordable price.
Mine Clearance in a Virtual Environment
Laurent Todeschini, Thérèse Pasquier, Pascal Hue and Paul Gorzerino
Etablissement Technique d’Angers
route de Laval, BP 36
49460 Montreuil Juigné
At the same moment as France completed destruction of
its stock of anti-personnel mines (21/12/99) in
accordance with the 1998 Ottawa Agreement, in more
than 60 countries there were 100 million live, buried
“permanent sentinel” mines continuing to mutilate the
inhabitants of mine-infested regions, most of the
wounded being children (600,000 people affected over
20 years, one person killed every 20 minutes by these
devices designed to terrorise civil populations during the
war, whose effects persist for a long time afterwards).
Paradoxically, confronted by the sophisticated manufacturing techniques of these “cowardly weapons”,
French sappers use a rudimentary mine clearance
technique to render zones viable for the civil population.
With the aid of a bayonet-type tool, the operator probes
the ground until he hits a suspect device. This task is
carried out blind and one of the problems is identifying
the presence of a mine and distinguishing it from a false
alarm. This technique, demanding 100% results, based
on the skill and experience of the mine disposal team, is
taught by the Minex Centre of the Applied Engineering
Applications College.
The Human Factors division of ETAS (Etablissement
Technique d’Angers), a part of the DGA, has built and
tested version 1 of a demonstrator and virtual
environment for teaching this technique. One group
under training now has been able to distinguish the
methods for discriminating shapes after several contacts
of the probe with the mine.
In its version 2 (addition of force feedback), this
demonstrator has become a genuine teaching tool for
mine clearance strategy, enabling the instructor to
validate the relevance of the students’ probing, to
minimise the amount of probing and therefore to
increase the reliability of the decisions during an actual
operation. In due course, this tool will also enable the
technique to be taught to civilian populations and thus
accelerate the process of decontamination which still
takes a long time, costs a lot of money and, especially,
costs lives.
Technology development is already enabling us to
consider version 3, a portable system which uses
mathematical analysis of the probing geometry during
real operations, and by comparison with a database,
offers genuinely enhanced assistance to making
decisions and taking action.
1. Problems of Mine Clearance
The difficulty of mine clearance is that of DRI
(detection, recognition, identification) associated with
some action.
The main problem is detecting the device: the mine
clearance expert probes the ground in a systematic
manner in a 5×3 triangular grid arrangement to try and
detect the presence of a foreign body. If the probe hits
something, the mine clearance expert halts his movement. He then probes in order to discover the extent of
the object and to determine its shape, which will enable
him to recognise the presence of an object. He then
clears away the soil covering the object and identifies it
as being a mine or not, a munition or some unknown
If a device is present, a specialist intervenes who, after
having made a detailed identification, detects any booby
traps, analyses the condition and mode of triggering, and
decides to deal with the device either by destruction or
by rendering safe.
This paper is only concerned with the DRI task in the
initial phase; it is a difficult task, performed blind,
necessitating the mobilisation of sensors in seeking
stimuli which are indicators both for the accomplishment
of the task and for a perfectly controlled motor activity;
indeed the relevance of the indicators is dependent on
the steadiness of the prodding (the angle of incidence of
the probe must remain constant), this angle is a safety
factor and makes it possible to attack the mine at its
edges and not from above where the initiator is generally
situated. (angle of incidence lies between 30 and 35°).
2. Aim of the Research
In order to design a simulator for teaching manual
probing specific to the activities involved in mine
clearance, the Human Factors division of ETAS, in
association with the Angers Cognitive Psychology
Research Laboratory, engaged in research into learning
conditions in a virtual environment. This research was
the subject of a thesis entitled: Influence of sensorial
methods and individual characteristics on the conduct of
target detection in compared environments: the case of
virtual and actual environments.
The conduct of sensorial learning was observed for a
task in a real environment and then in a virtual
environment. The experiments were conducted in real
For information, phone +33 (0) 241 936 831.
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
time and took three aspects into account: the detection of
shapes; the detection of textures; and the detection of
sounds. Detection was based on visual, kinaesthetic and
auditory indicators. The experiments conducted in a
virtual environment were limited to a shape-detection
Nowadays, virtual reality based on visual immersion
makes it possible to explore and augment visual information in an enhanced manner and it also allows, at a lower
level, transmission of multi-modal sensorial information
by haptic, tactile or auditory feedback systems, close to
those felt in real life.
For qualitative and financial reasons, this thesis was
restricted to a part of the significant information of a
blind, sensori-motor task. The mine clearance expert's
task was not reproduced identically. A selection was
made of the sensorial modalities of exploration.
Emphasis was placed on the guiding of movements
aiming at the target using visual and auditory indicators
excluding any use of a haptic interface. Therefore the
strategy of probe movement was tested in virtual space:
1. by observing and analysing a shape detection task;
2. by comparing the learning of this task when
conducted in a real and in a virtual environment.
3. Development of Virtual Reality and Research in a
Multi-disciplinary Team
The number of articles concerning new virtual reality
interfaces shows the diversity of applications and their
development from the leisure domain into the professional world and firmly establishes virtual reality as
man’s new environmental tool.
What effect does confrontation with the virtual world
have on human behaviour? Is such immersion neutral, or
does it influence behaviour or generate different
Scientific research was directed towards specific prototypes for learning in a virtual environment with the aim
of preparing the user for a new real environment and to
encourage his adaptation by devising a new generation
of training facilities.
Few multi-disciplinary scientific teams incorporate
experimental psychologists, cognitivists and neurophysiologists for studying the perceptive, sensorial and
cognitive effects on man in the virtual environment and
reflect, inter alia, on man’s ability to transfer learning
from the virtual world to the real world. Such teams are
still not very numerous today (Kalawsky, 1999 and
Fusch, 1999) in spite of a strongly expressed need to
integrate human factor dimensions, both cognitive and
sensorial, into technological research.
The development of such collaboration is becoming
urgent in the field of man–machine interactions in order
to accelerate the development of knowledge and
understanding of the virtual reality interfaces which are
still limited today. Evaluating the “human factor”
component is complex and covers numerous aspects
such as performance linked to sensorial capacity and
cognitive processes.
It was from this viewpoint, multi-disciplinary team and
performance evaluation, that the Human Factors division
of ETAS became interested in virtual reality and
developed a virtual reality platform with the aim of
testing the upper level interfaces in learning specific
4. Experimental Conditions
4.1 Selection of subjects and the experimental task
Experiments were conducted with a group of 27
subjects, all of whom were mine clearance experts. The
task was that of target DRI. This simple gestural
movement was defined as an activity of blind probing
aimed at identifying the structure of a hidden target,
using a probe, an intermediate tool extending the
operator’s hand. The mine clearance expert had to
identify three shapes of concealed targets which were
rectangular, triangular and round. The mine clearance
expert probed into a container placed in front of him,
respecting the angle of probing used in mine clearance
(between 35 and 45°) and had to identify each shape
several times in succession in order to measure the effect
of learning.
The mine clearance experts were split into two subgroups one of which had the benefit of a visual aid (a
special virtual reality feature which made it possible to
show impacts on the target) while the other did not. The
chosen variables were the detection time, the number of
probings and the quality of the response.
Virtual reality made it possible to display the incidence
of the probe with respect to the terrain and to monitor
4.2 The conditions for exploring the virtual
The subject wore a helmet with stereoscopic vision, a V6
from Virtual Research, each channel being connected to
a graphic map so that the image was retransmitted in 3D.
The mine clearance expert’s real probe was used to
reproduce direct contact between the hand and the
exploration tool. The position of the helmet and the
displacement of the probe were controlled (6D) by
“flock of birds” position sensors.
The frame of reference was established as a function of
the environmental context, which undergoes significant
changes in the virtual environment, and the subject's
visual capabilities. In order to construct the virtual
environment special attention was paid to the selection
of the essential markers for forming the frame of
a. at the visual level, according to the concept of
identifying the shape and to remain faithful to the
analysis of a simple task, the visual markers had
basic geometric shapes concealed in an environment
with filtered geometrical data and colours. Visual
representation of the size of the probe was
proportional to its actual size.
b. at the haptic level, according to the concept of
transfer of sensorial modalities, the markers were
partially transformed into auditory markers. The
collision detection points were signalled by sounds
which symbolised the times of contact between the
probe, the environment and the targets. For gestural
guidance and accuracy reasons, the real medium was
replaced by a substitute real homogeneous medium in
which the mine clearance expert carried out the
c. at the kinaesthetic level, the constituted virtual
environment gave the opportunity of traversing the
walls and therefore influenced the kinaesthetic and
proprioceptive movement of the subject, who lost the
notion of rigidity of the wall.
5. Theoretical Approach
5.1 Sensori-motor and cognitive domain
The observations of this study of a shape detection task
were conducted in a sensori-motor learning context. In
the real environment, to read spatial information the
sensori-motor act is associated with the subject's
cognitive capabilities (Paillard, 1985). The treatment of
spatial information cross-refers to the detection
capabilities and therefore to the attention the subject
applies to discriminating sensorial space. This space is a
function of the information reflected by the environment.
Thus, the sensori-motor act is linked to the attention
capabilities of the subject and to the mental loading due
to processing the information received from the
environment. The subject's performance will also depend
on the mental representation of the action undertaken.
This information processing occurs in three stages:
• a perceptive stage which corresponds to processing
the stimulus;
• a motor stage which is the transmission of the action
undertaken on the medium;
• a response processing stage which is the subject’s
stimulus–response translation.
The identification of an object is the subject of a multimodal processing (visual, auditory, kinaesthetic and
tactile). Similarly, a subject may recode information
under several sensorial modalities. However, in order to
identify an object each individual will recode according
to his particular sensorial predisposition (Ohlmann, 91),
which enables the different approaches to be
differentiated. Research has emphasised the interactions
between the various modalities and the major
implications in co-ordinating sensorial activity. Study of
the relationship between perceptive systems describes a
variation of the predisposition of perceptive systems
according to the object of the study (Hatwell, 1994).
The individual mobilises “decoders” as a function of the
data to be extracted and of his sensorial capabilities in
processing information. Recognition of the shape of an
object or stimulus is defined by the object’s specific
intrinsic characteristics (by its shape, dimensions and
colour) and by its extrinsic characteristics (its position
and orientation in space).
Given that the visual dimension is predominant in the
virtual environment and given that the priority of
perceptive systems can change according to the type of
task in the real environment, can this perceptive priority
be modified in passing from one environment to the
How does the individual process information when
immersed in a virtual environment? Are the cognitive
processes employed in the real world automatically
efficient when the person is immersed in virtual reality?
The work of Morineau, Boujon, Papin and Le Bouedec
(1996) tends to show that the adult plunged into a virtual
world for the first time appears to use the cognitive
processes coming under the preoperative structures of a
5-year-old infant. These results project the idea that
immersion in the virtual world requires acclimatisation
or learning.
The cognitive dimensions of the personality may
intervene in processing information and have been the
subject of numerous papers (Huteau, Marendaz &
Ohlmann). In this context the concept of “dependence
and independence with regard to the visual field” offers
relevant explanations in the real environment.
The DIC (dependence and independence with regard to
the visual field) is a theory on the personality factors
presented among cognitive styles referring to the work of
Witkin (1948). Exploration strategies differ with the IC
(independent with regard to the visual field) and DC
(dependent with regard to the visual field). Huteau
(1985) developed the theory of the DIC and qualifies the
IC by higher discriminative capacities and level of
vigilance, basing this on egocentric factors and their own
perception built on gravitational, proprioceptive or
kinaesthetic factors, whereas the DC use more visual
factors for referencing themselves in space. They will be
very attentive to the positions of others, referring to
external factors. Ohlmann and Marendaz (1991) studied
this same theory from perceptive conflicts.
Before action, the operator employs a conduct, a manner
of proceeding and of giving a reasoning whose degree of
complexity varies with the task. In addition to
environmental factors, personal factors and the specific
nature of the action are going to influence the subject.
The reference point of this study relates to the concept of
restricted spaces in a static situation. The subject relies
on his capabilities of spatial representation. In order to
recognise a shape and characterise it the subject must be
capable of selecting stimuli that can be arranged in a
simple or complex fashion.
These will be differentiated by combinations of specific
information about a shape whose basic identifying
markers will be based on arrangements of points
characterised by the distance separating them, their
orientation, intersection and movement. The subject will
mobilise his attention to create grouping factors so as to
determine the boundaries and define the contours. In
order to perceive a shape and to construct a
representation, each subject has need of information
which may be total or partial. The person's strategy is
based on a representation using part of or all of the
constituents of the shape.
Image processing cross-refers to perceptive models of
basic features to guide a discriminatory behaviour
between global information and more analytical local
information. In differential psychology, cognitive styles
are evoked by global or analytic strategies in analysing
various activities such as learning, memory attentiveness
or games strategies.
5.2 A specific feature of the task: remote
As described above, the mine clearance task is
performed blind. The target is masked off from the
visual field and probing is carried out with a tool, the
probe. The probe is a link enabling three types of
sensorial information to be transmitted:
• visual, by the presence of marks left on the surface of
the soil enabling the shape to be identified;
• tactile, which makes it possible to detect collisions
and identify textures;
• auditory, during collisions for identifying materials.
The problem is to correlate these various sensations.
The mechanoreceptors situated in the hand and at the
ends of the fingers possess perceptive acuity which is
strongly discriminatory and makes it possible to decode
the detailed information which is characteristic of the
objects dealt with. Is the acquisition of information as
powerful when the hand is not in direct contact with the
Recent studies on professional situations where
interaction with the world necessitates an intermediary
contact object were aimed at measuring the performance
of haptic spatial recognition in a real environment
(Lederman & Klatzky, 1998). This work was aimed at
providing information on tactile manipulation of
intermediate interfaces in remote control or in virtual
5.3 The rapid development of the virtual environment
The rapid development of technological and computer
facilities over the past few years has resulted in a
possible skewing between the initial analyses and recent
analyses conducted in virtual environments. The
conditions for visual and haptic exploration have
advanced and so we can state that the virtual
environment is fundamentally different when we
experiment with interfaces of different generations. The
creation of illusion effects specific to each system makes
it possible to assume that the exploratory conditions are
not similar and that comparison and transfer of
information is difficult.
Applied research conducted on the processing of spatial
information in a virtual environment sometimes conveys
conflicting data in the learning domain. The exploration
conditions cross-refer to two different types of space:
a. large action spaces (representation and orientation)
b. spaces with restricted action (detection and
manipulation of simple objects).
This work shows that:
• the choice of reference markers is important for
constructing a visual space which becomes the
medium for spatial representation for the subject;
indeed, a badly monitored activity could affect the
subject's representation capabilities;
• the performance is sensitive to spatial distortion,
restriction of the visual field and to the effects of
The perceptive conflicts between movement and vision
hamper the precision of the gesture and may modify the
speed of movement of the gesture (Coello, Decety,
Leifflen & Orliaguet, 1996) and, because of this, the
concept of learning transfer between a virtual and a real
world is compromised. On the other hand, the individual
may acquire a performance on a particular sensorial
The integrity of cross-reference in the virtual
environment is an important factor. Is the perception of
information received in a real environment faithful to the
perception of information received in a virtual world?
This concept of environmental fidelity involves the
psychological judgement of the subjects plus the
technological concept.
A virtual environment which makes it possible to lift the
mask over the hidden task offers the subject the
opportunity of memorising more complete information
and enriching his spatial representation by the effects of
2D and/or 3D visualisation and of “rejects”; the
hypothesis that the subject is able to transfer this
information by limiting the affects of spatial knowledge
may then be raised.
The virtual environment makes it possible to substitute
one sensorial factor for another on the concept of
amodality and in this way to isolate a sensorial process
in order to obtain a better understanding of it. These
artefacts can also make it possible to limit the mental
load on the subject facing a heavy use of the equipment.
There are numerous controversies on:
• whether there is a need to reproduce identically the
needs felt in a real environment for a simulation tool
when there is a risk of its resulting in a heavy mental
loading on the subject;
• the employment of cognitive capabilities in coordinating information from visual space and motor
space in the virtual environment;
• the development over time of processes used in
virtual environments;
• the need for modulation for subjects’ interpersonal
dimensions during information processing in virtual
6. The Results of Experiments in Compared
The objective of the thesis was to compare conditions for
exploring a shape under real and virtual environments
using simple interfaces for assessing subjects’
performance on the acquisition of spatial information.
6.1 The needs of subjects in the real environment
The three shapes proposed require the operator to make a
double category recognition: object is round or angular;
if angular decide the aperture angle. These are exocentric
factors that the subject will seek to identify in order to
6.2 Comparison of learning in real and virtual
environments when making a detailed gestural
manipulation remotely
The results were obtained from variance analyses in
order to measure four main effects which were: the
environmental condition; learning; visual modality; and
interpersonal variability.
6.2.1 The effect of the environmental condition enabled
us to distinguish between real and virtual performance
a: Probing time
explain the subject’s strategy when he is referring
mainly to the visual field;
• individuals have identification strategies for breaking
down a shape according to the shape strategy
b: Number of probings
% correct responses
Time (s)
Nb of Probings
differentiate the three shapes. This observation has
facilitated the breakdown of the gestures into these
different translational and rotational movements.
We then observe that, in order to detect a shape:
• performance is not linked to expertise, which enables
us to formulate the hypothesis that the subjects’
personal strategies would have a discriminatory
nature independent of the training received;
• the subjects’ performance varies as a function of the
simple shapes to be identified;
• the subjects’ performance varies as a function of the
environments concealing the shapes;
• processing the information for detecting a shape and
its texture could be first class in the various unisensorial or bi-sensorial modalities while revealing a
graduation in performance measurement;
• the cognitive style of dependence and independence
with regard to the visual field is insufficient to
c: Percentage of correct responses
Figure 1: Probing times, number of probings and percentage of correct responses of shapes
for 1 test as a function of environment conditions.
The detection of a shape in the virtual environment
required three times as long as in the real environment.
The average time in the real environment was 42 s
whereas it was 123 s in the virtual environment. Whereas
the number of probings remained similar, the mean
deviation between the two situations was three points. In
contrast, the quality of the responses varied between the
two situations. In the real environment we had 62%
correct responses whereas in the virtual environment we
obtained 46% correct responses (average for the
different shapes).
The shape influences the subject’s performance. The
triangle, with a quick detection time, had the best
success percentage in both conditions. The rectangle and
the round shape showed detection conditions were more
difficult, increasingly so in the virtual environment both
for time and for correctness of response (Figure 1).
In order to determine a geometric shape the probing time
increased considerably in the virtual environment, but,
however, without, the action undertaken by the subject
being changed significantly, and without this increase in
time affecting the quality of the response. Although there
was no increase in motor activity (that is in the number
of probings), we did note that the time spent on
concentration, attention or reflection was longer for an
achievement of detection capability which was less
difficult than one in a real environment. Although the
subjects spent far more time to achieve an identical
result, we can reckon that this difference in time is
marked by the modification of sensori-motor activity
and/or cognitive activity in order to compensate for the
search for new identifying markers.
6.2.2 The learning effect
Learning was measured by systematic repetition of three
tests for all subjects. This enabled us to eliminate the
effect of chance.
Here, we found that the deviations in detection time
between the first and third test were significant (Figure
2). For the first test we recorded 123s and for the third,
111 s. Differentiating the two environments we observe
that the time curves decrease in parallel with the tests
(Figure 1).
In terms of the quality of the response, the percentages
increased during the three tests. They varied from 54.5%
in test 1 to 66% in test 3, and the percentage of correct
responses in the virtual environment showed a greater
increase between test 2 and test 3.
Once again, for the number of probings we identified a
slight deviation between test 1 (17.5%) and test 3 (15.5)
which is not significant. Repeating the tests made it
possible to commence learning under both environmental conditions (Figure 3).
Essai 1
Essai 2
Essai 3
% correct responses
Nb of probings
Time (s)
Essai 1
Number of test
Essai 2
Essai 1
Essai 3
Essai 2
Essai 3
Number of test
Number of test
a: Detection time
b: Number of probings
c: Percentage of correct responses
Figure 2: Detection times, number of probings and percentage of correct responses of shapes
for the three tests in real and virtual environments.
Tps es3 Ré Tps es3 Vi
% es3 Ré
% es3 Vi
Figure 3: In the learning situation, comparison of time and % of correct responses
for shapes between real (Re) and virtual (Vi) for test 3 (T3).
This learning is identified in real and virtual
environments. Although the changes in time and number
of probings are small we observe there is a marked
increase in the percentage success. The display of the
third test results also shows disparities in the detection of
6.2.3 The effect of visual modality in the learning
The contribution from this visual modality is measured
on a special aid provided by the virtual environment.
After each test, half of the subjects received a display of
the probing points that they were able to superimpose
over the shapes sought.
The sub-group benefiting from the visual aid detected
shapes more quickly (99 s) than the sub-group without
the aid (119 s), still with a deviation of 3 points on the
number of probings. In contrast, with quicker detection
the subjects using the aid answered with 70% correct
responses whereas the sub-group without the visual aid
had only 50% correct responses.
The presentation of visual information was envisaged as
an aid to detection. This visual window offered space for
reflection, allowing the subject to readjust the processing
of the information obtained blind in order to confirm or
reject the shape detection decision.
It appears that the subject reinforces his action at each
new test in the virtual environment by still mobilising his
sensori-motor activity just as much. In contrast, we
found a slight reduction in the time, in parallel with an
increase in the quality of the responses. The visual aid
became an aid for representing the shape. It caused
reflection on the action and enabled the subject to
readjust his strategy for the next test.
6.2.4 The personal effect
This effect was measured initially after the subjects were
divided into 4 sub-groups in accordance with the GEFT
(group embedded figures test) which is a perceptive test
measuring the capability of subjects to extract a simple
shape from a complex figure.
On a test and per sub-group, the average detection times
ranged from 68.62s to 112.36s for the number of
probings from 50.5 to 82, and for percentages of correct
responses from 47.5 to 80.
The variation in data between sub-groups did not
correspond to the results expected. The performance
specified by sub-group 1, identified as dependent with
regard to the field, distinguished processing strategies
defined in reference to Huteau’s concept. In contrast,
sub-group 4, categorised as independent with regard to
the field, represented an “economic” processing strategy
for the three variables: time; number of probings; and
percentage success. These data are currently being
studied to break down the perception strategies.
7. Subjects’ Perception of the Virtual Environment:
Review of Conversations
Subjects’ conversations during the experiments provided
us with the following information.
7.1 Visual perception of the context
The visual perception of an object in our case revealed a
dispersion on the size of the object which was assessed
at between 5 and 30 centimetres (the true dimensions
being 15×15). Evaluation of the size of an object in a
virtual environment was often unrealistic and varied
from person to person with some over- or underestimating, but others correctly assessing the dimensions
of the target.
The visual perceptive conflict is the search for a visual
compromise between the intention of accurately aiming
a probing point and the possibility of reaching this
precise probing point. In the present case, the subjects
had to seek to align several probing points in order to
create visual markers of the shape and to trace a curve, a
straight line or an angle.
All subjects mentioned visual fatigue after the repetition
of three tests interspersed by returns to a real
environment. It was for this reason that we limited the
learning to three tests.
7.2 Sensori-motor perception of the context
The time taken to perform a task under motor control.
The relationship of speed/accuracy of arm and hand
movements, measured by the time between picking up
the probe in the hand and identifying the target, was
clearly modified in the virtual environment. For the
detailed and precise gestural movements of the mine
clearance expert, the subject will have to slow down his
movement by continual monitoring in order to adjust his
gesture. This means that the subject is going to have to
adapt his sensori-motor movement by developing a
slower gestural movement in order to achieve success or
otherwise of the task undertaken.
The lack of sensori-motor and haptic information.
The virtual presentation of the target (visual and auditory
factors) lacks haptic information, located mainly on the
edges of the target. The need for this is revealed in the
mine clearance expert’s gestures by the manner of
proceeding to obtain accuracy for the angular or rounded
8. Conclusion
For this paper, comparisons of conduct and learning in
two environments enable us to report that the
performance acquired when making precise gestural
movements in a virtual environment is lower than the
performance achieved in a real environment. However,
we can state that repeating the tests enhances the
speed/accuracy factor. The improvement seen in the two
situations demonstrates that the subjects adapt and
develop with this new environment. One advantage of
the virtual environment in learning compared with the
real environment is that it offers the possibility of
calibrating the task on one or more modalities in order to
measure the significant individual and collective
performance on simple tasks. It could become a
simulation tool of benefit to mankind, offering the
possibility of isolating or combining several types of
information in order to verify the specific needs of the
9. Development
The results of this study have enabled us to specify the
changes to the virtual environment needed for the mine
clearance expert’s DRI task. The new application uses:
• a Proview 60 helmet which, combined with a more
powerful machine and graphics cards, has made it
possible to improve the visual aspect and to stabilise
the image;
• modelling of two real mines;
• a Phantom 1.5 3DoF force feedback arm which
makes it possible to provide haptic effects (especially
when detecting collisions), a more precise
manipulation of the probe position (which enables
the detail of gestural movement to be increased).
For budgetary reasons when specifying these changes,
the force feedback was limited to translational
movements. While it is necessary, theoretically, to
constrain the probe to 5 degrees of freedom (3
translations and 2 rotations — pitch and yaw) for
guiding the probe over the ground, this is not possible
with the 1.5 3DoF version of the Phantom. In the
absence of guidance, the displacement of the point of
intersection of probe with the surface of the terrain, due
to lack of gestural accuracy, is visualised in the virtual
world and leads to a perceptive conflict.
In order to overcome this technology limitation, as soon
as the end of the probe contacts the ground it is subject
to guidance by a point within a tube along the probe axis
of incidence, and the image of the probe is locked to this
This artefact serves as a decoy for the operator's sense
which, when the probe is free to move in rotation, works
along a single translational axis.
In version 2 (addition of force feedback) this
demonstrator could become a genuine tool for learning
mine clearance strategy, enabling the instructor to
validate the relevance of probings (searching for limits,
width and height, detecting contours, enabling the shape
to be identified), minimising the number of probings by
developing strategies depending on the sensorial data
received and thereby increasing the reliability of
decisions in a real operation. In time, this tool could also
make it possible to teach the technique to civilian
populations and thus accelerate the decontamination
process which is still long, costly in terms of money and
also of human life.
Technology development already permits us to envisage
version 3, a portable system which, by mathematical
analysis of the probing geometry and comparison with a
mine database, can offer a genuinely improved aid to
decision making and processing in real operations. The
greatest problem is to obtain a system which is not liable
to trigger the mine irrespective of the latter’s technology
and therefore this means a system which does not emit a
signal or signature of any kind.
With sociological problems overcome, we can envisage
using a master force feedback arm to remotely operate a
slave arm fitted with a probe; while retaining the skill
aspect of the sapper’s job, it would then become possible
to shift the task towards the rear and thereby make mine
clearance operations safer.
It nevertheless remains true that, beyond technology, the
best way of obtaining terrain completely free from the
presence of mines is not to mine it in the first place.
Acquiring Real World Spatial Skills in a Virtual World
Bob G. Witmer, Bruce W. Knerr and Wallace J. Sadowski Jr.
US Army Research Institute1
12350 Research Parkway
Orlando, FL 32826-3276
In rehearsing specific missions, soldiers frequently must
learn about spaces to which they have no direct access.
Virtual Environments (VE) representing those spaces
can be constructed and used to rehearse the missions, but
how do we ensure their effectiveness? The US Army
Research Institute was among the first to demonstrate
that spatial knowledge acquired in a virtual model of a
building transferred to the real world. While route
knowledge was readily acquired in a VE, configuration
knowledge (distance and direction to locations not in the
line-of-sight) was not. Spatial learning in the VE was
hampered not only by disorientation resulting from a
narrow FOV and multiple collisions with walls, but also
by participants’ inability to accurately estimate distances
in VEs. Poor distance estimation in VE was linked to the
reduced VE FOV and to verbal report procedures for
making the estimates. Some improvement in distance
estimates was obtained by adding auditory compensatory
cues for distance and by using the non-visually
locomotion technique for obtaining distance estimates.
Armed with knowledge that some VE characteristics
adversely affect distance estimation and configuration
learning, we conducted research to determine if unique
capabilities of VEs could compensate for those
characteristics. We developed three VE navigation
training aids: local and global orientation cues, aerial
views, and division of the VE into distinctive themed
quadrants. The aids were not provided when testing
configuration knowledge. Training included a guided
tour, free exploration of the VE and searching for
designated rooms. Configuration knowledge tests
included a shortest route test, a pointing task, and a map
construction task. An aerial view was the most effective
navigation aid, though its effectiveness depended on how
it was used. Those participants who used aerial views to
organize the VE and learn its layout during free
exploration performed quite well, while participants who
used it as a crutch to locate a particular destination
performed worse than those without an aerial view. To
ensure that VEs train effectively, we must recognize
VEs’ deficiencies, compensate for deficiencies whenever
possible, and exploit VEs’ unique training capabilities.
The U.S. Army has invested heavily in the use of virtual
environments (VE) to train combat forces, to evaluate
new systems and operational concepts, and to rehearse
specific missions. While the Army has focused mainly
on simulations for mounted combat, there is also a need
to train infantry and other dismounted soldiers. In
training dismounted soldiers there are occasions (e.g.,
rehearsing a hostage rescue mission) in which the
soldiers must learn about strategically important spaces
to which they have no immediate access. Virtual
environments can be constructed as a substitute for these
spaces, but how effective are they? This paper describes
a series of experiments that investigated the limitations
of using VE for training spatial knowledge and how VE
might be improved to meet Army human performance
Although VE technologies such as helmet-mounted
visual displays, head trackers, 3-D sound systems, haptic
devices, and powerful graphics image generators have
the potential to immerse dismounted soldiers directly in
virtual training environments, their capability to provide
effective training has yet to be ascertained. The effective
use of VE for training requires more than just VE
hardware and software. It also requires a body of
knowledge that identifies the characteristics of VE
systems that are required to provide effective training
and the training strategies and features that are most
appropriate for use with VE. In order to develop this
body of knowledge, the U.S. Army Research Institute for
the Behavioral and Social Sciences (ARI) Simulator
Systems Research Unit, initiated a program of
experimentation to investigate the use of VE technology
to train dismounted soldiers in 1992.
Experiment 1: Transfer of Spatial Knowledge
We were among the first to conduct research
demonstrating transfer of spatial knowledge from VE to
a real world environment (Witmer, Bailey, Knerr, &
Parsons, 1996). For this research, a detailed model of a
large office building was constructed using Multigen and
World Tool Kit. The model was rendered using a Silicon
Graphics Crimson Reality Engine and displayed via a
Fake Space Lab Boom. The Boom consists of a highresolution binocular display on the end of an arm that
allowed six degree-of-freedom movement and thumb
buttons for controlling forward and backward motion.
For correspondence with author: [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
The participants were sixty college students who had no
previous exposure to the building. Participants first
studied route directions and photographs of landmarks,
either with or without a map, then were assigned to one
of three rehearsal groups. These were (1) a VE group
that rehearsed in the building model, (2) a building
rehearsal group that rehearsed in the actual building, and
(3) a symbolic rehearsal group that relied on verbal
rehearsal of the route directions. Participants were then
tested in the real world building for transfer of route
Differences in training transfer were evaluated using a
MANOVA with rehearsal mode, map, and gender as the
independent measures. Only the main effect for rehearsal
mode was significant (p<.001). A follow-up ANOVA
indicated that this effect was significant for each of the
dependent measures: route traversal time (p<.001);
number of wrong turns (p<.001); and total distance
traveled (p<.05). Participants trained in the building
made fewer wrong turns (t=3.25, p<.005) and traveled
less distance (t=2.9, p<.01) than did participants who
were trained in the virtual environment (VE). VE
participants, in turn, made fewer wrong turns (t=-4.77,
p<.001) and took less time to traverse the route (t=-5.82,
p<.001) than those who were trained symbolically.
In practicing the route, participants were expected to
acquire some knowledge about the overall layout of
building (i.e., the building configuration). Configuration
knowledge was measured using the projective
convergence technique (Siegel, 1981; Kirasic, Allen, &
Siegel, 1984) and by measuring the capability of subjects
to exit the building quickly using an unrehearsed route.
The projective convergence technique requires
participants to estimate the distance and direction to
target locations not in the line of sight, and uses these
estimates to determine the participant's perceived target
location. The participants either draw lines to indicate
the distance and direction to targets (in a non-immersive
mode) or point to indicate bearing and verbally report
their distance judgments in standard or metric units (in
an immersive mode). Errors in estimated bearing and
distance using this method may either be due to poor
distance estimation skills or disorientation and a lack of
knowledge regarding the designated target location.
Hence it is not a pure distance estimation measure.
MANOVA was used to assess differences in the amount
of configuration knowledge. Surprisingly, there were no
significant differences among the various rehearsal
conditions (p=.135) and no significant differences as a
function of map use (p=.688). Only the effect of gender
was significant, with males performing better than
females (p=.015). No significant interactions were
The results suggest that individuals can learn how to
navigate a real world route by training in a virtual
environment. While the VE used in this experiment was
not as effective in training subjects as the actual
building, it was much better than verbally rehearsing
route directions, even for subjects who had previously
studied a map. The effectiveness of the VE for acquiring
route knowledge was probably limited by the display
reduced field of view and by disorientation after
collisions with virtual objects. These factors along with
an unnatural interface that controlled movement through
the VE. These factors along with participants’ inability
to judge distance in VEs may also have adversely
affected the acquisition of configuration knowledge.
Experiments 2–5: Judging distance in Ves
To better understand why participants were unable to
accurately judge distance in the VE, ARI investigators
conducted a series basic research experiments in the area
(Kline & Witmer, 1996; Witmer & Kline, 1998; Witmer
& Sadowksi ,1998). Kline & Witmer (1996) and Witmer
and Kline (1998) used magnitude estimation to measure
participants’ ability to estimate distances in a VE. The
task was performed in a virtual office corridor with
various floor and wall patterns and textures. Participants
first estimated the distance to a standard stimulus (e.g., a
cylinder at 100 feet)2. They received no feedback
regarding the accuracy of their distance estimates to the
standard stimulus, but were told that all subsequent
estimates should be made relative to that standard.
Actual distances varied from 1 to 12 feet in one
experiment (Kline & Witmer, 1996), from 10 to 110 feet
in another, and from 10 to 280 feet in a third (Witmer &
Kline, 1998). The basic measure for all of these
experiments, with the exception of Witmer and
Sadowski (1998), was the reported target distance in feet
or meters. The amount of error in these estimates was
calculated as the difference between the estimated and
true distance divided by the true distance. This error
measurement is called relative error because it is the
amount of error relative to the true target distance.
Kline & Witmer (1996) investigated how accurately
stationary observers could estimate distance to a wall in
a VE as FOV, texture, and pattern were varied. The
observer’s view was fixed (i.e., no head tracking}. The
distances being judged were between 1 and 12 feet. The
results indicated that a wider FOV (140H × 90V
degrees) produced more accurate estimates than a
narrow FOV (60H × 38Vdegrees), F(2,23)=5.85, p<.01.
Distances were typically underestimated with the wide
FOV and overestimated using the narrow FOV. For
example, a target placed 5 feet from the observer was
judged to be at 2.68 feet with the wide FOV and 8.73
feet with the narrow FOV. Significant two-way
interactions of distance with texture, F(44,1054)=2.53,
p<.001, pattern, F(22,3)=14.1, p<.05 and FOV,
F(44,1054)=2.5, p<.001. indicated that these variables
affected depth perception only at the shorter distances.
Note: All distances are given in feet. Multiply by .3048 to
convert to meters.
In another experiment, Witmer & Kline (1998)
investigated the effects of floor texture and pattern on
distance judgements to a cylinder for distances up to 110
feet. The observers were stationary and had a fixed view
of the target scene (i.e., no head tracking). Participants
grossly underestimated the target distance; the estimates
averaged about 50% of the true target distance. This
compares to estimates of approximately 75% of the true
distance in a comparable real world environment.
Cylinder size, F(1,22)=38.67, p<.001, distance,
F(5,18)=5.87, p<.01, and the interaction of cylinder size
and distance, F(5,18)=3.97, p<.05, significantly affected
the magnitude of the VE estimates. The estimates were
more accurate for the small cylinder than for the large
cylinder. For example, a target placed 50 feet from the
observer was judged to be 22.57 feet for the small
cylinder and 18.91 feet for the large cylinder. Floor
texture did not significantly affect either the distance
estimates or the magnitude of the relative errors.
Witmer & Kline (1998) also reported the results of an
experiment in which moving observers judged distance
traversed for distances up to 280 feet. Half of the
participants received compensatory cues (an audible tone
every 10 feet) to help them calibrate their distance
judgements to the true target distances. Although these
cues were provided on only half of the trials, they
improved performance to levels approaching perfect
performance, F(1,60)=11.49, p<.001. The judgments
averaged 96% of the true target distance when
compensatory cues were present but only 67% of the
target distance when compensatory cues were absent.
The mode of locomotion used in moving through the VE
(treadmill, joystick, or teleport) did not significantly
influence the accuracy of the distance estimates, but
speed of movement had a significant impact on
estimation accuracy, F(1,60)=36.15, p<.001. Distance
judgments were more accurate at the slow speed than at
the fast speed. For example, a distance of 280 feet was
judged to be 267 feet on the average when moving at the
slow speed and 241 feet when moving at the fast speed.
Accuracy of the distance estimates generally decreased
as distance to the target increased, F(7,54)=482.53,
The extremely poor VE distance estimates made by a
stationary observer and the lack of substantial
improvement in the accuracy of the estimates when
observer movement was added (Witmer & Kline, 1998)
suggests that either verbal estimates of distance are not
very accurate or that VEs degrade distance estimation to
a large degree. The ability of participants to accurately
report distances in feet or meters varies widely among
participants, and may be independent of their perception
of target distance. These individual differences may
inflate the amount of error observed in estimating target
distance. To determine how much of the problem is due
to the requirement to provide verbal estimates of
distance and how much is due to VE factors, Witmer &
Sadowski (1998) used non-visually guided locomotion
(NVGL) to obtain distance judgements in VE and real
world environments. Participants viewed a target for 10
seconds from a stationary position, forming a mental
image of the target’s location. They were then
blindfolded and asked to walk to the target’s location,
keeping the target's location in their minds as they
approached it and stopping when they thought they had
reached it. They were asked not to count steps or time
mentally. The distance judgments were performed both
in a real world officer corridor and in a virtual office
corridor modeled to simulate the real world corridor. The
target, a construction cone, was clearly visible and
distinct from the background at all distances. Participants
made judgements for targets placed at distances between
15 and 105 feet. The distance judgements averaged
about 85% of the true target distance in the VE and 92%
of the true target distance in the real world environment.
The differences between the distance judgements in the
VE and in the real world were significant, however,
F(1,20)=4.41, p<.01. The magnitude of the errors in the
VE was nearly twice those obtained in the real world.
Implications of the learning transfer and distance
estimation experiments
Our initial investigation of configuration learning
(Witmer et al., 1996) suggested that distance estimates in
VE were poor. Witmer and Kline (1998) confirmed this,
showing that distance estimation in a VE is significantly
less accurate than in the real world. Kline & Witmer
(1996) demonstrated that reducing the FOV for one of
the devices (BOOM2C) could affect not only the amount
of error in distance estimates, but also the direction of
that error (underestimates vs. overestimates). The
hypothesized that narrow FOV produced less accurate
estimates by reducing or eliminating linear perspective
cues. Witmer & Kline (1998) found that manipulation of
textures did little to eliminate the observed deficits in
performance. Although target size did influence
performance, manipulation of the size of unfamiliar
objects is not a practical solution. Taken together, these
studies suggest that VEs distort monocular or
stereoscopic distance cues, negatively impacting the
distance judgements in those VEs.
We had anticipated that providing the cues for distance
associated with movement would compensate for the
distortion of other distance cues in VE, resulting in
substantial improvements in performance. However,
Witmer & Kline (1998) found that neither movement
method nor edge rate markedly changed the distance
judgments. These results indicate that proprioceptive
cues and visual flow cues may not play a major role in
making distance judgements in a VE. In contrast,
movement speed clearly influenced distance judgments,
suggesting that the time spent covering a distance
changes one's perception of distance traveled. This
research also suggested that distance perception in VE
could be recalibrated cognitively by providing
compensatory cues for distance. This cognitive
recalibration may or may not extend to other distances or
to other environments, however. Witmer & Kline (1998)
did not collect data that would answer questions about
transfer of estimating skill to other distances or
Using NVGL to evaluate the accuracy of VE distance
estimates altered our working hypothesis regarding how
much VE degrades distance estimates. This procedure
yielded more accurate VE distance estimates, suggesting
that the use of verbal distance estimates is partly
responsible for the poor performance observed in our
research. However the magnitude of the errors in VE
using the NVGL procedure was still twice that observed
in the real world, establishing beyond any reasonable
doubt that VEs are distorting perceptual judgments of
Factors influencing VE distance judgements
What factors might be responsible for this distortion? In
our search for an explanation it is important to remember
that the performance decrements were found across
various VEs using different display devices, and with
varying movement conditions. It is also important to
keep in mind the distances investigated in each
experiment, because the effective range of various
distance cues vary with the distance being judged.
To understand why VE distorts distance perception at the
target distances investigated, we need to know which
distance cues are effective at those distances, and to
assess the extent to which these cues were present or
absent in our research. Cutting and Vishton (1995) have
identified which depth cues are most effective at
different distances and related these cues to three
egocentric regions or zones of space: (1) personal space
extends just beyond arms reach and refers to space used
by a static observer; (2) action space extends to about
100 feet and refers and includes distances in which an
observer can throw an object to another person or easily
talk to others; and (3) vista space extends beyond 100
feet. Kline & Witmer (1996) studied both personal and
action space. In personal space the most important depth
cues are occlusion, binocular disparity, relative size,
convergence and accommodation. The remaining studies
investigated action space and vista space. The primary
distance cues in action space and vista space are the
pictorial cues, including occlusion, height in the visual
field, convergent linear perspective, relative size, and
relative textural density. In addition, two other distance
cues, binocular disparity and motion perspective are
effective distance cues in action space. Note that
accommodation and convergence are not effective depth
cues in action space or vista space.
Witmer & Kline (1997) have shown that while relative
textural density influences distance estimates in VE, its
effects are typically too small to account for the
differences between real world and VE distance
estimation performance. Similarly adding observer
movement, which provides motion perspective and other
movement related cues does not eliminate the deficits in
performance in VEs (Witmer & Kline, 1998). Research
by Wright (1995) and Witmer & Kline (1996) suggests
that simply using a high resolution or wide FOV VE
display cannot erase the deficits in perceived distance.
Although occlusion is probably the most powerful depth
cue in action space, it was not a factor in our distance
estimation tasks. Of the remaining distance cues listed
by Cutting & Vishton (1995), height in the visual field,
convergent linear perspective, relative size, and
binocular disparity appear to be the most likely
candidates for explaining the observed discrepancies
between VE and real world judgements of distance.
The National Research Council (1997) has suggested
that the restricted FOV provided by VE displays must
degrade height in the visual field and convergent linear
perspective as cues for distance at some point. The
limited vertical FOV found in most VE displays (ranging
from 40 to 90 degrees) may be responsible for this
degradation. By comparison, the real world vertical FOV
is approximately 120 degrees. A reduced vertical FOV
may result in distant objects appearing closer in VE than
they would in the real world because these objects would
be compressed into a smaller visual frame as they recede
into the distance. Kline & Witmer (1996) showed that a
reduced horizontal FOV could also adversely impact the
accuracy of distance estimates by reducing or
eliminating linear perspective cues. Because linear
perspective cues are among the most effective distance
cues in simulated environments (Surdick et al., 1997),
reducing or eliminating these cues can have a major
impact on the accuracy of distance estimates.
In VEs, emulation of binocular disparity is achieved by
presenting different images to the two eyes with some
central area overlap. While this technique may provide
the illusion of depth in VE, it may not faithfully
reproduce real world depth. Cutting & Vishton (1995)
noted that early stereoscopic pictures enhanced the
distance between the eyes to show large expanses and
cityscapes, diminishing the effective size of the objects
seen. Relative size may be important factor at the closer
distances because the perceived size of an object
accelerates as the distance to the object decreases,
yielding a looming effect. Accommodation and
convergence cues are not accurate in VEs, a fact that
researchers often use to explain poor distance estimation
in VEs. However, these cues are only important for
judgments in personal space and at the shorter distances
within action space.
Additional research is needed to determine which of the
distance cues operating in action space are most
responsible for degrading distance judgements in VE.
Once the causes of this degradation are isolated, we can
begin working toward a solution. The solution may be as
simple as increasing the VE display vertical or horizontal
FOV, or adjusting the overlap in VE stereoscopic
viewing devices. On the other hand, it may involve
major technological advances, such as inventing new
techniques for emulating binocular disparity in VE
Having identified some of the factors that affect distance
judgements in VE, we turned our attention back to how
to best use VEs for training configuration knowledge.
Our approach was to utilize unique capabilities of VE
that might compensate for its inherent deficiencies (e.g.,
VE’s tendency to distort distance judgements).
Enhanced VEs for spatial knowledge acquisition
A computer model of one floor of a large office building,
used in previous research (Bailey & Witmer, 1994;
Witmer et al., 1996) was adapted for this experiment. All
passageways in the virtual building were widened to
reduce collisions, an improved collision detection
algorithm was introduced that decreased the need to back
away from objects following a collision, and additional
rooms were modeled. Separate VE models were
constructed to represent the standard and enhanced
environments. The enhanced environment was created
by adding theme objects and sounds to the standard
environment model. The models were created using
Multigen II software and rendered by a Silicon Graphics
Onyx with eight 200MHz processors and three
RealityEngine2 Graphics Pipes. Both models were
displayed using a Virtual Research V8 Helmet-Mounted
Display (HMD). Locomotion through the VE was
achieved by virtual walking in the safety pod shown in
Figure 1. Head and body movements were independently
or enhanced VE, received orientation cues or did not,
and could chose to view the VE from an aerial
perspective or was restricted to viewing the VE from the
normal perspective. Orientation cues included an arrow
projecting from the chest of the participant’s avatar and a
flagpole visible throughout the environment.
Groups having an aerial perspective could view the VE
from heights of 49, 98, and 394 feet for a period of up to
one minute. After one minute, they automatically
returned to the normal perspective view. The viewing
heights were selected such that participants could see
either the whole third floor layout at once at 394 feet or
parts of the layout at 39 and 98 feet. More objects in the
environment could be recognized at the lower viewing
heights. Figure 2 shows the VE from a viewing height of
98 feet. While in the aerial mode participants could
further explore the environment by flying to other aerial
locations (accomplished by walking in place). To return
to ground level they pressed the thumb button on their
hand controller, and gradually descended to reenter their
virtual body at the exact location where they left it when
they started to fly.
Figure 2: Aerial View of Third Floor Viewed at 98 feet
Figure 1: Safety Pod for Virtual Walking
The participants were sixty-four college students who
had no previous exposure to the building. Following a
brief train-up, the participants were randomly assigned to
one of eight treatment groups, who received different
levels of navigation aids. Depending on group
assignment, a participant experienced either the standard
The enhanced environment model was divided into four
themed quadrants or districts. Groups exposed to the
themed environment encountered sights and sounds
associated with the themed quadrants. Each destination
had a memorable theme object located inside the room
and an associated sound that became louder as the
participant approached the destination room. Additional
theme objects were positioned along the building
corridors, but no sounds were associated with these
additional objects. The themes embedded in the
quadrants were a tropical island theme, a wild animals
theme, an extraterrestrial (or outer space) theme, and a
sports theme. Upon encountering a theme object located
inside one of the destination rooms, participants were
asked to identify the theme represented by that object.
This encouraged participants to associate destination
rooms with their location in a particular quadrant.
The orientation cue groups were asked to relate their
current position to their starting position marked by a
virtual flagpole. This was accomplished by facing the
flagpole upon reaching each destination. The flagpole
served as a global orientation cue that allowed
participants to continually update their current position
based on their known starting position. Participants were
told to use the arrow projecting from the chest of their
avatar as an indication of their current heading and as a
way of aligning their virtual body so as to avoid
collisions with walls and doorways.
Individual training and testing phases comprised the
research. During the first training phase participants
followed a virtual tour guide through the VE, pausing at
each destination room, and identifying it by name. The
tour guide verbally described the ‘non-theme related’
distinguishing features of each destination. In the second
training phase, participants explored the VE freely, while
trying to locate and identify each previously visited
destination. In the final training phase, participants
attempted to take the shortest route from the third floor
lobby to each named destination. If the participants did
not find the destination within three minutes, they were
verbally guided to it. Knowledge of the building
configuration was tested by asking participants to
complete the following tasks: (1) take the shortest route
between designated rooms, (2) estimate the distance and
direction to locations not in the line-of-sight, and (3)
place room cutouts in their correct locations on a map.
Similar to the NVGL procedure, participants estimated
distance by walking the straight-line distance between
their current location and the perceived location of the
destination without vision. Navigation aids were not
provided during the testing phase. A follow-up room
placement test was given one week after the initial test to
examine retention of configuration knowledge.
The purpose of the navigation aids was to offset the
effects of VE deficiencies that interfere with the
acquisition of configuration knowledge in a VE. The
orientation cues had no significant effects on
configuration knowledge acquisition, F(4,51)=2.05,
p=.10. Participants receiving the enhanced environment
performed better during training than those who received
the standard environment, F(4,51)=2.80, p<.05, but not
on the tests of configuration knowledge. Only the
participants who received an aerial perspective view
performed significantly better both during training,
F(4,51)=5.69, p<.001, and on the configuration
knowledge tests, F(6,50)=3.44, p<.01. Participants with
an aerial view during training also performed better on
the 1-week retention test, F(1,51)=9.76, p<.01.
The effectiveness of the navigation aids, including the
aerial view, seemed to depend on how the participants
used the aids. When the aids were used as a crutch to
quickly find a room, they were not effective. Similarly in
those cases where the navigation aids increased the
workload beyond what the participants could handle, no
performance gains were realized. The navigation aids
seemed to work best when participants were able to use
them to mentally structure the environment. For
additional discussion of the effects of these navigation
aids, see Witmer, Sadowski, and Finkelstein (in press).
What then must be done to ensure that training in virtual
environments meets military human performance goals?
The first step is identify the shortcomings of VE that
adversely affect VE training effectiveness and link these
shortcomings to specific performance deficiencies. For
example, in spatial learning, a reduced FOV in VE was
linked to poor distance estimation and spatial
disorientation, ultimately impairing the acquisition of
route and configuration knowledge. The next step is to
determine if the deficiency can be addressed directly, or
if not, how to compensate for the deficiency. Currently
increasing the FOV for VE displays is an expensive
proposition and large FOV devices may sacrifice
resolution for the larger FOV. We used auditory cues to
compensate for poor distance estimation in the VE and
showed that the estimates were improved even when the
cues were not present. We adopted the NVGL procedure
to reduce the affects of individual differences on distance
estimation tasks, and used it to measure distance in the
projective convergence test. We took steps to reduce
collisions in VE, thereby reducing the amount of
disorientation that occurred with a narrow FOV display.
We also increased the effective FOV by providing
participants with an aerial view leading to improved
acquisition of configuration knowledge. In searching for
effective compensatory mechanisms, some promising
factors had little practical effects. A more realistic
walking interface (i.e., a treadmill) did not improve
distance estimates and dividing the environment into
themed quadrants or districts did not improve the
performance on tests of configuration knowledge. This
demonstrates the importance of evaluating VE interfaces
and training enhancements in controlled experiments
before implementing them in military training
Bailey, J.H. & Witmer, B.G. (1994). Learning and
transfer of spatial knowledge in a virtual
environment. Proceedings of the Human Factors and
Ergonomics Society 38th Annual Meeting. Santa
Monica, CA: Human Factors and Ergonomics
Society, 1158-1162.
Cutting, J.E. & Vishton, P.M. (1995). Perceiving layout
and knowing distances: The integration, relative
potency, and contextual use of different information
about depth. In W. Epstein & S.J. Rogers (Eds.),
Handbook of perception and cognition: Volume 5,
Perception of space and motion (pp. 69-117). New
York: Academic Press.
Kirasic, K.C., Allen, G.L. & Siegel, A.W. (1984).
Expression of configurational knowledge of largescale environments: Students’ performance of
cognitive tasks. Environment and Behavior, 16 (6),
Kline, P.B. & Witmer, B.G. (1996). Distance perception
in virtual environments: Effects of field of view and
surface texture at near distances. Proceedings of the
Human Factors and Ergonomics Society 40th Annual
Meeting, 1112-1116.
National Research Council (1997). Tactical display for
soldiers: Human factors considerations. Washington,
DC: National Academy Press.
Siegel, A.W. (1981). The externalization of cognitive
maps by children and adults: In search of ways to ask
better questions. In L.S. Liben, A.H. Patterson & N.
Newcombe (Eds.), Spatial representation and
behavior across the life span: Theory and
application. (pp. 167-194). New York: Academic
Press, Inc.
Surdick, R.T., Davis, E., King, R.A. & Hodges, L.F.
(1997). The perception of distance in simulated
visual displays: A comparison of the effectiveness
and accuracy of multiple depth cues across viewing
distances. PRESENCE: Teleoperators and Virtual
Environments, 6 (5), 513-531.
Witmer, B.G., Bailey, J.H., Knerr, B.W. & Parsons, K.C.
(1996). Virtual spaces and real world places:
Transfer of route knowledge. International Journal
of Human Computer Studies, 45, 413-428.
Witmer, B.G. & Kline, P.B. (1997). Training efficiently
in virtual environments: Determinants of distance
perception of stationary observers viewing stationary
objects (ARI Research Note 97-36). Alexandria, VA:
US Army Research Institute for the Behavioral and
Social Sciences.
Witmer, B.G. & Kline, P.B. (1998). Judging perceived
and traversed distance in virtual environments.
Environments, 7 (2), 144-167.
Witmer, B.G. & Sadowski Jr., W.J. (1998). Nonvisually
guided locomotion to a previously viewed target in
real and virtual environments. Human Factors,
40 (3), 478-488.
Witmer, B.G., Sadowski Jr., W.J. & Finkelstein, N. (in
press). Training dismounted soldiers in virtual
environments: Enhancing configuration learning
(Draft ARI Technical Report). Alexandria, VA: US
Army Research Institute for the Behavioral and
Social Sciences.
Wright, R.H. (1995). Virtual reality psychophysics:
Forward and lateral distance, height, and speed
perceptions with a wide angle helmet display (ARI
Technical Report 1025). Alexandria, VA: US Army
Research Institute for the Behavioral and Social
Advanced Air Defence Training
Simulation System (AADTSS)
Virtual Reality is
Reality in German Airforce Training
M. Reichert
Federal Office of Defence Technology and Procurement
FE I 4
Ferdinand-Sauerbruch-Str. 1
D-56073 Koblenz
This article describes the AADTSS simulation system
and explains the reasons why it was realised with Virtual
Reality technology.
The requirements
The programme started with the following main
• STINGER team training (commander, gunner)
• transportable, mobile
• size of scenarios: 360°×130°
• 8 targets, 8 effects, 2 missile firings at the same time
• long range aircraft detection and identification
• fast database generation system.
Why Virtual Reality?
As you can see, there are the contradictory requirements
size of “scenario” and “transportable”.
These requirements make it impossible to use a normal
dome-display-system. The solution is Virtual Reality.
The technical solution
The AADTSS simulator is integrated in a container.
Both, the commander and the gunner are wearing HeadMounted-Displays (HMDs). Because of the requirement
“long aircraft detection/identification”, the resolution per
eye is 1280×1024 pixels. The HMDs are without a seethrough-option, because of the better contrast and the
advantage, that there is no need to switch of the lights in
the container.
The two team members need to communicate
acoustically and optically. Because of the closed HMDs
the students cannot see each other. This problem was
solved by modelling the commander and the gunner as
This solution might be a little bit funny, but it is well
accepted by the soldiers.
The commander is tracked by an inertial tracking
system, the gunner is tracked by an optical tracking
system. Magnetic tracking systems are not suitable for
use in environments like containers made of metal.
Orientation rings for the commander and the gunner are
integrated in the container. This solution is necessary
because of the HMDs without see-through-option.
Database generation system
The generation of databases is based on stereoscopic
photos. It allows the generation of scenarios, targets and
flight paths and is independent of the simulator. It
consists of one workstation, the stereo-camera-system
and one control-PC.
The main advantage of this system is, that there is no
need for geographical data like maps or DTED and
DFAD data.
• 04/1995
• 11/1997
• 05/1999
First requirement
Troop trial unit
Final configuration
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
“What is Essential for Virtual Reality
to Meet Military Performance Goals”
Performance Measurement in VR
Lt. Jim Patrey1, Robert Breaux, Andrew Mead & Elizabeth Sheldon
Virtual Environment Training Technology (VETT) project
Naval Air Warfare Center Training Systems Division (NAWCTSD)
12350 Research Parkway
Orlando, FL 32826-3275
One of the unique attributes and potentially greatest
assets of virtual environments is the unique ability to
comprehensively measure human performance. In the
real environment, measuring human behaviors is usually,
though not always, feasible and typically extremely
effort intensive and cost-prohibitive. Similarly, there is
substantial environmental variability that can have
pervasive effects on human performance, but is beyond
any feasible, economic data capture. Virtual
environments instill the capability for comprehensively
monitoring both user inputs and interactions and the
environment (as well as control the virtual environment
and thereby eliminating confounding variables with
precision beyond that of real environment lab research).
Monitoring and measuring human behavior in this
fashion provides three invaluable elements. Firstly, it
furnishes a valuable research tool for the development of
outcome measures for performing research. Secondly,
performance measurement has training value for
assessment and evaluation. The derivation of accurate
performance measures can enable improved proficiency
and reduced training time when implemented in a
training curriculum. Finally, the development of
performance measures can facilitate the development of
intelligent tutoring systems and thereby cost-effective,
stand-alone training systems. Measuring human
performance can be of great use in the facilitation and
maximization of training.
Performance measurement involves three distinct
processes: Identification, Monitoring, & Evaluation.
Identification is the determination of the significant
measures of performance for a given task. This is
typically accomplished via cognitive task analysis and
intense subject-matter expert (SME) interviews and
observation and/or statistical analytic techniques
occurring after the observation of real world
performance. These approaches are the two traditional
approaches to performance measure development.
of performance measures - cognitive model driven and
data-driven approaches. Cognitive models enable a new
method of performance measurement. Through
traditional approaches (such as SME interviews) a
cognitive model can be developed for a given task (in
truth, a cognitive task analysis is a variant of a cognitive
model, typically represented in GOMS format). There
are a host of cognitive modeling approaches (discussed
in detail in Pew & Mavor, 1996), but they all generally
afford identification of cognitive variables not easily
discernable through traditional approaches. However, the
usefulness of such models for performance measurement
is dependent on the accuracy of the model and the
development of cognitive models can be resourceintensive, particularly for complex tasks.
Data-driven approaches are also afforded by virtual
environments. The ability to thoroughly monitor and
record all actions and interactions in a virtual
environment enables data mining approaches to provide
value to the determination of performance measures.
There are numerous data-driven techniques for mining
data (such as neural networks, genetic algorithms,
evolutionary computing, etc.), but it is fuzzy sets theory,
or fuzzy logic, which may hold the most promise for
identifying crucial aspects of human performance.
Unlike other approaches, fuzzy logic preserves the
semantic value of the input variables. Output from fuzzy
models meaningfully represents human behavior and can
be directly applied to performance measure development
(Cowden, Burns, Casey, & Patrey, 2000).
It is likely that all of these approaches should be
integrated to fully profit from virtual environments for
the identification of performance measurement. Ideally,
we will someday be able to place a SME in a VE to
perform a task and have hybrid models (of both topdown cognitive models and bottom-up data-driven
models) monitor the virtual world and generate
performance models that produce measures of
The advent of virtual environments has fostered the
development of two new approaches to the development
For correspondence with author: [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
VE-based performance measures cannot be developed
Accomplishing this requires monitoring behaviors and
their consequences within the VE. Behaviors include
active behaviors such as control inputs and verbal
commands as well as passive behaviors such gaze
surveys. The consequences of these actions include
movement through the VE and interactions with and
within the VE resultant from user behaviors. The
principal behaviors and consequences must accurately
represented, inherently measurable, and recorded for the
effective use of VR for performance measurement.
Implicit in this is the indispensability of adequate
modeling of the VE. All salient cues must be represented
with suitable fidelity within the VE for the performance
measures reaped to represent real world performance.
This may be the greatest challenge for the practical use
of VE for performance measurement. It generally
behooves VE developers to minimize the fidelity in
order to minimize processing demands and cost. The
level of fidelity should be mapped to the task fidelity
requirements so that 'training' fidelity, the level of
fidelity required to meet training requirements, can be
attained. The role of the SME cannot be underestimated
in fulfilling this balance between minimal fidelity and
requirements. Achieving this necessitates thorough frontend analysis prior to significant investment in
development of the VE.
Finally, the effective use of VE in performance
measurement should also provide performance
Evaluation. Beyond identifying and monitoring
performance measures is the need to discriminate
good/expert performance from bad/novice performance.
This is most meaningful for VE in the context of
developing intelligent tutoring systems (ITS), but also
permits structured, empirically based, objective feedback
in any circumstance.
Derivation of evaluatory measures of performance
(MOPs) is generally accomplished through methods
similar to identifying performance measures. Traditional
methods include SME ratings of performance (typically
gathered through observation of another's performance)
and statistical analysis. Cognitive model and data driven
approaches also hold promise for evaluating
performance (particularly in contrasting novices and
experts), but they have not been applied as extensively in
this domain.
Traditional performance measure development for
virtual underway replenishment
An immersive virtual environment has been developed
for underway replenishment (UNREP) with a U.S. Navy
Cruiser (see Davidson, 1997 and Martin et al., 1998 for
more information on the virtual UNREP). An UNREP
involves the transfer of fuel, stores, ammunition, and
people from one vessel to another while underway. It is
comprised of four distinct phases (see Figure 1): 1)
Approach - from awaiting station to bow-stern crossing
(overtake oiler & attain lateral separation), 2) Slide-in transition from approach to alongside (match velocity),
3) Alongside - stationkeeping (maintaining proper lateral
separation and matched velocity), & 4) Breakaway separation of own ship from oiler.
Figure 1. Depiction of the phases of Underway
The ship is controlled by the Conning Officer via verbal
commands to a virtual helmsman. The verbal commands
are broken down into two main types: control commands
and requests for information. Control commands include
engine commands such as all stop, all back, all ahead,
indicate knots, increase turns, & decrease turns and
rudder commands such as rudder amidships, steer
course, left rudder, & right rudder. The Conning Officer
can also make “requests for information” regarding
rudder angle, relative bearing, true bearing, heading,
speed, & range. These shiphandling behaviors provide a
solid foundation upon which to develop MOPs.
Iterative inputs from SMEs also identified ship dynamic
features indicative of good performance. These
parameters vary depending upon the phase stage
(approach, slide-in, alongside, or breakaway), but
generally include relative positional data (vertical
separation, lateral separation, & bearing) and relative
velocity. The following depicts the statistical analyses
conducted in pursuit of MOP identification.
Twenty-six (26) male Navy personnel (students &
instructors) of the Surface Warfare Officer’s School
(SWOS) participated as subjects. Due to technical errors
in the VE data collection process, data from eight (8)
subjects, were not included in the analysis. The level of
duties represented in the sample were Ensign (ENS, n =
6), Division Officer (DIVO, n = 4), Department Head
(DH, n = 3), and Commanding/Executive Officer
(CO/XO, n = 5). Further description of the subject
demographics can be found in Martin, Sheldon, Kass,
Mead, Jones, & Breaux (1998).
positioned 1000 yards directly behind the supply ship,
and both ships were traveling on a heading of 130o at a
speed of 15 knots (the UNREP course and speed).
Subjects received a review sheet (an informative briefing
of the VE ship’s characteristics, general reminders
regarding hydrodynamic effects, and rules of thumb
applicable to UNREP) to study prior to the experiment.
Figure 2. Virtual Underway Replenishment.
The VE testbed was comprised of the following
hardware: Dual Processor Octane R 10000 Processors,
MXI Graphics, Octane Channel Option, and Indigo2
Impact R 10000 IDS by Silicon Graphics, Inc. Subjects
used a VR4 Head Mounted Display (HMD) by Virtual
Research, and IS600 Inertial Tracker by Intersence to
view the graphics. The commercial software components
were dVise by Division and Vega Marine by Paradigm.
Further specifications can be found in Davidson
(1996,1997a, 1997b).
Subjects were administered six questionnaires: PreQuestionnaire, Demographics Questionnaire, PreExposure Symptom Checklist, Scenario Review, PostExposure Symptom Checklist, and Debrief. The PreQuestionnaire and Demographics Questionnaire were
completed prior to the experimental session. The PreQuestionnaire solicited comments regarding the critical
points of an UNREP, UNREP performance measurements, typical UNREP strategy, and a diagram of the
UNREP outlined in the strategy. The Demographics
Questionnaire gathered background information on
shiphandling, UNREP, and VE experience. The Scenario
Review was administered between the performance of
the two VE UNREPs to obtain the subject’s appraisal of
the first UNREP and planned strategy modifications for
the second pass. The Debrief was given after the
performance of the second UNREP to acquire a
comparison of the two UNREPs and usability comments.
The results of the usability comments are described in
Martin et al. (1998). The Pre- and Post- Exposure
Symptom Checklists, an adaptation of the Simulator
Sickness Questionnaire (SSQ, Kennedy et al., 1993;
Lane & Kennedy, 1988), were used to examine the
occurrence of simulator side effects and will be
described in a future report.
VE UNREP Scenario
The scenario task was to execute an UNREP from the
port bridgewing of a guided missile cruiser (CG) and
conn the ship alongside a supply ship, maintain the
alongside position (at 120 feet lateral separation) for two
minutes, and breakaway from the supply ship (see Figure
2 for alongside view). At the scenario start, ownship was
The session began with the subject’s review of written
instructions describing the task and pictures of the
location of the supply ship’s UNREP station displayed
on a PC monitor. The subjects were instructed to issue
commands and requests for information as in the real
world. These commands and information requests were
input to the simulator by an experimenter via keyboard
strokes. Replies to commands were made by a prerecorded speech system, and replies to requests for
information were provided verbally by the experimenter.
The subjects completed two UNREPs and were given a
brief rest period between the UNREPs in which they
completed the Scenario Review. It took approximately
1.5 hours to complete the entire experimental session.
The first UNREP was considered a practice trial
enabling subjects to adapt to the VE. The second
UNREP was used for all subsequent analyses.
Following UNREP performance, SMEs were solicited to
rate UNREPs presented as plot tracks. Six experienced
Surface Warfare Officers rated performance by
evaluating a printed track of each subject’s UNREP
performance. Each track was assigned a rating of 0 to
100. One rater who demonstrated poor internal
consistency and poorly correlated with the group was
dropped. The mean inter-rater correlation of the
remaining five raters = .68; ranging from .56 to .78. The
ratings from the five remaining raters were averaged to
derive a final performance rating for each UNREP.
The experience level of the sample was diverse, ranging
from ensign to commanding officer with a median of 8
years shiphandling experience. The median number of
deployments completed was 4 and the median elapsed
time since the last deployment was 3 years. A typical
UNREP has an extended duration. Depending on the
type of ship, an UNREP can last as long as12 hours
(though 1 to 3 hours is more typical), therefore several
officers assume the conn during a single evolution. The
subject’s UNREP experience included completion of a
median of 17 approaches, a median of 22 alongsides, and
completion of a median of 10 breakaways.
Pursuit of good performance measures began with
evaluation of requests for information, engine & rudder
commands, & ship dynamic characteristics.
Requests for information (RFI)
Difference comparisons between novice ensigns (no
shiphandling experience; n=6) and experienced
shiphandlers (n=12) were made for RFI (rudder angle,
relative bearing, true bearing, heading, velocity, &
range). Novice shiphandlers made significantly more
requests for velocity (Novices = 6.3, Experts = 3.0; Oneway ANOVA, F=2.40, p<.05) and relative bearing
(Novices = 11.2, Experts = 3.3; One-way ANOVA,
F=6.99, p<.01). These differences are consistent with
rules of thumb that novices are taught to judge relative
positions; experienced shiphandlers rely instead on
“seaman’s eye” (Crenshaw, 1965) and rarely use these
rules and therefore don't make the same RFIs.
In order to determine whether any RFIs were predictive
of performance, a linear regression model of SME
ratings from RFI was conducted and produced an R=.48
(F=0.59, ns). No individual RFIs were statistically
significant in this model. This suggests that RFIs are not
effective measures of performance, though they do
appear to be indicative of experience.
Engine & Rudder commands
One-way ANOVAs were conducted comparing novice
and expert shiphandlers on their cumulative use of
shiphandling commands; none of the comparisons on
these engines and rudder commands were statistically
significant. Furthermore, a linear regression predicting
SME ratings from these shiphandling commands was
also not significant (R=.47, F=0.91, ns).
Ship dynamics
Candidate measures of ship dynamics as characteristic
performance measures were gathered from SME
interviews and prior shiphandling dynamics analyses
(Martin et al., 1998, Patrey et al., 2000). The most
meaningful single relative position, based upon these
prior analyses, is within the transitional slide-in phase; in
particular, the ship dynamic characteristics (lateral
separation, bearing, velocity, & acceleration) at
approximately 100 feet astern of the stationkeeping
position appears to be the single most distinguishing
point. Additionally, measures from the alongside phase
for minimum lateral separation (LS), maximum LS, root
mean square (RMS) LS, & RMS vertical separation (VS)
were included as potentially significant measures.
A linear regression predicting SME ratings from these
ship dynamic characteristics was highly significant
(R=.98, F=15.76, p<.001). In order to create a more
parsimonious model, a backward elimination linear
regression predicting SME ratings from this host of
variables reduced the model to velocity, relative bearing,
LS, maximum LS, & RMS LS (R=.92, F=12.99,
Performance measures were successfully identified for
virtual UNREP using a traditional approach of
identification. Indices of relative position (LS, RMS LS,
& maximum LS), relative velocity, and relative bearing
significantly predict SME evaluation of performance.
Iterative development of the VE coupled with feedback
and inputs from SMEs and data analysts enabled the
monitoring of salient measures of performance (such as
ship dynamics). Furthermore, this has provided a basis
for empirically driven performance evaluation.
This clearly demonstrates the functionality of using VE
as a tool for deriving performance measures for a real
world task. Collecting this quality of data in the real
world is a daunting task (though efforts are underway to
accomplish this to validate matching between real and
virtual UNREPs). While possible to collect this data in
the real world, it is difficult and uneconomical to do so,
particularly when VE affords an alternative, potentially
more effective, method for accomplishing this.
While this particular performance measure derivation
effort was primarily driven by a traditional approach to
knowledge extraction, virtual data was manually
processed with standard statistical methods to glean
performance measures that were not wholly apparent
from SME interviews. This highlights the need, for at
least some types of task, such as those heavily perceptual
in nature and not easily verbalized, for additional
methods of knowledge elicitation.
Data and cognitive model driven approaches were
discussed as potential methods of facilitating and
streamlining the knowledge acquisitions process.
Currently, both approaches are being investigated for
virtual UNREP. Fuzzy logic, as a data driven approach,
and COGNET (Cognitive Network of Tasks, Chi
Systems Inc.), as a cognitive modeling approach, are the
platforms of choice for virtual UNREP and will provide
some guidance as to the value in using these powerful
tools for performance measure extraction.
This is likely where one of VE’s great potential can be
realized — as effectual and inexpensive generators of
performance indicators, monitors of performance, and
ultimately providers of performance evaluation. As these
data mining cognitive modeling tools continue to
develop, their integration within VE, particularly VE
training systems, may prove to be the cornerstone in the
revolution in training.
Cowden, A.C., Burns, J. J., & Patrey, J. (2000). Datadriven knowledge engineering. To be presented at
the Interservice/Industry Training, Simulation, &
Education Conference.
Crenshaw, R.S., Jr., CAPT, U.S. Navy (Retired). (1975).
Naval Shiphandling. Annapolis, Maryland: Naval
Institute Press.
Davidson, S. (1996). Software Design of a Virtual
Environment Training Technology Testbed and
Virtual Electronic Systems Trainer (Technical
Report 96-002). Orlando, FL: Naval Air Warfare
Center Training Systems Division.
Davidson, S. (1997a). Design of an Open Water
Shiphandling Software Testbed (Technical Report
97-007). Orlando, FL: Naval Air Warfare Center
Training Systems Division.
Davidson, S. (1997b). Development of a virtual
environment software testbed using commercial off
the shelf software components. In Proceedings to
NATO VR Conference, (December, 1997), Orlando,
Kennedy, R.S., Lane, N.E., Burbaum, K.S., & Lilienthal,
M.G. (1993). A simulator sickness questionnaire
(SSQ): A new method for quantifying simulator
sickness. International Journal of Aviation
Psychology, 3 (3), 203-220.
Lane, N.E. & Kennedy, R.S. (1988). A New Method for
Quantifying Simulator Sickness: Development and
Application of the Simulator Sickness Questionnaire
(SSQ). (Technical Report EOTR 88-7). Orlando, FL:
Essex Corporation.
Martin, M.K., Sheldon, E., Kass, S., Mead, A., Jones, S.,
& Breaux, R. (1998). Using a virtual environment to
elicit shiphandling knowledge. Proceedings to '98
I/ITSEC Conference, Orlando, FL, December 1998.
Patrey, J., Sheldon, E. M., Breaux, R.B., & Mead, A. M.
(2000). Quantifying performance of a dynamic
shiphandling perceptual-action task. Technical
Report in preparation.
Pew, R. W., Mavor, A. S., Modeling human and
organizational behavior: application to military
simulations, National Academy Press, Washington,
D.C. 1998.
Appropriate Use of Virtual Environments
to Minimise Motion Sickness
Willem Bles & Alexander H. Wertheim1
TNO Human Factors
Kampweg 5
3769 DE Soesterberg
The Netherlands
1. Introducton
With the current fast rate of technological developments
and the high requirements for training with sophisticated
apparatus, the military has become more and more
involved in working with simulators. The term
“simulator” here means: a systems that has the potential
to create sensations of passive or active self movement
in a simulated environment. This definition of the term
“simulator” not only applies to the traditional flight
simulators, both with and without a moving base, but
also to Virtual Environments (VE) set-ups implemented
in Head Mounted Display (HMD) systems, which no
doubt will become part of future flight training
Apart from the obvious usefulness of such simulators,
they also have a serious disadvantage: it turns out that
they expose users to discomforting and unwanted sideeffects, that might well affect training efficiency. One of
the most important and well known problems is that
these simulators often induce motion sickness, which
severely interferes with behaviour and thus with training.
Motion sickness causes lowering of motivation, usually
resulting in a considerable slowing down of work rate, a
disruption of continuous work, or even its complete
abandonment. In fact, motion sickness in simulators is
currently the main factor limiting the use of simulators.
There are various kinds of motion sickness, such as air
sickness, sea sickness, car sickness, space sickness, and
some people may even get sick in trains or elevators.
Simulator sickness is basically a form of motion
sickness. It has been defined as motion sickness which
occurs in a simulator, but which would not occur in the
real world in the same circumstances as those which are
simulated [28]. For instance, if a person gets sick in an
aeroplane and also in a simulator, which validly mimics
the flight movements, then this would not classify as
simulator sickness. We only speak of simulator sickness
if that person would become sick in the simulator but not
in the aeroplane. The same reasoning applies to motion
sickness in virtual environments.
In order to be able to minimise the incidence of motion
sickness in virtual environments, it is necessary to
understand the reasons for simulator sickness, and thus
for motion sickness in general. Therefore we will briefly
review our present view on motion sickness. This will
then allow us to understand why some factors are
important to lower the motion sickness incidence in
virtual environment applications.
Finally we will discuss other, often related, human factor
problems that happen frequently in virtual environments,
such as headaches, eye strain and after-effects, and
mention what might be done to minimise these effects.
2. Motion sickness in general
Motion sickness may vary among subjects: within
individuals, there is no direct correlation between
sensitivity to various forms of motion sickness.
Sensitivity to any particular form of motion sickness also
varies largely among humans. Moreover, motion
sickness may develop fast or slow. Women are generally
somewhat more sensitive than men. There seems to be
an effect of age as well. Sensitivity for motion sickness
is very low with children a few years old, then increases
and at old age decreases again [36].
It is known that, after its initial rise, motion sickness
eventually decreases with time despite ongoing motion
exposure. This adaptation may take a few hours up to a
few days, as with sea or space sickness. But again, the
time it takes for the symptoms to disappear differs
among individuals. With approximately 5% of
humankind adaptation does not take place at all.
All this makes it difficult to understand the nature of the
provocative motion stimulus. In a series of experiments,
carried out in a Ship Motion Simulator (SMS),
McCauley et al. suggested that it is mainly the vertical
component of ship motion that causes sea sickness [34].
For sinusoidal vertical motion they found motion
sickness to be most prominent between 0.05 to 0.8 Hz
(maximum at 0.2 Hz) and with amplitudes of over 1
m/s², the incidence of motion sickness increasing further
at higher amplitudes. On the basis of their data these
authors developed a descriptive mathematical model of
sea sickness [31]. More recently another mathematical
motion sickness incidence model has been proposed by
Griffin, allowing also for complex vertical motion
patterns [23] (for comparison of these two models, see
[16, 17]). These models became the basis for the
international standards. The main premise of these
descriptive models is that varying vertical accelerations
are an important factor in the generation of motion
For contact with authors: [email protected] and [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
Fig. 1 Instead of the conflict vector c from the Oman model [35], the subjective vertical motion sickness model
considers the vector d to be the conflict vector for generating motion sickness. The modules V are necessary for the
computation of the subjective vertical (thick lines). The dotted lines represent the internal model.
It has also been shown that motion sickness may develop
as a result of horizontal linear movements [22, 25].
Furthermore, the notion that only vertical movements are
sea sickness provoking, has been challenged in a series
of experiments by Wertheim with a ship motion
simulator [43]. He observed motion sickness even when
vertical movements — which, according to the above
mentioned mathematical theories, were too weak to
generate motion sickness — were accompanied by low
frequency pitch and roll motions. Head movements still
further increased the motion sickness incidence [42].
Ergonomic measures to counter motion sickness at sea
included the design of a working place such that head
movements could be minimised [4].
Head movements which changed the orientation of the
head with respect to gravity also proved to be very
provocative in subjects who had been submitted
previously to constant hyper gravity in a human
centrifuge (2–3 g for 1.5 hrs [7, 33]).
These examples illustrate the view that the vestibular
system plays a crucial role in the generation of motion
sickness [21, 36]. In fact, it has long been known that the
one necessary requirement for any kind of motion
sickness is a functioning vestibular apparatus. People
who do not have a functioning vestibular apparatus
(because of particular illnesses) simply cannot become
motion sick [e.g. 29].
However, vestibular-visual interactions are also very
important in provoking or preventing motion sickness:
the driver of a car does not get sick, whereas the
passenger reading in the back seat may have a fair
chance of getting sick on a curved road. Somatosensoryvestibular interactions also prove to be important in the
incidence of motion sickness as was demonstrated with
(Pseudo-)Coriolis effects [6]. Especially with VE these
interactions are very important.
This is not the proper place to present a detailed
description of how the vestibular apparatus works. Many
good texts on the subject ore available elsewhere (e.g.
Guedry [24], or Howard [26]. Here it suffices to note
that the central role of the vestibular system is
recognised in what are currently the most well known
explanatory theories of motion sickness, like the theory
of intersensory mismatch [35, 36].
A more specific version of this theory assumes that
motion sickness results from only one mismatch, the one
between the expected vertical and the vertical as
determined on the basis of the incoming sensory
information [8]. There are some other alternative
theories based on ecological perspectives [39, 44], and
there are ideas about cognitive influences on motion
sickness [19], but here we will focus primarily on the
view that motion sickness arises when there is a
mismatch in the determination of the gravity
According to the sensory mismatch theory from Reason
and Brand [36], motion sickness occurs when the
sensory systems provide the brain with more than one
kind of self-motion information which do not match each
other. This could be either an intra- or an inter-sensory
The sensory mismatch theory offers some remedies for
motion sickness. For example, in the case of a ship at
sea, the incidence and severity of sea sickness under
deck should be reduced when the visual system is
provided with an optic pattern which remains stable not
relative to the eyes, but relative to the real world. This
was proposed by Bittner & Guignard [4] and it fits the
experience that standing on deck with view of the
horizon is less provocative than standing under the deck.
In fact there have been some attempts to investigate
possible motion sickness reducing effects of an artificial
horizon [10, 38].
3. The subjective vertical mismatch concept
Although many examples of conflicts between and
within sensory systems can be described, leading to
disorientation and motion illusions indeed, there is
plenty of evidence that motion sickness is primarily
provoked in those situations where the determination of
the subjective vertical, the internal representation of
gravity, is challenged. Therefore the sensory
rearrangement theory on motion sickness was redefined
to: “All situations which provoke motion sickness are
characterised by a condition in which the sensed vertical
as determined on the basis of integrated information
from the eyes, the vestibular system and the nonvestibular proprioceptors is at variance with the
subjective vertical as predicted on the basis of previous
experience” [8, 9]. In Fig. 1 this concept is illustrated.
Since with this model in principle motion sickness
incidence can be described for every stimulus condition,
such an approach would be more useful than the
descriptive models as discussed above for sea sickness,
since these descriptive functions only apply to particular
stimuli which have to be determined first. We therefore
explain this model in more detail for the situation of
walking towards a certain position.
In Fig. 1 we see that, in order to obtain the desired
position xd, muscle activity (m) is generated leading to a
position x due to the body dynamics (B). This signal,
together with the external noise ne, is detected by the
senses (S) resulting in sensory information a. The
internal model consists of the same components
(indicated with a hat) and computes the expected sensory
information â. Differences between the vectors a and â
are calculated, and are fed back into the system. In this
way an optimal estimate of the actual walking path can
be obtained.
The Subjective Vertical conflict model extends the
Oman model [35] with a network V which constructs the
sensed vertical, vsens, based on the incoming sensory
information. Similarly, in the internal model a network V
^ or
is added which constructs the expected vertical, v
vexp, based on previous experience and expectation. The
difference vector d between vsens and vexp is used to
update vexp, and is in our view the conflict vector which
generates motion sickness [8] (see Fig. 1).
For analysis of the provocativeness of motion conditions
it is of great importance to know how the representation
of the vertical is accomplished [8, 11].
This is in fact the basic vestibular problem for the central
nervous system. In Fig. 2 it is shown how this could be
accomplished on the basis of psycho-physiological
evidence. The vestibular (semi-circular canals, SCC, and
otoliths, OTO), the visual (VIS) and the somatosensory
(SOM) system all provide information on spatial
orientation. In order to obtain only one unique spatial
orientation it is assumed that all this sensory information
is integrated (INT) into basically three signals, indicating
the sensed rotation (SR), the sensed translation (ST) and
the sensed vertical (SV) as shown in Fig. 2 [8].
The integration of rotatory motion information is rather
straightforward, because the sensory systems provide
complementary information. A more complex problem
for the central vestibular system is to extract the gravity
information out of the sensed gravito-inertial force
vector. In view of normal human movements and
locomotion, it was hypothesised that low-pass filtering
(LP) of the signal representing the gravito-inertial force
vector could preserve gravity. This is a sensible
approach, provided that the angular motion information
helps to compensate for the consequences of fast head
tilts. Mathematically this compensation is accomplished
by a transformation R of the co-ordinate frame with the
otolith vectors, over the angle of the head tilt indicated
by the rotation sensors. Such a manipulation keeps the
input to LP unchanged, the sensed vertical after the head
tilt being determined by the rotatory motion information
due to the inverse transformation R-1 as shown in Fig. 2.
Fig. 2 Integration of sensory information.
Fig. 3 Differential effects of congruent and incongruent visual and somatosensory motion stimulation on the magnitude
of the vestibular Coriolis effect (5 is the standard magnitude of the discomfort of the vestibular Coriolis effect).
It is assumed that the internal model uses a similar neural
network as on the sensory side, the values of the
different parameters being determined by previous
experience. To illustrate this point, the observation is of
interest that experienced pilots are suffering less from
motion sickness in real flight than student pilots,
whereas they are more prone to simulator sickness than
student pilots. The internal model of an experienced pilot
apparently has parameter settings that match quite well
the motion signals which are sensed by the sensors
during real flight, but they do not match to the
information as sensed by the sensors in, for instance, a
fixed-base simulator environment. For student pilots the
argument goes the other way around: they have no
particular experience as for the in-flight environment.
Thus, in the simulator the match is better then during
real flight, where they sense motion signals which are
not expected.
To summarise, difference vectors between sensed and
expected linear and rotatory motion are not a trigger for
motion sickness: this may only result in disorientation.
Only differences between the sensed and expected
vertical provoke motion sickness.
This is illustrated in modern architecture where fully
listed buildings are popping up more and more: In a
stationary listed environment (visual frame information
not coinciding with the gravity vector) head movements
were found to be provocative to motion sickness. This
was described by Kitahara & Uno [30] and we
confirmed this observation: Walking in a stationary
listed environment (max. 20 degrees) made about 10%
of the subjects motion sick within 15 minutes. Especially
the turning around proved to be provocative [11]. In this
condition it is noteworthy that a stationary subject
doesn’t get motion sick, despite the continuing
conflicting information from the visual frame and the
otoliths about the direction of the vertical. According to
the SV conflict model, the sensed and the expected
attitude converge due to the feedback in this situation:
Only when the subject starts to move around, differences
are to be expected between these two vectors. This is a
common observation in many motion sickness provoking
surroundings: Moving around or making head
movements enhances motion sickness (see section 2).
4. Factors causing nausea in virtual environment
Somatosensory-visual-vestibular interactions. With
the principle of the subjective vertical mismatch one can
analyse the different virtual environments concepts on
provocativeness for motion sickness. To illustrate the
meaning of this concept, the results of laboratory
experiments which are of direct relevance for the use of
HMD and VE system concepts, are shown in Fig. 3. The
results stem form experiments done by Brandt et al. [14]
and Bles [5].
In these experiments the magnitude of the nausea of the
Coriolis effect obtained by lateral head tilt during
constant velocity rotation at 60 V ZDV VWXGLHG XQGHU
different visual and somatosensory stimulus conditions.
The pure vestibular Coriolis effect, head tilt in darkness,
served as a reference and had a magnitude of 5. It shows
that the Coriolis effect is minimal if there is sight on the
earth-stationary visual surround. This is comparable to
walking conditions with a HMD with a perfect earthstationary virtual environment. The nausea increases if
the visual surround rotates together with the chair, which
is compatible to rotating with a HMD with a head fixed
display. If the surround rotates with twice the chair
velocity, the nauseating effect of a head tilt is very
strong. This demonstrates what happens if the HMD
provides non-earth referenced motion information.
Inspection of the right frame in Fig. 3 indicates that
manipulation of somatosensory motion information as
obtained by stepping in circles in darkness provides
similar results as manipulation of the visual information
when sitting. If in these conditions the visual and
optokinetic motion information is combined, the
modulation of the vestibular Coriolis effect is even more
pronounced (Fig 3 right frame, open dots). This shows
how important it is to take into account the
somatosensory information, if present, otherwise the
analysis may lead to predictions which are completely
different from the experimental data. The Subjective
Vertical model as shown before fully accounts for these
experimental results [6]. The SV model also perfectly
applies to the concept of the closed cockpit aircraft [11].
Fixed base vs. moving base. In order to minimise
simulator sickness for HMDs with virtual environment
applications, the same rules apply as for fixed and
moving base simulators. Moving in a virtual
environment of a HMD may be accomplished by turning
or walking on a treadmill, or by means of a joy-stick.
These changes of propagation means, together with
irregular motion velocity patterns using the joy-stick
may even be more demanding from the human
equilibrium system than a normal 6DoF flight simulator.
In fact, keeping in mind the frequency characteristics of
the different parts of the equilibrium system, the model
in Fig 1 may help to analyse the stimulus patterns on
their provocativeness to motion sickness. It is no surprise
that a HMD training facility on board of a moving
platform with motion which has absolutely nothing to do
with the training scenario, is due to be more provocative
than on a non-moving platform.
Destabilisation of the visual world. If one makes a
head movement while wearing a HMD, the image in
front of the eyes will move with the head. In other
words, in such situations the visual world loses its
stability [40].
An additional complication here is that when we make a
rotatory head movement, the eyes rotate in the head in
counter direction. This so called Vestibulo-Ocular
Reflex (VOR) normally serves to maintain ocular
fixation on an object in our environment during head
movements. The VOR is very fast and has a latency of
approximately 10 ms. However, to maintain ocular
fixation when the object moves with the head (as in a
HMD) the VOR must be suppressed. The necessary
enervation of the ocular musculature is relatively slow
and frequency specific. With head movements up to 1Hz
the VOR can be properly suppressed, but at higher
frequencies the VOR dominates, blurring the visual
image on the retinas and causing visual discomfort. If the
blur stems from very fast retinal motion its direction
cannot be perceived, which may have consequences for
the computation process as indicated in Fig. 2.
Image magnification (or minimisation). Similar
problems may occur when an outside image, projected
inside an HMD (e.g. the image of a night vision goggle),
is magnified or minimised. Normally the VVOR (VOR
with full sight on the visual surround) has a gain of 1,
which means that the velocity and amplitude of this
reflexive eye movement is equal to that of the
counterdirective head movement. If the head movement
is fed back to move a magnified image in a HMD in
counter direction to the head, the velocity of the image
shift is higher than expected, while with a minimised
image it is lower. This means that the visual information
contributing to the computation in Fig. 2 may not
properly match the vestibular inputs in the computations,
which may also lead to discrepancies in the
determination of the representation of gravity. Such
situations resemble the case where one scans the scenery
with binoculars in which case the visual image moves
across the retinas with a much higher speed than is
normally the case during head movements. The same
happens when wearing new spectacle glasses. But since
glasses are usually worn continuously the visual
vestibular interaction may adapt back to normal in due
time. However, as long as such adaptation is not
complete nausea might persist. Unless one wears a HMD
for quite a long time similar adaptation may not easily be
obtained. Thus it is recommended not to use a
magnification or minimisation factor in the design of VE
or HMD visuals with outside image representations.
There is another discomforting problem related to image
magnification or minimisation. The point is that when
we stand upright we normally make small body
movements (body sway). Here the visual system helps. It
feeds these small retinal image shifts back to the system
which maintains body posture. When those image shifts
do not really correspond to how the body really moves
(because of their optical magnification or minimisation),
they are still fed back to our musculature with which we
maintain our postural equilibrium. Thus we may end up
making much larger body sway motions, which poses a
threat to our postural stability and may create feelings of
insecurity with respect to our equilibrium (in fact it is
this mechanism which causes fear of heights — in which
disproportionally small, because of the very far distance
of objects in the visual environment [13].
Time delays. In many VE simulations head movements
are fed back to the visual display, with the purpose of
moving the image across the display in the direction
counter to the head movement. This should ensure that
the virtual environment remains stationary relative to
earth (i.e. relative to gravitation and compass-fixed)
during head movements. However, in many simulators,
including HMD-systems, this coupling is less than
perfect, which may cause severe nausea. The point is
that the visual image must move across the display
surface in precise temporal synchrony with the
movements of the head. Otherwise a phase difference
between the visual and vestibular inputs to the CV and
SV occurs which may cause them to deviate from each
other, causing severe nausea. However, it always takes
time to record (and filter) head movements and to
calculate the movements of the image inside the HMD
on the basis of these records. This manifests itself in a
temporal delay of the required visual image changes,
especially with very large and detailed displays. During
the delay period there is a large discrepancy between
visual and vestibular information: the head movements
are properly registered by the vestibular system, but the
visual world moves with in stead of against the head.
Even with delays as brief as 46 ms, the resulting
visual-vestibular mismatch, which may easily cause a
CV vs. SV mismatch, may already be extremely
nauseating [20, 27].
This reasoning is in line with empirical results from
recent experiments in which the gain and phase relations
of visual and vestibular information were manipulated
separately, using an artificial environment set up
mounted on a sled for linear motion [32]. The data
clearly suggested that phase differences are much more
provocative than gain differences, and that, in
contradistinction to visual phase-leads (relative to the
vestibular stimulus), small visual phase-lags are already
highly provocative.
Vection. Visually induced sensations of self-movement,
known technically as “vection”, are of course key
phenomena in simulators. However, since visual
suggestions of self-motion may easily affect the Sv
through the integration INT with the SCC and SOM
information (see Fig. 2), they always form a potential
risk of motion sickness. In this section we will review
the properties of visual displays and images that affect
vection, and which thus have to be considered in
evaluating the risk for the development of nausea in
Screen size. Vection is strongest with peripherally
moving visual flow fields. Hence, large screens carry
higher risks of motion sickness. With full-field flow
fields almost everyone will experience strong sensations
of vection. Thus as a general rule, the smaller the visual
image (or display) the lower the chance of motion
sickness. From laboratory experiments it has been
concluded that the risk of vection is minimal with
images extending a visual angle less than approximately
30°. A normal standard 17 inch computer screen viewed
at a distance of 50 cm encompasses 34° and therefore
will not easily generate vection.
Foreground/background. A necessary condition for
vection to occur is that the inducing visual pattern is
perceived and interpreted as a background. Normally,
when walking past an object which we fixate with our
eyes, its background moves in our visual periphery, the
central area of the visual field is occupied by the
retinally stationary object. However, when we move in a
vehicle (e.g. a car), the situation is reversed. Here the
peripheral parts of our visual field are occupied with
objects that remain stationary on the retinas (e.g. the
hood of the car, the frame of the windshield, the
dashboard etc). In such situations vection is caused by
image motion across the central area of the visual field.
Experiments have shown that such centrally evoked
vection is possible only if the visual flow is perceived as
background, that is, as further away in depth than the
stationary objects in the periphery. Hence, in exception
to the above mentioned rule, visual patterns covering
small visual angles may still evoke vection if they are
perceived as a background. Thus small displays in
simulators, which simulate “out-of-the-window” views
may facilitate vection.
Pattern motion. As should be clear by now, moving
visual patterns always carry with them a certain chance
that vection develops. With a constant velocity pattern
vection normally develops with a latency between up to
20 seconds (depending on various stimulus parameters)
after which vection velocity does not increase any
further and the pattern appears earth stationary. At this
point vection is said to be saturated. The forcefulness
with which vection is experienced and the perceived
velocity of vection depend not only on the size of the
vection inducing pattern, or on whether or not it is
perceived as a background, but also on its velocity.
Perceived vection velocity increases with the velocity of
the stimulus pattern up to approximately 60º/s, after
which it is reduced rapidly and the visual pattern is
perceived as unstable or just moving.
Vection also depends on the motion frequency of the
inducing pattern. As mentioned above, its latency can be
relatively long, implying that low frequencies are more
powerful than high frequencies. With sinusoidal pattern
motion frequencies up to 0.1 Hz vection can normally be
induced. At higher frequencies vection rapidly decreases.
Thus if one wants to prevent vection it is important to
keep this cut-off frequency of 0.1 Hz in mind.
5. Other discomfort factors in head mounted displays
Image flicker. Typical computer work complaints such
as eye-strain, visual fatigue, headache and blurred vision,
are common also when working with HMDs. The reason
for these complaints are not always clear, but one of the
causes often suggested is image flicker. Our sensitivity
for image flicker is higher in the visual periphery than in
the central visual field. Causes for image flicker are long
times needed for computing the motion of images in the
HMD (update frequency), especially when these
computations must be carried out on the basis of on-line
head movement registrations, and the refresh rate of the
particular screen used in the HMD. It is advisable to
avoid screens that have a refresh rate of less than 80 Hz.
Traditional video screens are too slow (50 Hz). To
reduce the risk of perceiving flicker it is also advisable to
reduce the luminance of the images in the HMD to less
than 50 cd/cm2 and to keep luminance contrasts
relatively low as well.
Image acuity and depth perception. Bad image acuity
may also yield complaints of headache and eye-strain,
especially when text has to be read. Image screens
should have a resolution at least comparable with that of
a 1024 × 768 pixel 17 inch computer monitor.
Traditional video screen technology has too low a
resolution to be acceptable in HMDs, especially with
wide angle screens.
With 3D VR systems, the two eyes receive separate and
slightly different images, which are fused by the brain to
perceive depth. It is advisable to facilitate the fusion
process as well as possible, by positioning the image
optically at 2 to 4 m distance from the eyes. The
necessary ocular accommodation is then 0.5 to 0.25
dioptres and the necessary convergence of the eyes then
covers 0.9 to 0.4 degrees of visual angle. If the two
images are not placed at the correct position relative to
the eyes eye-strain will result from the additional oculomuscular effort required.
To keep a reasonable visual acuity in such 3D VR and
HMD systems, the following criteria should apply with
respect to corresponding details in the two images
(correct adjustments of the rims of the images is less
• The (rotational) difference between corresponding
details should not exceed 1°.
• The vertical position of corresponding details should
not exceed 0.5°.
• Divergence between corresponding details should be
no more that 0.5°.
• The size of corresponding details should not differ by
more than 3%.
• The difference in required accommodation of the two
eyes should not exceed 0.25 dioptres.
However, if the head movements of the individual inside
the simulator are not fed back to affect the image on the
display in a similar manner, the concurrent vestibular
sensations may not always match the changes in the
image. For example, imagine a person inside such a
simulator who moves the head closer to a visual display
unit that is supposed to simulate a window through
which a visual outside scene is seen. The eyes then get
closer to the screens. In normal situations more of the
visual environment will then become visible from behind
the rims of the window and the size of the retinal images
of far away objects will not change much. Conversely, if
such a forward head movement is made in a simulator,
where the observer’s head position is not fed back to the
image on the screen, no new parts of the environment
will become visible from “behind” the rims of the screen
and the images of all virtual objects will be enlarged
equally on the retinas, whatever their distance. Therefore
the changes in the visual information will not match the
vestibularly sensed head movements. This may cause
visual discomfort and, if lasting long enough, eye-strain
and headache. If that visual-vestibular mismatch
includes aspects of the subjective or sensed vertical, a
risk of motion sickness may evolve as well.
Smoothness of image motion. To avoid headaches and
eye strain in simulators it is necessary that smooth visual
motion will indeed be perceived as smooth. This is not
always the case. The same factors apply here as those
which cause flicker. When calculations necessary for
generating moving images take relatively much time
(low update rate), or screen refresh rate is low, the
movements will be seen as consisting of small steps.
This is visually quite discomforting.
Control device system lag. When using a computer
mouse, a joy stick, roller ball or any other control device
to affect the image on a visual display in a simulator
which is used in an interactive man-in-the-loop mode,
performance may be affected when delays between the
action and its effect on the screen become too large.
Such delays are not discomforting in the sense that they
might cause motion sickness, headaches etc, but they
may well have a deteriorating effect on tracking and
steering performance.
No hard limits can be given for maximum lags because
they also depend on the kind of vehicle model used in
the simulator (see for a review: Ricard [37]). However in
general steering performance is assumed to deteriorate
when control device system lags increase beyond 100 ms
[1, 2], while lags over 300 ms may induce oscillations
[3]. With respect to normal computer use, lag times for
the use of a mouse should not become larger than 50 ms,
while the lag between pressing a key on a key-board and
the appearance of a letter on the display should not be
longer than 100 ms (DERA defence standards [18]).
Motion parallax. On the flat surface of visual displays
there is no real depth. It must be simulated. Not only by
proper perspectives which change during simulated ego
motion, but more importantly, by concurrent relative
motion between the objects in the surroundings (motion
parallax). If motion parallax is not properly
programmed, it may create impressions of self motion
which do not properly fit vestibular cues from the
motion base. For example, most simulator systems use
visual display systems in which the movements of the
vehicle (e.g. an air plane) are fed back to change the
visual image on the display in such a manner that it
appears stationary with respect to the real world.
After-effects. When trainees spend many hours inside a
simulator there is a risk of after- effects once they exit
the simulator. Such after-effects include not only a
continuation of nausea, but also postural imbalance and
headaches (see for a review: Wertheim [41]). They may
have negative effects on performance in normal
everyday behaviour (e.g. driving), or may aversely affect
special skills such as are involved in flying an air plane.
This issue has been recognised in the literature as having
juridical consequences for those responsible for
simulators and trainees. They might find themselves
liable if trainees cause accidents after a simulator
training. Only recently has research started on such after-
effects and currently there is not much specific
information available as to their exact nature and the
risks involved. However, after-effects may last for many
hours [3].
6. Conclusions
Head Mounted Displays still easily provoke discomfort.
The known visual problems in using HMDs which are
due to the technical limitations of the display and
computing limitations, will most probably be solved by
technical improvements in the near future. As long as
that is not the case, the factors described in section 5
should be taken into account.
In developing HMD application concepts one should be
aware of the motion sickness consequences of
orientation cues which lead to false visual verticals,
because of the fact that a discrepancy between the sensed
and expected representation of gravity is considered to
be the primary motion sickness provoking conflict.
Qualitative analysis with the model on the
provocativeness of the application taking into account
what is known on the sensory interactions is very useful
already. Quantitative analyses by Bos & Bles [12] have
shown that the model accounts for the sea sickness data
of O’Hanlon and McCauley [34]. This is a very
promising accomplishment, since the international
standards (see section 2) are based on descriptive
Allen, R.W. & DiMarco, R.J. (1984). Effects of
transport delays on manual control system
performance. Paper presented at 20th Annual
Conference on Manual Control, NASA Ames
Research Center, 12-14 June, 1984. System
Technology, Inc., 13766 South Hawthorne
Boulevard, Hawthorne, CA., USA.
Bailey, R.E., Knotts, L.H., Horowitz, S.J. &
Malone, H.L. (1987). Effect of time delay on
manual flight control and flying qualities during inflight and ground-based simulation. AIAA paper
87-2370 at the AIAA Flight Simulation
Technologies Conference, New York. American
Institute of Aeronautics and Astronautics. Calspan
Advanced Technology Center, Buffalo, New York,
Baltzley, D.R., Kennedy, R.S., Birbaum, K.S.,
Lilienthal, M.G. & Gower, D.W. (1989). The time
course of postflight simulator sickness symptoms.
Aviation Space and Environmental Medicine, 60,
Bittner, A.C. & Guignard, J.C. (1985). Human
factors engineering principles for minimizing
adverse ship motion effects: theory and practice.
Naval Engineers Journal 97-4:205-213
Bles, W. (1981). Stepping around: circular vection
and Coriolis effects. In: Long, J.; Baddeley, A., eds.
Attention and Performance IX. Hillsdale, NJ:
Lawrence Erlbaum Ass.
Bles, W. (1998). Coriolis effects and motion
sickness modelling. Brain Research Bulletin, 47
(5): 543-549.
Bles, W., Bos, J.E., Furrer, R., De Graaf, B.,
Hosman, R.J.A.W., Kortschot, H.W., Krol, J.R.,
Kuipers, A., Marcus, J.T., Messerschmid, E.,
W.J.Ockels, W.J.Oosterveld, J.Smit, A.H.Wertheim
& C.J.E.Wientjes (1989). Space Adaptation
Syndrome induced by a long duration +3Gx
centrifuge run. Rept. IZF-1989-25, TNO Human
Factors Research Institute, Soesterberg, The
Bles, W., Bos, J.E., De Graaf, B., Groen, E.L. &
Wertheim, A.H. (1998). Motion sickness: Only one
provocative conflict? Brain Research Bulletin,
Bles, W., Bos, J.E. & Kruit, H. (2000). Motion
Sickness. Current Opinion in Neurology 2000, 13,
p. 19-25.
Bles, W., De Graaf, B., Keuning, J.A., Ooms, J., De
Vries, J. & Wientjes, C.J.E. (1991). Experiments on
motion sickness aboard the M.V. "Zeefakkel". Rept
IZF-1991-A-34, TNO Human Factors Research
Institute, Soesterberg, The Netherlands.
Bles, W. & Tielemans, W.C.M. (2000). Motion
sickness consequences of flying closed cockpit
aircraft. Countering the directed energy threat: are
closed cockpits the ultimate answer? RTO Meeting
Proceedings 30 (RTO-MP-30 AC/323(HFM)
TP/10). 22:1-6
Bos, J.E. & Bles, W. (1998). Modelling Motion
Sickness. RTO/HFM’s Aircrew safety Assessment
Brandt, Th. (1999). Vertigo: its multisensory syndromes. -2nd ed. Springer Verlag, Berlin Heidelberg New York
Brandt, Th., Wist, E. & Dichgans, J. (1971).
Optisch induzierte Pseudocoriolis-Effekten und
Nervenkrankheiten, 214, 365-389
Conklin, J.E. (1957). Effect of control lag on
performance in a tracking task. Journal of
Experimental Psychology 53(4):261-268.
Colwell, J.L. (1989). Human factors in the naval
environment: a review of motion sickness and
memorandum 89/220, Canadian National Defence
Research Establishment Atlantic. Dartmouth.
Colwell, J.L (1994). Motion sickness habituation in
the naval environment. DREA Technical
Memorandum 94/211, Canadian National Defence
Research Establishment Atlantic. Dartmouth.
DERA Defence Standard 1996. Defence standard
00-25 (part 13)/Issue 1, 24 May 1996; Human
Factors For Designers Of equipment, Part 13:
Human Computer Interaction. DERA, Malvern,
Worcs, WR14-3PS, UK.
Dobie, T.G. & May, J.G. (1994). Cognitivebehavioral management of motion sickness.
Aviation Space and Environmental Medicine, 65,
10, section II, C1-C20.
Draper, M., Viire, E., Furness, T.A. & Parker, D.E.
(1998). Theorized relationship between vestibuloocular adaptation and simulator sickness in virtual
environments. Proceedings of the International
Workshop on Motion Sickness: Medical and
Human Factors, Marbella, Spain.
Eyeson-Annan, M., Peterken, C., Brown, B. &
Atchison, D. (1996). Visual and vestibular
components of motion sickness. Aviation, Space
and Environmental Medicine, 67, 10, 955-962.
Golding J.F. & Kerguelen, M. (1992). A
comparison of the nauseogenic potential of low
frequency vertical versus horizontal linear
oscillation. Aviation, Space and Environmental
Medicine, June issue, 491-497
Griffin, M.J. (1990). Handbook of human vibration.
Academic Press, London.
Guedry, F.E. (1974). Psychophysics of vestibular
sensation. In: H.H. Kornhuber (ed): Handbook of
sensory physiology Vol 6/2, Springer, NY. 3-154.
Horii, A., Takeda, N., Morita, M., Kubo, T. &nd
Matsunaga, T. (1993). Acta Otolaryngologica,
Suppl 501, 31-33
Howard, I.P. (1986). The vestibular system. In:
Boff, K.R., Kaufman, L. and Thomas J.P. (eds):
Handbook of Perception and Human Performance,
Vol I: Sensory processes and perception. John
Wiley, NY.
Kalawsky, R.S. (1993). The science of virtual
reality and virtual environments. Addison-Wesley
Kennedy, R.S., Berbaum, K.S., Allgood, G.O.,
Lane, N.E., Lilienthal, M.G. & Baltzley, D.R.
(1988). Ethiological significance of equipment
features and pilot history in simulator sickness.
NATO-AGARD conference proceedings No. 433:
Motion cues in flight simulation and simulator
induced sickness. Neuilly sur Seine, France, 1-1/122
Kennedy, R.S., Graybiel, A., McDonough, R.C.
and Beckwith, F.D. (1968). Symptomology under
storm conditions in the North Atlantic in control
subjects and in persons with bilateral labyrinthine
defects. Acta Oto-laryngologica, 66, 533-540
Kitahara, M. & Uno, R. (1967). Equilibrium and
vertigo in a tilting environment. Annals Otol (St
Louis) 76:166-178.
McCauley, M.E., Royal, J.W., Wylie, C.D.,
O’Hanlon, J.F. & Mackie, R.R. (1976). Motion
sickness incidence: exploratory studies of
habituation, pitch and roll, and the refinement of a
mathematical model. Human Factors Research Inc.
Technical Report 1733-2.
Mesland, B.S. (1998). About Self-motion
Perception. PhD thesis, University of Utrecht. The
Ockels, W.J.R., Furrer, R. & Messerschmid, E.
(1990). Space sickness on earth. Experimental
Brain Research, 79, 61-663.
O’Hanlon, J.F. & McCauley, M.E. (1974). Motion
sickness incidence as a function of the frequency
and acceleration of vertical sinusoidal motion.
Aerospace medicine, 45, 366-369.
Oman, C.M. (1982). A heuristic mathematical
model for the dynamics of sensory conflict and
motion sickness. Acta Oto-Laryngologica, suppl.
Reason, J.T. & Brand, J.J. (1975). Motion Sickness.
Academic Press. London.
Ricard, G.L. (1994). Manual control with delays: A
Bibliography. Computer Graphics, 28, 2, may
1994, 149-154.
Rolnick, A. & Bles, W. (1989). Performance and
well-being under tilting conditions: the effects of
visual reference and artificial horizon. Aviation
Space and Environmental Medicine, 60, 779-785.
Stoffregen, T.A. & Riccio, G.E. (1986). Out of
control: an ecological perspective on motion
sickness. Paper presented at the Fall meeting of the
International Society for Ecological Psychology,
October 18, Philadelphia, PA, USA.
Wertheim, A.H. (1994). Motion perception during
self motion: the direct versus inferential
controversy revisited. Behavioral and Brain
Sciences 17: 293-355.
Wertheim A.H. (1999). The assessment of
aftereffects of real and simulated self-motion:
motion sickness and other symptoms. Rept. TNOTM 1999-A074, TNO Human Factors Research
Institute, Soesterberg, The Netherlands.
Wertheim, A.H., Heus, R., & Vrijkotte, T.G.M.
(1995). Human energy expenditure, task
performance and sea sickness during simulated ship
movements. TNO- Report TM-1995-C-29; TNO
Human Factors Research Institute, Soesterberg,
The Netherlands.
Wertheim, A.H., Wientjes, C.J.E., Bles, W. & Bos,
J.E. (1995). Motion sickness studies in the
TNO-TM Ship Motion Simulator (SMS).
TNO-Report TM-1995 A-57, TNO Human Factors
Research Institute, Soesterberg, The Netherlands.
Yardley, L. (1992). Motion sickness and
perception: a reappraisal of the sensory conflict
approach. British Journal of Psychology, 83, 449471.
Keynote Address 2: Available Virtual Reality Techniques
Now and in the Near Future
(Unclassified and for distribution to all NATO nations)1
Grigore C. Burdea
Rutgers University
Human-Machine Interface Laboratory/CAIP
96 Frelinghuysen Rd.
Piscataway, NJ 08854-8088
This paper presents available virtual reality technology
as well as technology that is projected to become
available to NATO in the near future. Areas discussed
are new PC technology (graphics rendering and wearable
computers), personal and large-volume displays, large
volume tracking, force feedback interfaces, and software
toolkits. PCs presently render millions of polygons/sec.
Their reduced cost makes possible the distribution of
virtual environments at many sites and in many
countries. Large-volume displays are more expensive,
but allow more natural user interactions. They do require
large-volume tracking that is fast and accurate. Haptic
interfaces are a recent class of input/output devices that
increase simulation realism by adding the sense of touch.
This comes at a cost of more computing power and
better physical modeling. The modeling and programming needs of virtual reality are met by software toolkits
designed for such simulations.
1. Introduction
Virtual reality technology has experienced significant
advances in the late nineties, and now has many
characteristics that may be exploited by the military.
Virtual reality has the potential to significantly reduce
training costs and the risk to him. It also has the potential
to reduce team training costs, allowing multi-national
organizations, such as NATO, to have a unified training
system, without a unique training location. Virtual
reality, as a computerized training environment, allows
transparent gathering of data, and the remote access to
such data, at a much smaller time interval, and resolution
than allowed by manual data collection methods. For all
these reasons it is important to inform the military
decision-makers of what technology and methods are
available today, or what will become available in the
near future.
This report is based on the keynote address given by the
author at the NATO Workshop that took place in April
2000 in Hague. Then, as now, the time and space
available for such a review are limited. When trying to
condense all this material, which can easily take a
Semester to teach in college, certain things had to be
omitted. Thus the present review does not cover
networked communication as it applies to shared VR,
nor does it cover human factor trials of VR technology.
Such topics are covered in companion papers. Emphasis
here is on commercial off-the shelf technology, or
technology that is close to commercialization. Many
deserving research projects are omitted here, as a matter
of practicality. The interested reader who wants more
information on such research should consult the open
literature, such as the Proceedings of the IEEE Virtual
Reality Conference series (formerly VRAIS), and other
such publications.
Section 2 of this report presents significant changes in
the computing platforms that are (or may be) used in
VR. Section 3 describes the displays that output the
graphics scene to the user, whether such displays are
personal or large-volume. Large-volume displays, in
turn, require large-volume trackers, which are the subject
of section 4. Section 5 presents the newer haptic interfaces, which bring more realism to the simulation by
allowing the user to touch and feel virtual objects. The
modeling libraries needed by modern VR simulations
(including haptics) are detailed in section 6. Section 7
concludes this report.
2. The PC Revolution
Probably one of the most important changes that has
influenced the VR arena in recent years is the
tremendous increase in PC-based graphics rendering
speed. The closing gap between inexpensive PC-based
graphics and the high-end SGI engines is clearly
illustrated by Figure 1.
The measure of performance used for comparison here is
the number of polygons rendered by the computer in unit
time. When dividing this number by the scene complexity, one obtains the screen refresh rate in frames/
second (how many snapshots of the virtual scene the
computer can render per unit time). The more complex
the scene, the less frames/second, which in turn can
result in a disturbing saccadic graphics [Burdea &
Coiffet, 1994].
Based on the author’s presentation at RTA/HFM Workshop 007, The Hague, Netherlands, 13-15 April. © Grigore C. Burdea,
except for certain illustrations.
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
Figure 1: SGI graphics vs. PC-based graphics
In 1994 a 486 processor PC with SPEA FIRE board was
capable of 7,000 polygons/sec. A modern Pentium III PC
with Wildcat graphics board can do 6,000,000 polygons/
sec, and costs only 6,000 dollars or so. During the same
time the performance of high-end graphics workstations
produced by SGI rose from 300,000 polygons/sec. on a
Reality Engine in 1994 to 13,000,000 polygons/sec.
today on a multi-pipe Infinite Reality 2 [Real Time
Graphics, 2000]. While its performance is twice that of
the fastest PC rendering board, its price is two to three
hundred thousand dollars, which makes it affordable to
only a few! By significantly improving performance,
while actually reducing costs in the late nineties, the PC
industry made possible the much-desired widespread use
of desktop 3-D graphics.
The second important change in the computer industry is
the tendency to miniaturize the computer, to the point
that it becomes wearable on the user. Figure 2 shows just
such an example, namely the Mobile Assistant IV® produced by Xybernaut Co. (Fairfax VA, USA). It consists
of a CPU unit with a Pentium processor and simplified
keyboard, a head-mounted display, a microphone for
voice input, and a camera worn on the user’s head. By
coupling this with wireless communication, the user gets
freedom of motion within the range of the wireless
transmitter, and as a function of battery life.
User freedom of motion is very important to the VR
application designer, because it increases the naturalness
of the interaction, and thus the feeling of immersion that
the user has. At the present time the Mobile Assistant
does not have sufficient computing power to incorporate
graphics real-time rendering. Such a capability is
expected to appear in subsequent models of the device.
Figure 2: Mobile Assistant IV® wearable computer.
Courtesy of CAIP Center, Rutgers University.
Reprinted by permission
3. Graphic Displays
Another important component of VR systems are the
graphics displays, which present the computer, rendered
scene to the user. Such displays may be classified as
personal displays, for a single user, and large-volume
displays, which allow several users to view the same
scene in a given location. Both types of displays have
advanced significantly in the past decade, as will be
described next.
3.1 Personal displays
The most prevalent type of personal display available in
the nineties were head-mounted displays (HMDs), which
projected the image close to the user’s head. Early
HMDs were very bulky and heavy, weighing over two
kilograms in the case of the VPL “Eyephone.” Their
resolution was poor (360×240 pixels) owing to the LCD
technology of the time. Compared to this, modern
HMDs, such as the SONY Glasstron® shown in Figure 3,
have an SVGA resolution (832×624 pixels). The
improvement in image resolution was coupled with a
dramatic reduction in weight (120 grams for the
Glasstron). Unfortunately, the necessary miniaturization
means that the user’s field of view (FOV) is small
(30×22 degrees) compared to the Eyephone FOV of
90×60 degrees. Recently SONY has announced it will
stop producing Glasstrons. Its logical replacement is the
Olympus Eye-Trek HMD (37×22 degrees) weighing a
little over 100 grams [Olympus, 2000].
senses the 3-D aim of the binoculars and displays the
corresponding scene in real time.
Figure 3: The SONY Glasstron Courtesy of InterSense
Co. Reprinted by permission
The user’s natural field of view is 180 degrees horizontal
and almost as much vertical. The human vision system,
unlike the HMDs, has an uneven resolution over its
FOV. The highest resolution is in a central “foveating
area,” while the retina has much lower resolution away
from the foveating area. By rendering the image at
constant resolution the computer essentially wastes
pixels, since the eye cannot see them. Eye trackers allow
computers to detect where the user focuses on an image.
It is then possible to render the corresponding virtual
scene in high resolution, and the rest of the scene in
lower resolution. A review on the state-of-the-art in eye
tracking can be found in [Isdale, 2000]. Figure 4 shows
an HMD retrofitted with an eye tracker.
Figure 5: The V8 Binoculars HMD. Courtesy of
Virtual Research Systems Inc. Reprinted by permission
Other types of graphics displays, available today, are
“virtual windows” and auto-stereoscopic displays. The
WindowVR® produced by Virtual Research Systems
Inc., is shown in Figure 6. In has a flat-panel display (a
touch-sensitive display in some versions) with handles
and suspension cable. A tracker inside the display allows
the computer to change the scene and give the user the
sensation of looking at a virtual world through a
window. Buttons on the handles allow actions and
navigation within the VR simulation.
Figure 4: The SONY Glasstron fitted with an eye
tracker. Courtesy of VR News. Reprinted by permission
Military reconnaissance training applications can benefit
from a “customized” HMD, such as the V8 Binoculars
(Virtual Research Systems Inc., Santa Clara CA, USA)
shown in Figure 5. These binoculars integrate dual LCD
displays, with VGA resolution, and a FOV of up to 60
degrees. Its optics allows individual focus adjustment,
and its weight is 680 grams. By integrating a position
tracker (discussed later in this report), the computer
Figure 6: The WindowVR®. Courtesy of Virtual
Research Systems Inc. Reprinted by permission
Auto-stereoscopic workstations, such as the ones
produced by Dimension Technologies Inc. (Rochester
NY, USA), use backlighting of a flat panel to produce a
stereo image. As seen in Figure 7, the image appears to
float in space, without the need for HMDs. Its resolution
is 1280×1024, which is superior to that of LCD-based
displays [Dimension Technologies Inc., 2000].
Unfortunately, the stereo image can be seen from only a
small viewing volume and the brightness of the image
suffers owing to the lighting scheme used. Thus graphics
appears dim when compared HMDs or active glasses
(discussed later in this report).
Figure 9 shows a marine amphibious landing exercise
scene produced by a workbench-type display [Hix et al.,
1999]. The usual 2-D military symbols were replaced by
3-D icons of trucks, airplanes, ships, etc., shown on a 3D terrain map. Such a scene is much easier to
comprehend, and may reduce errors in a high stress
combat situation. Furthermore, the use of 3-D icons
coupled with haptics (not used in this particular training
scenario) opens the way for a different kind of C&C
Figure 7: An auto-stereoscopic workstation. Courtesy
of DTI Inc. Reprinted by permission
3.2 Large-volume displays
Large-volume displays offer a much larger stereo
viewing area, high resolution, and a way for many
participants to view and interact with the same virtual
scene. One class of large-volume displays is “virtual
workbenches,” such as the one shown in Figure 8. It uses
a CRT projector and mirrors to “place” the stereo scene
on top of its table. The integration of its projector within
the display table makes for a compact design, and the
tilting mechanism can change the user’s viewing cone.
The Baron can tilt from fully horizontal to fully vertical,
which transforms it essentially in a “virtual wall” type
display. Future designs will replace the CRT technology
with much brighter digital mirror technology. Then it
will be possible to use such displays without having to
reduce the room ambient lighting level.
Figure 8: The BARCO Baron® 3-D display. Courtesy
of BARCO Co. Reprinted by permission
Figure 9: Sea Dragon Marine landing exercise.
Courtesy of the Naval Research Laboratory,
Washington DC. Reprinted by permission
Using a haptic glove (discussed later in this report) the
military commander may then be able to grasp and feel
such 3-D objects. The force feedback addition to the
simulation has at least two important advantages for the
military decision-maker. First, he knows he has complete
and unique control over the unit whose symbol he
grasped. This is true even if he momentarily looks away
from the screen. Second, the hardness of the symbol can
give him valuable information on the unit’s state of
readiness/strength level. A tank 3-D icon that feels soft
may indicate that unit is at half strength, due to losses. A
tanker plane that feels hard may indicate that it is full of
fuel, etc.
An example of a C&C application using a haptic glove is
the system demonstrated by the CAIP Center at Rutgers
University, and shown in Figure 10 [Medl et al. 1998]. It
consists of a distributed architecture, with a multi-modal
interface. The user gives voice commands that are
detected by a microphone array placed on top of a PC.
He can select and move military symbols on a map using
either an eye tracker, or a force feedback glove (Rutgers
Master glove [Burdea, 1996]). The New Jersey National
Guard, with little prior training, tested the system
successfully in 1997.
Figure 10: Multi-modal interface C&C exercise.
Courtesy of the CAIP Center, Rutgers University.
Reprinted by permission
Figure 12: Stereo “active” glasses fitted with the
InterSense tracker. Courtesy of InterSense Co.
Reprinted by permission
A larger type of display than the workbench is the
CAVE® stereo display made by Fakespace Systems
(Ontario, Canada). As shown in Figure 11, the CAVE
consists of multiple wall-type displays assembled in a
cube geometry. Each wall has its own CRT projector,
driven by a separate graphics pipe of a multi-processor
high-end SGI or equivalent computer. The user enters
the CAVE and is looking at the display walls through
“active” stereo glasses, such as those shown in Figure
12. Infrared emitters located in the corners of the CAVE
control the opening and closing of shutters incorporated
in the stereo glasses. They alternately block the view of
each eye, which allows the brain to register the two
images rendered by the computer separately and create
the stereo effect.
With his FOV filled by the graphics the CAVE user feels
immersed in the virtual world. Furthermore, the work
volume in which the user sees stereo and can interact
with virtual “floating” objects is much larger than for a
workbench. These advantages come at a price, as the
cost of the CAVE is five times that of a workbench
display. T this is added the cost of the high-end graphics
computer, bringing the system close to one million
dollars at the time of this writing.
Recently Fakespace Systems introduced the “Reconfigurable Advanced Visualization Environment”
(RAVE) shown in Figure 13. Unlike the CAVE, which
has a fixed geometry, RAVE can change its
configuration depending on the user’s needs. Thus its 3
m × 2.9 m × 3.7 m modules can be assembled to form a
straight wall geometry, where three display units are
side-to-side. Other available configurations include a ushape, or a cube (CAVE-type geometry). Alternately, it
can separate itself into two half-cube independent
displays. As expected, the cost of RAVE surpasses that
of the CAVE.
Figure 13: The RAVE re-configurable stereo display.
Courtesy of Fakespace Systems Inc. Reprinted by
4. Large-Volume Tracking
The user’s ability to see graphics that fill most of his
FOV is a good start towards a more immersive virtual
environment. Another important requirement is to allow
the user to interact with virtual objects he sees. Thus the
computer needs to know as accurately as possible the
current 3-D position of the user’s hand(s), head, or
whole body within this large working volume.
Figure 11: The CAVE stereo display. Courtesy of
Fakespace Systems Inc. Reprinted by permission
4.1 Magnetic tracking errors
Computers determine the user’s position by interpreting
data fed by 3-D trackers worn on the body. The
overwhelming majority of today’s trackers are
electromagnetic ones, consisting of a stationary source of
pulsating magnetic fields, one to several receivers (coils)
worn by the user, and an electronic control box. The
voltages induced in the receivers are transformed in
absolute position/orientation values by the control box,
and then sent to the computer running the simulation.
An example of high-end magnetic tracker is the
MotionStar® wireless tracking suit produced by
Ascension Technology Co. (Burlington VT, USA),
shown in Figure 14. The suit incorporates 20 magnetic
tracker receivers placed at critical locations on the user’s
body, such as the wrist, ankle, hip, etc. The receivers are
wired and the electronic control/communication box
worn on a backpack. Owing to its own power supply (a
battery with two-hour life), the suit can work
independently and furnish up to 100 readings/sec. within
three meters from the tracker source. Such a range would
accommodate two RAVE modules, if placed side-byside, with the source centrally located.
Human-Machine Laboratory at Rutgers University. The
tripod allowed the height of the tracker source to be
varied, while precise position of the receiver was
measured mechanically. The errors grew geometrically
with the distance from the tracker source, as expected.
However, errors also varied depending on the source
height above the floor. The most accurate measurements
were obtained when the source was at 1.68 m above the
floor. Errors grew when the source was too close to
either the ceiling or to the floor, owing to the metallic
beams used in the laboratory room construction.
Additional experimental measurements showed that the
metal in the large-volume display (in this case a BARCO
Baron workbench) introduced more tracking errors.
Source 2.02 m above floor
Source 1.68 m above floor
Source 1.37 m
above floor
Figure 15: The Polhemus LongRanger tracking errors
[Trefftz & Burdea, 2000]
The above findings, and those of others, point out the
inadequacy of magnetic trackers when working in
typical large-volume display environments. Thus one is
left with two alternatives. The first is to build a special
structure, designed from the start to house large-volume
displays and the related trackers, and to redesign the
display to reduce the amount of metal. The second, and
an easier alternative, is to change the tracker.
Figure 14: The MotionStar® wireless tracking suit.
Courtesy of Ascension Technology Co. Reprinted by
There is however a problem with all magnetic trackers,
which affects their accuracy. This is due to interference
from other magnetic fields, or from metallic objects.
Such problems were reported with the MotionStar®
[Marcus, 1997], but also with the Polhemus
LongRanger® (Colchester VT, USA) [Trefftz & Burdea,
2000]. Figure 15 shows the magnitude of the error vector
for a LongRanger® installed on a wooden tripod in the
4.2 Inertial/ultrasonic trackers
In recent years a new generation of trackers has become
commercially available. These are hybrid 3-D position
trackers, such as the IS-600 shown in Figure 16,
manufactured by InterSense Inc. (Burlington MA, USA).
They use a combination of inertial and ultrasonic sensing
technology, with the inertial component used for position
measurements, and the ultrasonic component used to
provide a zero position and to correct for drift. One or
more inertial cubes are placed on the user, or on his
interface, together with sonic disks (as previously shown
in Figure 12 for active glasses). The inertial cube signal
is read by an electronic box, which also drives ultrasonic
receivers placed on the ceiling in a cross configuration.
Since these trackers do not use magnetic fields, they are
immune to the type of interference associated with
magnetic trackers.
Figure 16: The InterSense IS-600® inertial/ultrasonic
tracker. Courtesy of InterSense Co. Reprinted by
A recent addition to the InterSense tracking family is the
IS-900 LAT (large-area-tracker) [InterSense, 2000]. It
can extend its 6 m × 6 m × 3 m standard tracking volume
to a maximum tracking area of 900 m2 using up to 24
expansion hubs. Its measurement accuracy, resolution
and latency are better than for magnetic trackers.
5. Haptic Interfaces
Another important change taking place in current VR
technology is the addition of haptic feedback, namely
tactile and force feedback. Tactile feedback gives the
user the ability to touch and feel the smoothness of
virtual object surfaces, their temperature, slippage, and
contact surface geometry. Force feedback conveys
information on object weight, inertia, mechanical
compliance, degree of mobility, viscosity, etc. The
addition of haptic feedback clearly increases simulation
realism in general. Furthermore, haptic feedback allows
object manipulation in occluded, foggy or dark virtual
environments, a task that would otherwise be difficult or
even impossible to complete.
5.1 General-purpose haptic interfaces
Haptic interfaces may be classified as general-purpose
ones, which can be used for many tasks (including
military ones), and special-purpose haptic interfaces,
which are designed specifically for military applications.
An example of a general-purpose force feedback
interface is the PHANToM® arm Desktop produced by
SensAble Technologies Co. (Woburn MA, USA), and
shown in Figure 17. The interface measures the position
and orientation of the stylus 1000 times/sec, and applies
forces of up to 10 N to the user’s hand in response to
actions in the virtual environment. The high bandwidth
of the PHANToM allows it to combine force with tactile
feedback, such that the roughness or stickiness of a
surface can be simulated as well.
A typical application developed for the PHANToM is
“digital sculpting,” as illustrated in Figure 17. The user
is presented with a block of “digital clay,” which he
deforms, sculpts, polishes, using the stylus. The user
feels the resistance of the material, as well as the
influence of the change in virtual tool to which the stylus
is mapped.
Figure 17: The PHANToM® desktop force feedback
arm. Courtesy of SensAble Co. Reprinted by permission
Once the 3-D model is sculpted, its files can be
downloaded to a NC mill or similar equipment, to build
an actual prototype. This is also applicable to the weapon
design cycle, speeding up its mock-up phase.
Another use of the PHANToM is in mine detection
training, an application being currently developed by the
French Ministry of Defense (see companion paper by
Todeschini). The force feedback arm integrated with this
system is designed to replicate the tactile sensation the
trainee uses to detect a mine. Since in actual operations
such a task must have a 100% rate of success, it is clear
that a realistic trainer should be useful. The difficulty in
realizing such a system is to realistically replicate the
dynamic force “signature” associated with various mines
and ground conditions.
Figure 18: Digital sculpting with force feedback.
Courtesy of SensAble Co. Reprinted by permission
One drawback of the PHANToM arm is that it is not
able to provide finger-specific forces, such as those
present in dexterous tasks, when contact is at the
fingertip. Such tasks could be assembly training,
servicing of military hardware, or training in explosive
handling. For such instances a better haptic interface is a
force feedback glove, such as the CyberGrasp® glove
produced by Virtual Technologies Inc. (Palo Alto CA,
USA), shown in Figure 19.
much more complex, which may lead to system
Figure 20: The CyberGrasp glove in a CyberForce
configuration. Courtesy of Virtual Technologies Co.
Reprinted by permission
Figure 19: The CyberGrasp glove in a CyberPack
configuration. Courtesy of Virtual Technologies Co.
Reprinted by permission
The glove consists of a CyberGlove [Kramer et al.,
1991] used for position measurements on which is
retrofitted a force feedback exoskeleton driven by cables.
The tendons are routed to an electronic control box
housing electrical actuators and communication
hardware. The force output is about 16 N per finger,
which is larger than the PHANToM output. Unlike the
PHANToM, which sits on a desk, and limits freedom of
motion, the CyberGrasp glove is worn. Furthermore, the
CyberPack® configuration places the control box in a
backpack, such that the user can walk around and grasp
objects and feel their hardness. Its limiting factors then
are weight, (which can lead to user fatigue) and the
range of the tracker measuring wrist 3-D position.
Another limitation of the CyberGrasp haptic glove is the
lack of force feedback to the wrist. Thus grasped objects
seem weightless, with no inertia and no mechanical
restraints. Recently Virtual Technology announced the
CyberForce® haptic interface shown in Figure 20. It
consists of a six degrees-of-freedom force feedback arm
connected to the back palm. By combining wrist force
feedback with the force feedback glove, the ability to
simulate weight and inertia are added while the user
preserves his hand dexterity [Kramer, 2000].
Furthermore, there is no need for a wrist position tracker,
since the force feedback arm measures wrist position
faster and without metallic interference. Unfortunately,
the dimensions of the arm limit the user’s freedom of
motion. Furthermore, the overall system control becomes
In certain military applications of VR, such as infantry
training, there is a need to simulate running, or walking
uphill, or through uneven terrain. In such cases haptic
feedback to the body becomes important in order to have
realistic training. One system that addresses these needs
has been recently developed by Sarcos Co (Salt Lake
UT, USA) and the University of Utah [Hollerbach et al.,
1999]. As shown in Figure 21, the user is located in front
of a three-wall display filling most of his FOV and
stands on a treadmill. By tracking his walking/running
on the treadmill, the computer updates the virtual scene
accordingly. A force feedback arm is attached to the
user’s torso through a harness. The arm applies resistive
and inertial forces to simulate uneven terrain and other
effects. A rope attached to the ceiling prevents injury in
case of tripping and falling.
Figure 21: The treadport VR system. Courtesy of
University of Utah CS Dept. Reprinted by permission
Recently, Japanese researchers proposed the replacement
of the treadmill approach with an “active floor”, as
shown in Figure 22 [Noma et al., 2000]. The floor is
composed of modular actuator tiles that can change slope
under computer control. The user’s motion is tracked by
a vision system, and the tiles actuated as needed to
replicate uneven terrain. Thus, unlike the walking-inplace paradigm of treadmill systems, the active floor
approach allows natural walking over the whole surface
of the floor. There is no need for a force feedback arm
attached to the user’s back, and no need for a safety
rope. The limitation in this case is the size and amount of
slope that can be produced by the active tiles.
Figure 22: The active floor VR system [Noma et al.
2000]. © IEEE. Reprinted by permission
5.2 Special-purpose haptic interfaces
All the haptic interfaces presented so far are generalpurpose, since they can be used in military applications
but were not specifically designed for such. By contrast,
special-purpose haptic interfaces are designed from the
start to provide force/touch feedback to military VR
tasks. An example is the Stinger trainer prototype
developed at TNO (The Hague, The Netherlands) [Jense,
1993], shown in Figure 23. It consists of a plastic mockup of the missile launcher, which is instrumented to track
the user’s aim, and to sense when switches are
depressed. Furthermore, a virtual environment showing
the enemy aircraft is presented to the trainee on an
HMD. The advantage of this system is that a much more
compact set-up replaces the classical large-dome training
system. Furthermore, all user actions are stored
transparently and his performance data is available on
the computer. The force feedback sensation is produced
naturally by the plastic mock-up, without need for more
expensive (and heavier) hardware. The system is now
being used in training the German Air Force, as
described in the companion paper by Reichert.
Another example of special-purpose haptics is the antitank missile trainer system recently developed by the
Fifth Dimension Technologies Co. (Pretoria, South
Africa), which is shown in Figure 24. It uses a mock-up
of the rocket launcher, similar to the TNO Stinger
trainer, which provides direct tactile feedback. Other
similarities include the used of a HMD to display the
virtual battlefield to the trainee, and a 3-D tracker to
determine his direction of view.
Figure 23: The Stinger VR training prototype Courtesy
of TNO, The Netherlands. Reprinted by permission
Figure 24: The anti-tank VR training prototype
Courtesy of 5DT Co., Pretoria, South Africa. Reprinted
by permission
Another type of special-purpose haptic interface is the
parachute-training simulator developed by Systems
Technology Inc. (Hawthorne CA, USA). As shown in
Figure 25, the system uses a full-size parachute harness,
and an HMD showing a detailed 3-D jump scene (insert).
The scene moves in response to either head motion, or
the toggle of the parachute harness [Systems Technology
Inc. 2000]. Wind effects are added, to train the jumper in
coping with adverse landing conditions. Playback of user
actions and instructor actions are used to help acquire the
necessary skills.
with an earlier release may not run when the library is
updated (currently WTK is at release 9).
Figure 26: The tank interior created with WTK.
Courtesy of EAI Co. Reprinted by permission
Figure 25: The VR parachute training system. Courtesy
of Systems Technology Inc. Reprinted by permission
6. Modeling Tools
So far this report has reviewed the computing hardware
and the interfaces available to develop VR applications.
The third element needed is a VR toolkit, i.e. software
libraries specifically developed for programming virtual
environments. Such toolkits offer certain advantages to
the developer, namely drivers for most VR I/O devices,
certain 3-D graphics routines, ease of portability, etc. In
turn VR toolkits can be classified as general-purpose and
special-purpose libraries.
6.1 General-purpose Modeling Tools
The most used VR programming toolkit today, by far, is
“WorldToolKit” (WTK), produced by Sense8, a division
of Engineering Animation Inc. (Ames IA, USA). It
consists of over 1000 C/C++ object-oriented functions,
which are executed, in an infinite loop during the
simulation. An example of a scene created with WTK is
the tank interior simulation shown in Figure 26. By
importing CAD files, doing smooth shaded graphics,
textured surfaces, dynamic effects, WTK allows very
realistic simulations to be created.
Another facility provided by WTK (in its “World-up”
version) is graphics programming, as shown in Figure
27. Thus the kinematics dependencies and other virtual
object characteristics can be easily specified using a
scene graph. At run time the software goes through the
nodes of this scene graph.
For all its advantages WTK has at least two
disadvantages, namely cost and short-lived releases. The
license cost for WTK is an order of magnitude more than
for widespread PC software, reflecting the small market
for VR products. This is aggravated by numerous
releases, which many times are not compatible with
earlier ones. As such a military application developed
Figure 27: The World-up graph scene. Courtesy of EAI
Co. Reprinted by permission
A 3-D programming toolkit which is free is Java3D
produced by Sun Microsystems (Palo Alto CA, USA).
Java3D programming is also based on a scene graph.
However, the software is still under development, and
certain drawbacks exist, when compared with WTK.
One of the most important limitations of Java3D is its
inability to deliver a uniform rendering speed, as
uncovered by recent tests done at Rutgers University.
Figure 26 [Boian, 2000] shows the same scene being
rendered on a dual-processor 450 MHz Pentium PC,
using (a) WTK (release 8) and (b) Java 3D (release
1.1.2). The scene consisted of 40,000 textured polygons,
and collision detection was activated. When WTK was
used, the average time to render one frame was 123 ms
(8.1 frames/sec), with a standard deviation of about 10
ms. Interestingly enough, Java3D was 37% faster, with
an average rendering speed of 11.1 frames/sec. Its
average time to render a frame was only 90 ms.
Unfortunately, its standard deviation was 84 ms, or
840% larger that for WTK.
of 1000 Hz. The visual frame rate was 20 frames/sec,
using Boeing’s proprietary “FlyThru” rendering
Figure 29:
Figure 28: Comparison of frame rendering speed and
consistency between: a) WTK; b) Java3D [Boian,
2000]. Reprinted by permission
Generalizations can be risky, and certainly SUN
Microsystems will address some of these drawbacks in
newer Java3D releases. However such large standard
deviations in frame rendering time, as present in the
current Java3D release will adversely impact interactions
in the virtual environment, especially where force
feedback is concerned.
Force feedback calculation is preceded by a collision
detection step that is used by the computer to determine
if there is interaction in the virtual environment. Such an
algorithm needs to be both accurate and fast, which is
difficult in complex virtual environments. One example
is CAD analysis for accessibility. Complex assemblies,
such as “crowded” aircraft engines, are difficult to
design and even more difficult to service. Researchers at
Boeing Co. (Seattle WA, USA) have developed the
“voxel point shell” (VPS) method of collision detection
to cope for such application needs [McNeely et al.,
1999]. VPS builds a point shell around the surface of a
single moving object in a pre-computing stage. At run
time, this point shell is checked for collision with the
static environment, and the resulting force/torque applied
to the user. Tests done using a complex model of a
Boeing 777 with almost 600 thousand polygons, shown
in Figure 29, allowed haptic rendering at a constant rate
6.2 Special-purpose modeling toolkits
Special-purpose toolkits have been developed to help
certain types of simulations. For example, Virtual
Technologies have introduced the VirtualHand® Suite
2000, which is a library designed to work with the
CyberGlove, CyberGrasp, and CyberTouch interfaces
[Virtual Technologies, 2000]. It helps develop
applications where interaction with the objects is at the
level of the hand, and includes collision detection, a
force feedback API and networking capabilities.
Another special-purpose toolkit is the GHOST library
developed by SensAble Technologies for their
PHANToM arm. It allows the mixing of scene graph and
direct force field programming, in scenes with
complexities up to 250,000 polygons (mesh
configuration). Multiple PHANToM Desktop models
can be supported in a daisy-chain arrangement on a
single host communication port.
Finally, the DI-Guy library developed by Boston
Dynamics (Cambridge MA, USA) helps program
simulations involving dismounted infantry, special
operations and peacekeeping operation tasks by
providing an intelligent-agent based library [Boston
Dynamics Inc., 1997]. As can be seen in Figure 30, the
toolkit allows users to control avatars that respond to
real-time task-level control. Once they are given
behavior (walk, kneel, crawl, etc.) and travel parameters,
they execute the action through motion interpolation.
This allows multiple DI-Guy characters to be included in
a given virtual scene. The toolkit is currently supported
by WTK (Release 9) and by Vega (Paradigm
Simulations Inc., Dallas TX, USA). Vega LynX allows a
point-and-click interaction environment.
Figure 30: Scene created with the DI-Guy toolkit for
dismounted infantry training. Courtesy of Boston
Dynamics Inc. Reprinted by permission
7. Conclusions
There is no doubt that VR technology has been going
through a rapid change. A major impact on the
widespread use of this technology in the military and
other areas is the tremendous decrease in computer
prices, and increase in PC-based graphics speed. The
miniaturization of the PC in its present form allows for
portability, which results in increased user freedom of
motion and simulation realism. Large-volume displays
are also adding to the user ability to interact with large
simulation volumes. New trackers have overcome the
limitation of magnetic technology and can be used for
wide area tracking and interaction. Portable haptic
interfaces also add to realism, especially in tasks
involving manual dexterity. Programming toolkits now
offer a complex programming environment integrating
the various modalities of interacting with the virtual
world. All these developments point to more useful
military application of VR, primarily in training, but also
in C&C and weapon design/prototyping. Human factor
studies need to validate the technology and its
NATO travel support for delivery of the Key note
presentation at the Workshop in The Hague is gratefully
acknowledged. Author’s research reported here was
supported by grants from the National Science
Foundation, from Office of Naval Research (DURIP)
and from Rutgers University (SROA and CAIP grants).
Boian, R., “A Comparison Between WorldToolKit and
Java3D,” Rutgers University, ECE Dept., Project
Report, May 2000.
Boston Dynamics Inc., “Vega DI-Guy™,” Data sheet,
1997. Also at www.bdi.com.
Burdea, G. & Coiffet, P., Virtual Reality Technology,
John Wiley & Sons, New York, 1994.
Burdea, G., Force and Touch Feedback for Virtual
Reality, John Wiley & Sons, New York, 1996.
Dimension Technologies, Inc. “3-D Flat Panel Virtual
Window Display Family,” Company brochure, 4 pp.,
Rochester, NY, 1998.
Hix, D., Swan, E., Gabbard, J., McGee, M., Durbin, J. &
King, T., “User-Centered Design and Evaluation of a
Environment,” Proceedings of IEEE Virtual
Reality’99, pp. 96-103, March 1999.
Hollerbach, J., Thompson, W. & Shirley, P., “The
Convergence of Robotics, Vision, and Computer
Graphics for User Interaction,” The International
Journal of Robotics Research, vol. 18, no. 11, pp.
1088–1100, November 1999.
InterSense Co., “InterSense IS-900 Precision Motion
Tracker,” Company brochure, Burlington, MA, 2000.
Also at www.isense.com.
Isdale, J., “Alternative I/O Technologies,” VR News,
Vol. 9, No. 2, pp. 24-29, March 2000.
Jense, H., Personal communication, TNO Physics and
Electronics Laboratory, The Hague, The Netherlands,
August 1993.
Kramer, J., Lindener, P. & George, W., “Communication
System for the Deaf, Deaf-Blind, or Non-Vocal
Individuals Using Instrumented Glove,” US Patent
5,047,952, September 10, 1991.
Kramer, J., “The Haptic Interfaces of the Next Decade,”
Panel Session, IEEE Virtual Reality 2000
Conference, March 2000.
Marcus, M., “Practical Aspects of Motion Capture
Technology for the Entertainment Industries,”
www.itc.co.uk/mirage, 1997.
McNeely, W., Puterbaugh, K. & Troy, J. “Six Degreeof-Freedom Haptic Rendering Using Voxel
Sampling,” Computer Graphics Proceedings
(SIGGRAPH), pp. 401-408, August 1999.
Medl, A., Marsic, I., Andre, M., Liang, Y., Shaikh, A.,
Burdea, G., Wilder, J., Kulikowski, C. & Flanagan,
J., “Multimodal Man-Machine Interface for Mission
Planning”, Intelligent Environments — AAAI Spring
Symposium, March 23-25, Stanford University,
Stanford CA, pp. 41-47, 1998.
Noma, H., Sughihara, T. & Miyasato, T., “Development
of Ground Surface Simulator for Tel-E-Merge
System”, Proceedings of IEEE Virtual Reality 2000,
IEEE, 2000, pp. 217-224.
Olympus Corporation of America, “Eye-Trek
Specifications,” http://www.eyetrek.com, 2000.
Systems Technology Inc., “Parachute Flight Training
http://www.systemstech.com/paramain.htm, 2000.
Real Time Graphics, “High-Performance Image
Generators — A Survey,” vol. 8, no. 6, January 2000.
Trefftz, H. & Burdea, G., “Calibration Errors in LargeVolume Virtual Environments,” CAIP TR-243,
Rutgers University, 2000.
Virtual Technologies Inc. (2000). “VirtualHand® Suite
2000”, Palo Alto CA, www.virtex.com/products.
Simulating Haptic Information with Haptic Illusions
in Virtual Environments
Anatole Lécuyer1
Aerospatiale Matra CCR
12 rue Pasteur, 92152
Suresnes, France
Sabine Coquillart
INRIA Rocquencourt
Domaine de Voluceau, 78153
Le Chesnay Cedex, France
This paper presents a set of experiments in which a
human user feels haptic sensations. These sensations are
in fact haptic illusions, generated by a visual effect.
Then, these haptic illusions are described and analysed.
These haptic illusions were generated by the use of a
pseudo-haptic feedback system. It is a system combining
an isometric input device and visual feedback. The
experimental apparatus did not use any force feedback
The paper addresses the role of action in the perception
loop — subjects felt a reactive force corresponding to
their own sensory-motor command. In addition, subjects
had to “participate” in the illusion process by choosing
the cognitive strategy, which led to the illusion.
In the future, the use of the concept of illusion might
improve or simplify VR simulations and pave the way to
a better understanding of human perception.
Philippe Coiffet
10-12 avenue de l’Europe, 78140
Vélizy Villacoublay, France
The simulation of force feedback by pseudo-haptic
feedback can be considered as a phenomenon of haptic
illusion. An illusion is a non-veridical perception. It is a
mistake made by our brain and not by our senses. The
effect of illusion can be generated by means of art,
artefact or special effects. This effect can be perceived
but is not real.
The objective of the study is to analyse some haptic
illusions involved in the pseudo-haptic feedback, in
order to introduce the concept of illusion in the design of
virtual environments.
After addressing previous work on haptic illusion, this
paper describes two different experiments, which were
carried out to demonstrate the potential of pseudo-haptic
feedback. Then the paper studies the haptic illusions,
which are generated by these experiments. Finally, it
assesses the perceptual mechanisms involved in the
process as well as their potential.
1. Introduction
The challenge of VR technology applied to aeronautical
virtual prototyping is the backdrop to the study.
Nowadays, virtual prototype designers should take into
consideration the assembly and the support constraints as
early as possible in the development process. Indeed, the
operator (or the designer) should have the possibility to
feel and interact more physically with the mock up. It is
therefore essential to allow haptic feedback in virtual
assembly and support operation simulations.
Haptic feedback devices will soon provide new and
indispensable possibilities [4]. But today these interfaces
remain expensive and complex. Thus, there is a need for
other replacement solutions.
In the absence of a haptic interface, a previous paper
[10] studied the possibility to simulate force cues with an
input device within a virtual environment (VE). This
device is the 2003C model of the Logitech Spaceball
[3], which is an isometric device — “isometric” meaning
that the Spaceball is nearly static and remains in place
while a pressure is being exerted upon it. The force
feedback was simulated by using the mechanical
characteristics of the passive device: its internal stiffness
and its thrust — and by combining them with an
appropriate visual feedback. The result of this visiohaptic feedback was called “pseudo-haptic” feedback
[10]. The pseudo-haptic feedback was established
qualitatively and quantitatively following different
psychophysical experiments.
2. Previous Work
Some well-known optical illusions such as the MüllerLyer illusion (see Figure 1a) or the Zollner illusion are
extensively described in scientific works [8]. Many
examples of famous illusions can be found on the web
[1]. And there are even companies whose business is
devoted to developing educational and fun products
relating to visual illusions [2].
Figure 1: Müller-Lyer Illusion : the left
segment looks smaller than the right one [a];
Bourdon Illusion : the left border looks
slightly bent [b]
But illusions may occur on the other sensorial modes. An
auditory illusion [2], composed by Roger Shepard in
For contact with authors: [email protected], [email protected], and [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
1964, is a transposition of the famous endless stairs
drawn by the Dutch graphist M.C. Escher on the
auditory mode. Shepard played on a keyboard an
ascending or descending chromatic or diatonic scale
using four parallel octaves simultaneously. The tones
were perceived as continuously increasing or decreasing
in pitch, however after travelling over an octave, they
were the same in pitch as when first started.
The existence of haptic illusions can be revealed by
simple experiments. For example, considering three jars
of water; from left to right, the temperature varies from
warm, tepid to cold. When the hands are first dipped into
the outer jars the water is perceived as warm on the left
hand-side and as cold on the right hand-side. Then, when
dipping both hands into the middle container, one
perceives again two different temperatures, this time
however, in reversed order cold on the left and warm on
the right, though the water is neither warm nor cold but
The Thaler haptic illusion can also be simulated very
simply. One can observe that the temperature of an
object influences the haptic perception of its weight: a
cold coin seems heavier than a coin of the same size but
warmer [12]. Thanks, probably, to the fact that
perception of coldness and that of heaviness share
common neurones.
Another example described in [5], is the haptic
equivalent of the Bourdon’s visual illusion (see Figure
1b). Day used a 3D volumetric model of the Bourdon
Figure. When a person explores the two opposite
surfaces of the model with his/her thumb and forefinger,
he/she feels the upper and straight surface as being
slightly bent. On average, people felt a bend of 3.8
degrees visually and a bend of 3.5 degrees haptically.
showed that there were strong effects of the Müller-Lyer
illusion on grasping as well as on visual perception,
indicating that the motor system is also receptive to
visual illusions.
In a VR simulation, Hogan studied haptic illusions
occurring during the exploration of virtual objects and
their implications on the perceptual representation of
these objects [9]. He used a force feedback arm
constrained to move on the 2D horizontal plane. Subjects
grasped the handle of the arm and were asked to evaluate
the length and stiffness of virtual rectangle objects. The
handle was grasped and moved around the virtual
rectangles. During the task, the subject had to choose the
longer (or the stiffer) stimulus of two stimuli. One
stimulus was evaluated on the X-axis of the horizontal
plane, while the other was evaluated on the Y-axis (i.e.
each stimulus depended on a side of the rectangle).
Results show that for the same stimulus on the X- and Yaxis, a difference of perception occurs according to the
distance of the handle from the shoulder (as if the
vertical-horizontal visual illusion was projected in the
haptic mode, on the horizontal plane). Hogan stated that
these haptic illusions show that the internal model of
haptic perception is not metrically consistent. This
property should significantly modify and simplify the
performance constraints in forces computation.
It seems that very few VR papers explored the
possibility to use illusions directly in the conception of a
This paper presents and analyses haptic illusions, which
were showed by two VR experiments. The next part
describes the two experiments and their results.
3. Pseudo-Haptic Experiments
Researchers do not always agree on what causes
illusions, and many illusions remain “unsolved”. Ellis
and Lederman focused more precisely on the origin of
illusion as located on the visual mode or the haptic one.
They studied the famous size-weight illusion [7] and the
material-weight illusion [6] — the size-weight illusion
occurs when a large radius ball seems heavier than a ball
of the same weight but with a smaller radius. The
material-weight illusion is the influence of the texture of
an object on the perception of its weight. Ellis and
Lederman established these two illusions as a primarily
haptic phenomenon, despite the size-weight illusion was
traditionally considered as a case of vision influencing
haptic processing.
Some works deal with consequences of illusions on
perception or on the performances of our motor system.
Volker studied the influence of visual illusions on
grasping [15]. Different subjects were presented fins on
a monitor screen being directed either outwards or
inwards such as in the Müller-Lyer illusion (Figure 1a).
During the grasping task, subjects were told to grasp the
fin, and the maximal aperture between thumb and index
finger was measured. During the perception task,
subjects were told to adjust the length of a comparison
bar on the screen to match the length of the fin. Volker
The concept of pseudo-haptic feedback relies on
coupling the visual feedback with the internal resistance
of the isometric device, which naturally reacts to the
force applied by the user. The overall system returns a
force information called pseudo-haptic feedback.
For example, let us assume that an operator manipulates
a virtual pipe in a virtual environment within the frame
of an insertion task evaluation. The pipe is displayed on
the monitor, and moved by means of the Spaceball. It
is to be inserted into a virtual duct. As the pipe
penetrates the duct, its speed is slowed down. The user
instinctively increases his pressure on the ball, which
results in the feeding back of an increasing reaction force
by the static device. This combination of visual effect
and growing reactive force is then expected to generate
cues of friction.
In order to study the pseudo-haptic feedback concept,
different experiments were conducted. Two of them are
described in the following paragraphs.
3.1 The “Swamp” Experiment
The swamp is a quantitative evaluation of the pseudohaptic feedback. 18 people took part in this experiment.
Each subject was told to manipulate a virtual cube in a
3D virtual environment (see Figure 2). The cube was
manipulated in 2D on the horizontal plane with either a
classical 2D mouse or with the Spaceball.
As the cube moves over a grey area, its speed is
accelerated or slowed down. At this very moment, the
subjects were asked to describe and compare their
sensations when using the 2D mouse or the Spaceball.
Figure 3: Psychophysical experiment — Manual
discrimination between a virtual spring and a real one
Figure 2: The Swamp experiment display
A quantitative comparison between an isometric device
and an isotonic device must be taken cautiously since
these interfaces are not used in the same way. But the
swamp example did display some global tendencies.
A great majority of people logically found that the use of
the two interfaces were very different. The need for a
learning phase with the Spaceball generally disturbed
subjects when starting their manipulation.
The subjects systematically perceived the following
phenomena: friction, gravity or viscosity, when the cube
was slowed down with both devices. Conversely, they
perceived a sense of gliding or lightness when the cube
was accelerated.
Great majorities of people found that the sensations they
felt were different while using the Spaceball or the 2D
mouse. Nearly all of the subjects chose the Spaceball
as the interface with which the “forces” were more
perceptible. This sensation was less obvious when the
cube was accelerated — which is probably due to the
fact that the reactive force from the static device is more
efficient during a compression phase.
The quantitative indications provided by the swamp
experiment were very useful to show us the potential of
this concept, but they didn't measure the characteristics
of the generated feedback. It was necessary to evaluate
more qualitatively the pseudo-haptic information: to do
so a psychophysical experiment has been conducted
3.2 Discrimination between a Virtual Spring and a
Real One
The psychophysical task, which was chosen, is manual
compliance discrimination between a virtual spring and a
real one (see Figure 3). The real spring is embedded
inside a piston, like a “trumpet piston” (see Figure 3).
The virtual spring is a combination of the input device
and the visual feedback (see Figure 4). A hand-made
apparatus was fixed on the Spaceball to obtain the
same catching in the virtual environment and in the real
one. The virtual spring is visually displayed on the
computer screen. It is made to appear as similar as
possible to the real piston. The force applied on the ball
by the user controls the visual displacement of the virtual
spring. When pressing the virtual spring, the user’s
thumb barely moves, since the Spaceball is an
isometric — hence static — device.
Figure 4: Virtual spring set-up
27 people took part in this experiment. There were 972
trials per subject. During each trial, each subject was
asked to test a real spring and a virtual one and to select
the one, which seems to him to be the stiffer. There were
three possible real springs with three different degrees of
stiffness. And each real spring was compared with 12
different virtual springs.
Theoretically, a “stiffer” virtual spring corresponds to a
case when the force — which is required to move the
visual display of the piston on the screen along a certain
distance — is bigger than the one which is required to
move the real spring along the same distance.
(For a complete description of this experiment see [10].)
The large volume of collected data made it possible to
calculate a psychophysical parameter called the Just
Noticeable Difference (JND). The resulting average JND
for the manual compliance discrimination between a
virtual spring and a real one is equal to 13.4%. It is
consistent with previous studies on compliance
discrimination between two springs simulated within a
single environment [13].
This consistency shows quantitatively that a system,
which combines visual feedback and an isometric
device, can provide force cues, which are comparable
with real ones.
4. Illusions Observed
The whole concept of pseudo-haptic feedback relies on a
phenomenon of haptic illusion. In the course of the two
experiments, the haptic perception is mistaken by a
visual effect. The visual feedback generates a new haptic
interpretation of a virtual scene, thus a haptic illusion.
This assumption is confirmed by the simple fact that if
one closes one’s eyes during one of these two
experiments, the experimental task becomes impossible,
and the haptic sensations vanish.
During the first experiment, the perception of friction
when crossing the grey area in the virtual environment is
linked to the visual variation of the speed of the cube.
The whole set-up generates a haptic illusion of several
haptic attributes of the cube — heaviness, lightness — or
of the grey area — rugosity, viscosity, and friction.
In the course of the second experiment, without the
visual displacement the haptic perception of the virtual
spring remains the same, i.e. the Spaceball internal
stiffness. The pseudo-haptic set-up generates the haptic
illusion that different springs are being manipulated. It
becomes possible to perceive different stiffness with the
same Spaceball.
One more illusion phenomenon is revealed in the course
of the second experiment by a question asked to the last
ten subjects. These people were told to draw a straight
line corresponding to the maximum displacement of the
thumb when pressing a virtual spring. The result
indicates an average overestimation of 5 times their
actual displacement (see Figure 5). It means that they
completely assimilated the visual displacement on the
computer screen to their own thumb motion. In other
terms, it implies an illusion of their proprioceptive sense.
Segment 1
Segment 2
Figure 5: Illusion of the Proprioceptive Sense.
Segment 1 — real maximum displacement of the
user’s thumb; Segment 2 — estimated displacement
of the user’s thumb
5. Discussion
The pseudo-haptic feedback is not an illusion of force
feedback. There is actually a force feedback during both
experiments when actuating the Spaceball:
First, because in all manipulation tasks there is a force
feedback reaching the brain — in terms of pain or
fatigue for example. Broadly speaking, even when one
simply holds something in one’s hand or wants to grasp
an object, the motion command sent from one’s brain
activates the muscles of the arm, and makes the one feel
efforts or tensions in his/her muscles or his/her tendons.
These efforts being sent back to the brain via the afferent
neurone network.
The manipulation of the virtual cube with the 2D mouse
in the first experiment could then be considered as a case
of pseudo-haptic feedback with an isotonic device. The
speed of the cube was decreased when passing over the
grey area, then the user had to increase his arm motion,
spend more energy on this gesture, and this may lead to
the friction effect.
In addition to the afferent signals coming from the
different mechanoreceptors, some efferent mechanisms
play a role in human kinaesthesia. Such as the
“innervating sensation” [14], which occurs when one
overestimates the weight of an object when tired. It is a
distortion of force perception, which is due to our own
command system. Our will to achieve an action
generally makes us feel the anticipated result of this
action before it actually happens. And this introduces the
role of action in the perception loop of illusion.
Then, a force feedback from the static device is also
present. It is not an “active” force feedback — i.e. a
computed force feedback — different from other force
feedback systems such as the PHANToM [11] of
SensAble Technologies. The current force feedback is
provided by the reactive force coming from the
Spaceball. And since the Spaceball is nearly static, it
means that the reactive force is nearly equal to the force
applied by the user on the ball. It is a characteristic of the
pseudo-haptic feedback with an isometric device: the
force feedback is always equal to the force command.
This is illustrated on Figure 7.
Figure 6 and Figure 7 show the difference of information
flux in a pseudo-haptic system and a haptic one. In the
case of a classical haptic feedback system such as the
PHANToM (see Figure 6), the user transmits a motion
which is sensed by the optical encoders of the
PHANToM. The force feedback device sends back to
him the computed virtual force, which is a function of
the interference between the probe and the virtual
elements of the virtual environment. The visual feedback
doesn’t play a major role in the haptic perception
Visual Perception
Haptic Perception
Motor System
Figure 6: Haptic feedback system
In the pseudo-haptic case (see Figure 6), the user
transmits his force command to the simulation by means
of the force sensors of the Spaceball. At each
simulation step, the force fed back is constantly equal to
the opposite of the force applied. The user receives
exactly the same force as the one he has just applied. The
haptic profile of the force vs. displacement is given by
the Spaceball internal stiffness. This profile
corresponds to the one of a constraint gauge; this profile
is not linear. The final haptic perception is achieved by
combining the force information and the visual impact of
this information in the VE. It means that the Spaceball
internal stiffness is somehow “mapped” on a visual
event. In reverse, the visual effect gives sense to the
force information.
force command
Visual Perception
Haptic Perception
Motor System
force applied
on the ball when the cube motion is slowed down. This
probably happens if he/she decides to keep the virtual
cube at fast motion, which is a cognitive strategy relying
on many factors affecting the subject (passivity,
availability, stress, etc.). This would imply that pseudohaptic feedback and haptic illusion could also depend on
cultural or contextual reactions of the subject.
In the course of the compliance discrimination task, the
subject had to recompose the stiffness of a virtual spring
with information coming from different modalities. In
addition, there was a conflict concerning the spring
displacement between the proprioceptive information
and the visual one.
Since they were able to compare the final model of the
virtual spring with a real one, the result of the
experiment shows that subjects succeeded in
recombining sensory information. It implies that they
made the choice to use the visual displacement rather
than the proprioceptive displacement. This choice is the
result of an unconscious participation of the user in
pseudo-haptic simulation, and is the reason why the
illusion appeared.
For the time being it is difficult to know if this choice is:
• an example of a sensory substitution or sensory
dominance, which corresponds to the following
expression: Vision dominates Touch ⇒ I use the
visual displacement to evaluate virtual springs,
• or, an example of a choice between different
cognitive strategies. I must choose among all
possibilities one that can help me to perform my
discrimination task, and eliminate other strategies.
This rather corresponds to the second expression: I
must evaluate a virtual spring ⇒ I choose the visual
displacement (and not the proprioceptive one) which
makes it possible.
In other terms, is the proprioceptive illusion due to a
characteristic (or a limit) of human perception system (=
“peripheral” view); or is it due to a decision process
made in a strategic situation (= “central” view).
This alternative has a direct impact on the conception of
VE’s which are to be based on pseudo-haptic feedback
or sensory illusions. There is indeed, a need for further
investigation concerning the generation of illusions.
Figure 7: Pseudo-haptic feedback system
An obvious consequence of this characteristic being that
the pseudo-haptic experiments, which were described,
cannot work without visual feedback.
The pseudo-haptic feedback process system implies that
the user receives his/her own force command in return.
Indeed, the whole pseudo-haptic feedback depends on an
action as well as a participation of the user during
simulation: in the course of the swamp experiment, the
friction sensation occurs if the user increases his pressure
6. Conclusion
The paper has presented VR simulations in which force
cues or haptic behaviours are simulated with a pseudohaptic feedback. This pseudo-haptic feedback comes
along with phenomena of haptic illusions. It is not an
illusion of force feedback, but rather an illusion of using
a force-feedback device.
The analysis of a pseudo-haptic feedback system shows
the role of the sensory motor command in the perception
loop, and also points to the unconscious participation of
the user in the illusion, which is linked, to his/her
cognitive strategy during the experimental task.
Designers of virtual environments, who usually try to
recreate human stimuli in an anthropomorphic manner,
could envisage a wider use of this concept of illusion.
The method, should it exist, implies to revise the
simulation process and the use of human-computer
interfaces. The designer has to think in terms of sensory
information feedback. He/she has to decompose the
sensory information into its different sensory modalities,
and to reshape it into a new sensory distribution. To do
so, he/she can make full use of all the possibilities that
are known in the field of sensory illusions and sensory
It is necessary to facilitate the repositioning of the user
perception to an “implicit” solution. It means that this
implicit sensory alternative must be explicit enough to be
found quickly by the user.
For example, in the case of the second experiment, the
information needed was the displacement of the virtual
spring, and its implicit alternative was the visual
Future work must develop and evaluate more cases in
which sensory illusions are used for VE interactions. The
overall objective is to propose an empirical method to
incorporate illusions in the conception of VE’s.
The authors would like to thank Mr P.R. Persiaux, Mr.
D. Tonnesen and Mrs. M.J. PaškauskaitG for their
valuable remarks.
G. Burdea. Force and Touch Feedback for Virtual
Reality. John Wiley and Son, US, 1996
R.H. Day. The Bourdon Illusion in haptic space.
Perception and Psychophysics, 47, 400-404, 1990
R.R. Ellis and S. J. Lederman. Modality, Weight
and Grip Force Effects in the Material-Weight
Illusion. In Proc. of the Canadian Society for
Brain; Behavior and Cognitive Science Annual
Meeting , 1995
R.R. Ellis and S. J. Lederman. The Role of Haptic
versus Visual Volume Cues in the Size-Weight
Illusion. Perception and Psychophysics, 53(3):
315-324, 1993
E.B. Goldstein. Sensation and Perception.
Brooks/Cole, US, 1999
N. Hogan, B.A. Kay, E.D. Fasse, and F.A. MussaIvaldi. Haptic Illusions : Experiments on Human
Manipulation and Perception of “Virtual Objects”.
Cold Spring Harbor Symposia on Quantitative
Biology, 55:925:931, 1990
A. Lécuyer, S. Coquillart, A. Kheddar, P. Richard,
and P. Coiffet. Pseudo-Haptic Feedback : Can
Isometric Input Devices Simulate Force
Feedback? In Proc. of IEEE International
Conference on Virtual Reality, 2000
T.H. Massie, J.K. Salisbury. The PHANTOM
Haptic Interface : A Device for Probing Virtual
Objects. In Proc. of ASME Winter Annual
Meeting, Symposium on Haptic Interfaces for
Virtual Environments and Teleoperator Systems,
Sherrick and Cholewiak. A Finite Element
Formulation for Nonlinear Incompressible Elastic
and Inelastic Analysis. Computers and Structures,
H.Z. Tan, N.I. Durlach, G.L. Beauregard, and
M.A. Srinivasan. Manual Discrimination of
Compliance Using Active Pinch Grasp: the Roles
of Force and Work Cues. Perception and
Psychophysics. 57(4):495-510, 1995
C. Tzafestas. Synthèse de retour kinesthésique et
perception haptique lors de tâches de
manipulation. Ph.D. Thesis, Université de Paris 6,
Jul. 1998
F. Volker, M. Fahle, K.R. Gegenfurtner, and H.H.
Bülthoff. Grasping visual illusions: No difference
between perception and action? In Proc. of ARVO
Meeting, 1999
Tactile Displays in Virtual Environments
Jan B.F. van Erp1
TNO Human Factors
Kampweg 5
3769 DE Soesterberg
The Netherlands
Virtual Reality (VR) technology allows the user to
perceive and experience sensory contact with a nonphysical world. A complete Virtual Environment (VE)
will provide this contact in all sensory modalities.
However, even state-off-the-art VEs are often restricted
to the visual modality only. The use of the tactile
modality might not only result in an increased
immersion, but may also enhance performance. An
example that will be discussed in this paper is the use of
the tactile channel to support the processing of degraded
visual information. The lack of a wide visual field of
view in VEs excludes the use of peripheral vision and
may therefore degrade navigation, orientation, motion
perception, and object detection. However, tactile
actuators applied to the torso have a 360° horizontal
‘field of touch’, and may be suited to present navigation
1. Introduction
Developments in VR technology have mainly focussed
on the visual sense. In the last decade, enormous
improvements have been made regarding the speed and
resolution of the image generators. However, the human
senses are not restricted to the visual modality. Using the
auditive and tactile modality as well in a VE might have
several advantages. This paper will more specifically
discuss the tactile sense in relation to VE use. I will
restrict the tactile channel to ‘the skin as information
channel’. Thus, I will not include receptors in muscles
and joints as part of the tactile sense. When these are
included, one usually uses the term haptics. On the other
hand, tactile information is not restricted to ‘touching’
(i.e., feeling objects), but also comprises (passive) vibrotactile stimulation of the skin and temperature
Employing the tactile modality has several potentially
useful applications and advantages in VE, including the
1. The quality of the VE and user performance is likely
to improve if the information that is available to the
tactile sense in real life is present in the VE as well.
This is certainly true for information that is
predominantly perceived with the tactile channel,
such as roughness of objects, and small vibrations.
2. Employing the tactile sense will enlarge the
immersion of the observer in the VE. The VE is more
complete, and sensory information may become
congruent: I can feel what I see.
3. Tactile information can guide movements. An
example is the potential role of tactile information in
grasping. Users may have trouble in estimating the
distance between their (virtual) hand and the object
they want to grasp. Presenting a tactile gradient (i.e. a
tactile intensity or frequency field around the object)
which guides the user to the object and indicates the
Euclidian distance between the object and the user's
hand might support the degraded visual information
in VEs. After grasping the object, tactile information
may be used to indicate how much force must be
applied to the object (see next point).
4. Tactile information can be a substitute for force
feedback. Force feedback is essential for adequate
user performance in interacting with virtual objects
(e.g., instruments and weapons), but is also very
difficult to present with contemporary VR
technology. Tactile information as a substitution for
force feedback has already proven its effectiveness in
remote control situations.
5. The tactile sense may be helpful in overcoming the
weak points that even state-of-the-art VE systems
still have. For example, the field of view of the
visuals is still reduced compared to real life; using
the tactile sense to compensate for the lack of
peripheral viewing is one of the possibilities.
6. Finally, the tactile modality may be used as a general
information channel to present VE-related but not
specific information, e.g., warning information.
For all these applications fundamental and applied
knowledge is required for successful use in VEs, and
moreover, for successful development of devices. At this
moment, not all this knowledge is available or
applicable. Areas that deserve attention include:
• body loci other than hand and fingers,
• sensory congruency (below, an example shows that
this doesn’t come naturally),
• cross-modal interaction,
• perceptual illusions,
• attention.
A simple experiment by Werkhoven and Van Erp (1998)
showed that visual and tactile information is not always
perceived consistently. They investigated the perception
of open time intervals, either marked by visual stimuli
(blinking squares on a monitor) or tactile stimuli (bursts
For correspondence with the author: [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
of vibration on the fingertip. They compared standard
intervals of 200 ms with uni- and cross-modal intervals,
as is schematically presented in Figure 1 for the crossmodal condition.
The results of this experiment showed a large bias in the
cross-modal condition: tactile time intervals are
overestimated by 30% (see Figure 2). This indicates that
sensory congruency is a non-trivial aspect of integrating
sensory modalities in a VE.
Examples of tactile displays
This section gives a small and far from complete
overview of tactile display applications (see also Van
Erp & Van den Dobbelsteen, 1998). It focuses on two
application areas: that of sensory substitution and
navigation displays. This restriction is made, because
displays developed for use in VE are regularly described
in the open literature, e.g. see Boman (1995) or Ziegler
Overview of the paper
This paper focuses on the use of the tactile modality to
present navigation (i.e., direction) information. This
application can help VE users in orientating in VEs,
which may be difficult on the basis of restricted visual
information only.
In the next section of the introduction, some examples of
tactile displays are given. Chapter 2 describes some
basic neurophysiology and psychophysical knowledge.
An example of cataloguing spatio-temporal characteristics is given in chapter 3. Here, the spatial characteristics of the torso are described, including experimental
data. This cataloguing is of primary interest for the
application that is described in Chapter 4: using the torso
to present tactile navigation information. The torso has
three important advantages in this respect. First, it has a
large surface, reducing the need to minimise actuator
size or to keep the number of actuators low. Second,
information presented to the torso does not interfere with
actions performed with the hands, like controlling input
devices. And third, the torso is a volume, and thus
a priori interesting for presenting 2D or 3D information,
like geographical or navigational information.
Sensory substitution
Some examples of the earliest displays providing
complex stimuli are aids for the blind, including
miniature matrices of point stimuli used for reading of
text and pictures. ‘Tactile imaging’ is the process of
turning a visual item, such as a picture, into a touchable
version of the image, so that this tactile rendition
faithfully represents the original information.
Interval 1
Interval 2
Figure 1: Schematic presentation of the stimuli to
investigate the perception of open time intervals. The
intervals are marked by visual stimuli (marked V) or
tactile stimuli (marked T)
PSE (ms)
visual - visual
tactile - tactile
visual - tactile
Figure 2: Point of subjective equality for a 200 ms
standard open time interval experiment. The visual —
tactile condition shows that a 150 ms tactile interval is
judged to be equal in length to a 200 ms visual interval
• The Optacon. One of the most successful devices to
present ‘visual’ information to the blind was an
ink-print reading machine, the Linvill-Bliss Optacon
(OPtical-to-TActile CONverter). Bliss and his
associates (Linvill & Bliss, 1966; Bliss et al., 1970)
developed this reading device, which converts
printed materials into vibratory patterns. With the aid
of a small camera containing a matrix of 6 by 24
photocells, the device converts the image electronically to a tactile display, placed on the skin of a
• The Kinotact. Craig (1974) studied letter-shape
perception with the aid of a 10 by 10 matrix of
vibrators placed against the observer’s back. The
encoding system, called ‘Kinotact’, was a 10 by 10
matrix of photocells, wired one-to-one with the
vibrators. With the presentation of the tactile image
of block letters, subjects learned to identify this
‘pictorial mode’ letter patterns to an average criterion
of 80-90% correct in 300 trials. For related research,
see also Loomis (1974), and Craig (1980).
• TVSS. Bach-Y-Rita (1972) and associates developed
the Tactile Vision Substitution System (TVS system),
in which a visual image picked up by a TV camera is
transformed into a tactile one by means of a 20 by 20
matrix of vibrators mounted on the back of a dental
chair. It was found that subjects could immediately
recognise vertical, horizontal and diagonal lines.
Experienced users could identify common objects
and people’s faces. This is an example of a
perceptual phenomenon called distal attribution, in
which an event is perceived as occurring at a location
other than the physical stimulation site. With
self-induced camera movement, subjects use the
camera as part of a perceptual organ and learn to
locate the percepts subjectively in space, rather than
on the skin.
Another TVS system, called the Electrophthalm,
developed by Starkiewicz, Kuprianowicz and
Petruczenko (1971) is more applicable to space orienta-
tion and presents a 12 by 8 tactile image to the forehead.
However, TVS systems are not useful for acquiring
information from ‘cluttered’ visual environments and are
not presently useful for navigation purposes.
• Desktop tactile displays. The formerly described
systems are not designed to provide computer access
to the visually impaired, and are rarely used due to
uncomfortable or impractical displays and inefficient
information transfer (Kaczmarek & Bach-Y-Rita,
1995). An example of a new generation display,
which I will call desktop tactile displays, is the
Moose. This display is especially designed to provide
computer access. A prototype developed by
O’Modhrain and Gillispie (1997) presents a haptic
representation of a screen by reflecting forces when
navigating across the screen. Desktop tactile displays
are nowadays widely available in the consumer
electronics shops for as little as 100 US$.
Tactile navigation displays
A second important application of tactile displays is as
navigation display. Gilliland and Schlegel (1994)
conducted studies to explore the use of vibrotactile
stimulation of the human head to inform a pilot of
possible threats or other situations in the flight
environment. Rupert, Guedry and Rescke (1993) developed a matrix of vibro-tactors that covers the torso of the
pilot’s body (http://www.accel.namrl.navy.mil/default.
html). This prototype may offer a means to continuously
maintain spatial orientation by providing information
about aircraft acceleration and direction of motion to the
pilot. Within the pitch and roll limits of their torso
display (15° and 45°, respectively), the subjects could
position the simulated attitude of the aircraft by the
tactile cues alone. The Tactor Evaluation System (TES,
Engineering Acoustics Inc.) was developed to
demonstrate the use of vibrotactile information for divers
in conditions of low visibility: real time navigational
information (course, distance, and cross-track error) and
alarm information. Five tactors were used: left and right
side, back and chest, and on a wrist for miscellaneous
signals (http://www.eaiinfo.com/).
2. Cataloguing spatial sensitivity
An important parameter in the design and application of
tactile displays is the spatial resolution. There are two
main areas involved in spatial sensitivity research:
neurophysiology and psychophysics. Important determinants of spatial sensitivity are the sizes and forms of
the receptive fields of the mechanoreceptors, and the
representation of the body surface in the (somatosensory) cortex. This neurophysiological data is
presented in Section 2.1. The psychophysical measures
of spatial sensitivity used throughout the years and
experimental findings are presented in Section 2.2. For a
more elaborate overview, see for example Van Erp and
Vogels (1998). Basic research on the spatial sensitivity
of the torso for vibro-tactile stimuli (relevant for the
application under study) is presented in Chapter 3.
2.1 Neurophysiology
A comprehensive overview basic neurophysiology can
be found in Kandel et al. (1991). An important
contribution of this research area has been the
determination of the density of receptors, and the size
and form of the receptive field of a single peripheral
nerve fibre. Micro-neurographic recordings from nerves
innervating the glabrous skin have isolated four groups
of mechanoreceptive fibres (see Table 1 for an
After contacting a single afferent unit, a systematic
exploration of the receptive field is undertaken.
Unfortunately, this technique is only applied for the
human arm and hand; no data on the trunk are available.
Furthermore, the technique provides information on
single peripheral nerve fibres only, not on the spatial
sensitivity of the cutaneous sense as a whole. Applied to
the Pacinian body, the receptive field proves to be large,
with poorly defined borders and a single point of
maximum sensitivity. Even for the fingers, receptive
fields can be in the order of several square cm
(Bolanowski et al., 1988; Valbo & Johansson, 1978).
Table 1: Characteristics of the four types of
mechanoreceptive fibres in the human skin
fast adapting
Meissner corpuscle
• small receptive field
• NP I channel, not
sensitive to temperature
• 10–100 Hz
• temporal summation:
• spatial summation: yes
• local vibration and
perception of localised
Pacinian corpuscle (PC)
• large receptive field
• P-channel, very
sensitive to temperature
• 40–800 Hz
• temporal summation:
• spatial summation: yes
• perception of external
slowly adapting
Merkel cell (SAI)
• small receptive field
• NP III channel, sensitive
to temperature
• 0.4–100 Hz
• temporal summation: no
• spatial summation: no
• tactile form and
Ruffini ending (SAII)
• large receptive field
• NP II channel, sensitive
to temperature
• 15–400 Hz
• temporal summation:
• spatial summation: ?
• not in glaborous skin
Besides the receptive field sizes of single afferent nerve
fibres, one has also determined the receptive field sizes
of the different cortical regions involved in cutaneous
2.2 Psychophysics
Within psychophysics, two classic measures are applied
to determine the spatial resolving power: the two-point
limen (participants have to judge whether a stimulus
consists of one or two points) and the error of
localisation (e.g. participants judge two successive
contacts as the same or different in locus). Both methods
know different variants. Unfortunately, little data are
After the work of Weber and Vierodt, little attention was
given to this field until the 1960s. Weinstein (1968)
measured (pressure-) thresholds of two-point discrimination and tactile point localisation on several body loci.
Both thresholds were highly correlated, however. Acuity
found with two-point discrimination was three to four
times lower than with point localisation. Because the
methods of two-point discrimination and point
localisation are measures for spatial acuity and hyper
acuity, respectively, the results are in accordance with
data on visual acuity (e.g. see Snippe, 1991). Furthermore, Weinstein found significant effects of body locus.
Lowest thresholds were found for the fingertips: 2.5 mm
and 1.5 mm for two-point discrimination and point
localisation, respectively. Thresholds for the trunk were
approximately 40 mm and 10 mm, respectively.
Sensitivity decreased from distal to proximal regions:
fingers, face, feet, trunk, upper and lower extremities.
Thresholds correlated with the relative size of cortical
areas subserving a body part. Another important
observation was that good two-point discrimination did
not necessarily mean good sensitivity to pressure. Vierck
and Jones (1969; Jones and Vierck, 1973) stated that the
method of the two-point limen leads to an underestimation of the skin's real spatial sensitivity. They
showed that the discrimination of area stimuli and length
stimuli is about ten times better. In the 1970s, Loomis
and Collins (1978) found comparable results when the
stimulus was a gradual shift in the locus of stimulation.
Johnson and Phillips (1981) introduced alternative
methods, and measured two-point thresholds, gap
detection and discrimination of grating orientation for
the fingertips. They found thresholds of 0.87 mm and
0.84 mm, respectively. These results show that the
ability of subjects to discriminate stimuli is much finer
than is indicated by the two-point threshold of Weinstein
3. Cataloguing vibro-tactile spatial resolution on
the torso
Since only indirect data are available regarding the
spatial resolution of the torso for vibro-tactile stimuli,
basic research was needed to formulate the optimal
display configuration. On the one hand, one wants to use
the full information processing capacity that is available;
on the other hand, one wants to keep the number of
actuators to a minimum. Therefore, a concise discussion
of a series of experiments is presented (for details, see
Van Erp & Werkhoven, 1999).
Four male subjects (age range 28–39 years, mean 31)
participated voluntarily. In the experiment, 11 vibrotactile actuators were attached to the torso with sticky
tape (see Figure 3). The participants performed a
localisation task: Two stimuli were presented to the torso
and the participant was asked to judge the location of the
second compared to the first (left/right). The stimuli
were first presented to the dorsal side of the torso, and in
a second session to the frontal side. The inter stimulus
interval (ISI) was varied (0 ms, 56 ms, 196 ms, and 980
ms), as was body locus within a torso side (left, middle,
and right). The latter indicates the location of the
standard stimulus; each standard was combined with
four comparison stimuli. The responses of the subject to
each standard-comparison pair were counted in
proportion ‘to the right’ responses. These summarised
data were fitted to a cumulative normal distribution,
resulting in two parameters: µ (or bias) and σ (or
threshold), see Figure 4.
Figure 3: Placement of the tactile actuators on the back
proportion (C to the right of S)
available on vibro-tactile perception and on loci other
than the hand.
Weber and Vierodt did the first psychophysical research
on spatial acuity in the nineteenth century. It was Weber
who introduced the two-point limen and the localisation
error (Weber, 1834). Mapping of the whole body
revealed large differences in spatial acuity between
different parts of the body. Vierodt (1870) generalised
this to the ‘law of mobility’, which states that the twopoint limen improves with the mobility of the body part.
2 left
1 left
1 right
2 right
position of the comparison
Figure 4: Psychophysical method to determine the bias
(mu) and sensitivity (sigma) for a specific standard (S)
The results of the experiment (see Figure 5) showed that
the sensitivity for vibro-tactile stimuli presented to the
ventral part of the torso was larger than for stimuli
presented to the dorsal part. Furthermore, the effect of
body locus was present on both the frontal and the dorsal
part: the sensitivity near the middle is larger than to the
sides. Moreover, the sensitivity is larger than expected
on the basis of the psychophysical literature. The effect
of ISI showed that sensitivity increases with increasing
dorsal side
frontal side
(1 actuator unit equals 2 cm)
threshold (actuator units)
position of the standard
Figure 5: Results of the spatial accuracy of the torso
for vibro-tactile stimuli
4. Example of implementing a tactile display:
presentation of spatial information on the torso
When the first phase, cataloguing relevant perceptual
characteristics, is finished, basic research into possible
applications becomes actual. As discussed in the
introduction, the torso may be well suited to present 2D
geographical information. In the following experiment,
tactile actuators were attached around the participants
torso (except for the region around the spine, see also
Figure 6). During the experiment, one actuator was
activated. The observer could adjust a cursor to indicate
the external direction suggested by the actuator (see
Figure 7 for the experimental set-up).
Figure 6: Method to ensure correct
placement of the actuators
This direction determination task resulted in two
parameters: a bias in the indicated direction, and
variability in the answers (expressed in the standard
deviation of the responses). The latter parameter is of
course a measure of the precision with which the
observer perceives the stimuli.
Figure 7: Top view of the set-up for the direction
discrimination task. With a dial, the observer can
position a cursor (a dot projected from above) along a
white circle drawn on the table. The cursor should be
positioned such that it indicates the direction of the
tactile stimulus
The results are interesting in several ways. First of all,
none of the participants had any trouble with the task.
This is noteworthy since a point stimulus does not
contain any explicit direction information. The strategy
people use is probably equivalent to that of visual
perception, namely using a perceptual ego-centre as
second point. Several authors determined the visual egocentre (e.g., Roelofs, 1959), which can be defined as the
position in space at which a person experiences himself
or herself to be. Identifying an ego-centre or internal
reference point is important, because it co-ordinates
physical space and phenomenal space. A second reason
to determine the internal reference point in this tactile
experiment was the striking bias all ten participants
showed in their responses, namely a bias towards the
sagittal plane. This means that stimuli on the frontal side
of the torso were perceived as directions coming more
from the navel, and stimuli on the dorsal side of the torso
were perceived as coming more from the spine. Further
research showed that this bias was not caused by the
experimental set-up, the visual system, the subjective
location of the stimuli, or other anomalies. The most
probable explanation is the existence of two internal
reference points: one for the left side of the torso, and
one for the right side. When these internal reference
points are determined as function of the body side
stimulated, the left and right points are 6.2cm apart on
average across the ten participants, see Figure 8.
longitudinal position (mm)
5. Discussion
lateral position (mm)
Figure 8: The Internal Reference Points for the ten
observers in the tactile direction determination task
The third noteworthy observation is related to the
variance of the responses as function of the presented
direction. As Figure 9 shows (lower values indicate
better performance), scores in the front-sagittal region (–
50°—+50° in the graph) are very good with standard
deviations between 4° and 8°, and somewhat lower
towards the sides.
Standard deviation (deg.)
located in between two simultaneously presented
stimuli) is as good as that of real points,
• small changes in the perceived direction can be
evoked by presenting one point stimulus to the
frontal side, and one to the dorsal side of the
Potential beneficial areas of tactile displays in VE
systems were presented in Chapter 1. After choosing
what information the tactile display must be designed for
to present, the relevant perceptual characteristics of the
users must be determined. Although there is substantial
literature on tactile perception, the available knowledge
isn’t by far as complete as on visual and auditive
perception. Gaps in the required knowledge, e.g. on
tactile perception of body loci other than the arms,
hands, and fingers, must be filled before applications can
be successful. Besides data on fundamental issues such
as spatial and temporal resolution, perceptual illusions
might be an interesting area in relation to display design.
Illusions such as apparent position (which may double
the spatial resolution of a display), and apparent motion
(which allows to present the percept of a moving
stimulus without moving the actuators) offer great
opportunities to present information efficiently. Still
more illusions are discovered (e.g., Cholewiak &
Collins, 1999). After cataloguing all relevant basic
knowledge, specific applications must be studied to
further optimise information presentation and display
use. Another important point, which is not fully
addressed in this paper, is the interaction between the
sensory modalities, and sensory congruency. An
enhanced VE will be multi-modal, but the interaction
between the tactile and the other senses is an area, which
is only recently being addressed.
When these steps are taken carefully, tactile displays
may enhance the experience and effectiveness of the VR.
-130 -110 -90 -70 -50 -30 -10 10 30 50 70 90 110 130
-120 -100 -80 -60 -40 -20
20 40 60 80 100 120 140
Stimulus angle (deg.)
Figure 9: Standard Deviation of the tactile responses as
function of the stimulus angle. The horizontal lines
summarise the results of the post hoc test; pairs of data
points significantly differ when separated by two lines
Other experiments and analysis with the same display
are discussed more elaborately elsewhere (Van Erp,
2000). Relevant implications for the application of tactile
displays for spatial information are the following:
• observers can perceive a single external tactile point
stimulus as an indication of direction,
• although the consistency in the perceived direction
varies with body location, performance near the
sagittal plane (SD of 4°) is as good as with a
comparable visual display,
• direction indication presented by the illusion of
apparent location (the percept of one point stimuli
Bach-Y-Rita, P. (1972). Brain mechanisms in sensory
substitution. New York: Academic.
Bliss, J.C., Katcher, M.H., Rogers, C.H. & Shepard, R.P.
(1970). Optical-to-tactile image conversion for the
blind. IEEE Transactions on Man-Machine Systems,
MMS-11, 1, 58-64.
Bolanowski, S.J., Gescheider, G.A., Verrillo, R.T. &
Checkosky, C.M. (1988). Four channels mediate the
mechanical aspects of touch. Journal of Acoustical
Society of America, 84 (5), 1680-1694.
Boman, D.K. (1995). International Survey: VirtualEnvironment Research. IEEE, June 1995, pp 57-65.
Cholewiak, R.W. & Collins, A.A. (1999, in press). The
generation of vibrotactile patterns on a linear array:
influences of body site, time, and presentation mode.
Perception and Psychophysics.
Craig, J.C. (1974). Pictorial and abstract pictorial
cutaneous displays. In F.A. Geldard (Eds.),
Cutaneous communication systems and devices.
Austin, Texas, psychonomic society.
Craig, J.C. (1980). Modes of vibrotactile pattern
recognition. Journal of Experimental Psychology,
Human Perception and Performance, 6 (1), 151-166.
Gilliland, K., Schlegel, R.E. (1994). Tactile Stimulation
of the Human Head for Information Display. Human
Factors, 36 (4), 700-717.
Johnson, K.O. & Phillips, J.R. (1981). Tactile spatial
resolution. I. Two point discrimination, gap
detection, grating resolution, and letter recognition.
Journal of Neurophysiology, 6(6), 1177-1191.
Jones, M.B. & Vierck, C.J. (1973). Length discrimination on the skin. American Journal of Psychology,
86, 49-60.
Kaczmarek, K.A. & Bach-Y-Rita, P. (1995). Tactile
displays. In W. Barfield & T.A. Furness III (Eds.),
Virtual environments and interface design (pp. 349–
414). New York: Oxford University Press.
Kandel, E.R., Schwartz, J.H. & Jessel, T.M. (1991).
Principles of Neural Science. New York: Elsevier
Science Publishing Co.
Linvill, J.G. & Bliss, J.C. (1966). A direct translation
reading aid for the blind. Proceedings IEEE, 54,
Loomis, J.M. (1974). Tactile letter recognition under
different modes of stimulus representation.
Perception and Psychophysics, 16, 401-408.
Loomis, J.M. & Collins, C.C. (1978). Sensitivity to
shifts of a point stimulus: An instance of tactile hyper
acuity. Perception and Psychophysics, 24, 487-492.
O’Modrain, M.S. & Gillispie, B. (1997). The moose: a
haptic user interface for blind persons. Internet.
Roelofs, C.O. (1959). Considerations on the visual
egocentre. Acta Psychologica, 16, 226-234.
Rupert, A.H., Guedry, F.E. & Reschke, M.F. (1993). The
use of a tactile interface to convey position and
motion perceptions. AGARD meeting on Virtual
interfaces: reserach and applications, October 1993.
Snippe, H.P. (1991). Human perception of spatial and
temporal luminance structure. PhD thesis, Utrecht
University, The Netherlands.
Starkiewicz, W., Kuprianowicz, W. & Petruczenko, F.
(1971). 60-channel elektroftalm with CdSO4mphotoresistors and forehead tactile elements. In Sterling et
al. (Eds.), Visual prothesis: the interdisciplinary
dialogue. New York: Academic press, 295-299.
Valbo, Å.B. & Johansson, R.S. (1978). The tactile
sensory innervation of the glabrous skin of the
human hand. In Gordon (Ed.), Active touch: the
mechanisms of recognition of objects by
manipulation. A multi-disciplinary approach (pp.
29-54). Oxford: Pergamon press.
Van Erp, J.B.F. (2000, in press). Accuracy of and bias in
2D direction perception of tactile stimuli presented to
the torso (Report TM-00-B0?). Soesterberg, The
Netherlands: TNO Human Factors.
Van Erp, J.B.F. & Dobbelsteen, J.J. van (1998). On the
design of tactile displays (Report TM-98-B012).
Soesterberg, The Netherlands: TNO Human Factors.
Van Erp, J.B.F. & Vogels, I.M.L.C. (1998). Vibrotactile
perception: a literature review (Report TM-98B011). Soesterberg, The Netherlands: TNO Human
Van Erp, J.B.F. & Werkhoven, P.J. (1999). Spatial
characteristics of virbo-tactile perception on the
torso (Report TM-99-B007). Soesterberg, The
Netherlands: TNO Human Factors.
Vierck, C.J. & Jones, M.B. (1969). Size discrimination
on the skin. Science, 163, 488-489.
Vierodt, K.H. (1870). Abhängigkeit der Ausbildung des
Raumsinnes der Haut von der Beweglichkeit der
Körpertheile. (Dependence of the development of the
skin's spatial sense on the flexibility of parts of the
body). Zeitschrift für Biologie, 6, 53-72.
Weber, E.H. (1834). De pulsu, resorptione, auditu et
tactu. In H.E. Ross & D.J. Murray (Eds.), E.H.
Weber, on the tactile senses. Hove, UK: Taylor &
Weinstein, S. (1968). Intensive and extensive aspects of
tactile sensitivity as a function of body-part, sex and
laterality. In D.R. Kenshalo (Ed.), The Skin Senses
(pp 195-218). Springfield: C.C. Thomas.
Werkhoven, P.J. & Van Erp, J.B.F. (1998). Perception of
vibro-tactile asynchronies (Report TM-1998-B013).
Soesterberg, The Netherlands: TNO Human Factors.
Ziegler, R. (1996). Haptic Displays — How can we feel
Virtual Environments? — Imaging Sciences and
Display Technology, SPIE Proceedings vol. 2949, pp
Virtual Cockpit Simulation for Pilot Training
Dipl-Ing. Kai-Uwe Dörr, Dipl.Inform. Jens Schiefele & Prof. Dr.-Ing. W. Kubbat1
Institute for Flight Mechanics and Control
Technical University Darmstadt
Petersenstraße 30
64287 Darmstadt
For some of today’s simulations very expensive, heavy,
and large equipment is needed. Examples are driving,
shipping, and flight simulators with huge and expensive
visual and motion systems.
In order to reduce cost, immersive ‘Virtual Simulation’
becomes very attractive. Head Mounted Displays
(HMD) or CAVEs (Computer Animated Virtual
Environments), Datagloves, and cheap ‘SeatingBucks’
are used to generate a stereoscopic virtual environment
(VE) for the trainee.
IVS enhances training quality and quantity for
classroom-teaching and Computer Based Training
(CBT). It allows to visualize and animate teachingmaterial in a more natural stereoscopic environment.
Data of before unseen complexity can be revealed and
complex models easily visualized. For the first time, the
trainee himself can interact with a Data-Glove in the
environment and collect cockpit experience long before
his maiden flight. CAVEs and Immersive Projection
Screens enable “group training” to collect personal and
shared experience while further enhancing training
With increasing maturity of VR-gear IVS will allow to
generate new training metaphors for immersive flight
simulation. This might include the enhancement or
partial replacement of conventional flight simulators by
High fidelity pilot training simulators are designed as
training tools for one specific aircraft type. They demand
authentic instrumentation and system layout for the
simulated aircraft type including huge outside vision
systems and cumbersome motion systems.1 Because of
these reasons, traditional simulators are very expensive,
inflexible, and difficult to reconfigure. The high cost
factor in buying and maintaining them causes air carriers
to purchase either just a single simulator for every
aircraft type they own or to buy expensive training hours
from other companies.1
Virtual Simulation
To overcome some of the problems in the field of pilot
training the Air Force Institute of Technology developed
a Virtual Cockpit (VC) for fighter pilot training.2 Pilots
are immersed in a stereoscopic VE, wearing a HMD and
a pointing device to interact with the virtual cockpit
devices.3 A VC can be easily reconfigured by simply
switching the cockpit model database and the attached
flight mechanics.4 At the Institute for Flight Mechanics
and Control, Darmstadt University of Technology this
concept was extended to be suitable for an Airbus A340
Cockpit-IVS using hi-resolution HMD, “Seating Buck”,
cyberglove, and stereoscopic projection screens for a
natural interaction metaphor.5 As pilot outputs for
principle navigation and Instrument Flight Rule (IFR)
testing a virtual Primary Flight Display (PFD), a virtual
Navigation Display (ND), and a virtual civil Head-Up
Display (HUD) are available. In addition, a simplified
outside visual is rendered to the pilot. These displays are
sufficient to run principle Instrument Flight Rule (IFR)
tests with the virtual cockpit.
The problem of lacking force feedback in IVS was
significantly reduced by developing a “Seating Buck”.6
Only side-stick, pedals, flap-lever, and thrust-lever are
physically available. All other buttons, dials, and
switches are simulated by simple plastic panels. In a test
series the concept and implementation proved to reduce
interaction time significantly.1,6
Other examples for operator training using Virtual
Training methods are Astronaut training to repair the
Hubble Space Telescope7, submarine outlook training to
practice maneuvering in a harbor8, support pilots
classroom education9, or caterpillar training. Instead of
HMDs very often CAVEs10 and BOOMs11 are used to
avoid heavy intrusive head gear and limited Field of
Views (FOV).
Human Machine Interface in Virtual Simulation
Transfer of training from virtual into real space still has
to be proven for pilot training. For simple Cola can
sorting in a CAVE transfer of training from virtual into
real space was shown.12,13 Also, people trained in VR
have a better orientation in buildings than map trained
persons14. Therefore, it can be assumed that training in
virtual environments might be useful to train trainees at
different requirement levels.
Training quality limiting factors due to today’s hardware
equipment such as Field of View (FOV)15,16,17,21, tracker
For contact with authors: [email protected]; [email protected]; [email protected]; tel. +49 6151 16
2890, fax +49 6151 16 5434
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
latency18, presence19, and missing force feedback1,6 were
investigated in principal. Only few research exists
determining the limitations on training caused by the
complete VR-Human Machine Interface (HMI) design
and hardware.20
VR-HMI research has already been conducted
concerning force feedback, HMD FOV, and HMD
resolution for Cockpit-IVS.1,6,21
A good general overview describing VR-HMI research
is presented in 22.
Conventional Computer-Based Training
Computer Based Training and Procedure Training (PT)
use PCs with a 2D image, sound, a mouse, and a
keyboard for interaction. A trainee sits in-front of a PC
screen and interacts by clicking with the mouse. CBT is
split into different chapters such as radio navigation,
flight planing, flight performance, electronics,
instrumentation, and engines. Further enhanced systems
allow partial simulation of functionality. For each
individual aircraft type a different program is available.
Each individual chapter is split into different learning
• Overview
• Components and Control
• System Operation
• Abnormal Operation
• Summary
• Mastery Test
In different learning units the trainee gets a multimedia
presentation of the learning material. After the
introduction, the trainee can interact with the system by
clicking with the mouse on interaction devices. In the
Mastery Test multiple choice questions have to be
answered and tasks performed. The trainee can practice
all units on his own personal learning pace. The test can
also be individually repeated.
Such a training metaphor helps to support individual
training. “Fast learner” are not frustrated by a low pace
and “slow learner” are not overrun.
CBT with a VR Training Environment
In order to enhance CBT, a 3D Virtual Cockpit model is
generated. All interaction devices such as side stick,
pedals, thrust-lever, knobs, buttons, and dials are
modeled as 3D geometry. All other parts and surfaces
are formed by simple textured geometry.5 This 3D model
is rendered to a pilot wearing a tracked high-resolution,
large field of view (FOV), stereoscopic HMD.
Figure 1: Flight management computer training
with CBT
The trainee does not have any immersive experience
towards the real geometry and functionality of the
cockpit. The position of interaction devices in real 3D
space is unknown to him. Familiarization in 3D space
can not be realized with today’s CBT systems.
Figure 2: IVS Cockpit based on modeled geometry
and textures
For interaction the pilot wears a tracked data-glove
recognizing hand position, orientation, and finger
bending. The trainee can virtually interact with all
cockpit devices, dials, and buttons. The system response
on the input can be visualized.
The same image is also rendered to a large stereoscopic
projection screen (Shutter classes) enabling observers to
watch the trainee and his interaction. This allows later
discussion on the trainees performance.
Figure 3: Demonstration room with projection screen
(three shutter signals)
The same concepts and structures known from ordinary
CBT are applied. The only difference is that the trainee
is immersed into the scene allowing him to interact
naturally with his environment. Learning aids and eye
catchers such as symbols, markers, and any desired
virtual information can be visualized within the 3D
virtual cockpit as well. For instance, after toggling the
gear lever the virtual gear unit is displayed and the
actuator changes visualized. Therefore, beyond the VR
simulation of an ordinary cockpit, virtual information
can be incorporated and used as didactical metaphor for
pilot training.
Therefore two different Training methods are feasible
with this environment. In the so called Class Room
Training, collective and cooperative learning in front of
a single projection screen enables trainees to work and
learn together in the same environment.
As a second training method, a single VR-Pilot training
environment was developed. Therefore the trainee wears
a HMD and a Data Glove. Both devices were tracked
with a tracking system. So a naturally interaction with
the virtual scene is possible. The trainee is guided trough
the different lessons depending on his interaction with
the cockpit The trainee itself define which lesson he
wants to work through . Additionally it is possible for
the trainee to fly with this virtual cockpit because of its
full functionality. Both Methods uses the same Training
lessons with minimum changes in interaction
possibilities. Example lectures were realized for both
training methods and will be described later on.
Classroom Training
The didactical methods for training vary depending on
the airlines and the training facilities. Training is often
based on the conventional concept of “frontal teaching”.
Teachers give lectures with varying didactical materials.
Dependent on the training facility this can be simple
transparencies, video-tapes, sketches, boards, and small
mockups. After each lecture the trainee has the
possibility to re-read the taught lecture from printed
material. At the end of each training chapters pilots must
pass a written multiple choice test.
Figure 4: Stereoscopic projection screen
for classroom training
Hence, the understanding of complex aircraft systems
strongly depends on the imagination of a trainee and the
teaching skills of a teacher. In order to enhance teaching
quality stereoscopic projection screens can be used to
visualize complex aircraft systems and technical
dependencies in a natural way. A teacher can fly through
a model, hide obstructing parts, or animate complex
functionality’s. The trainee himself becomes part of the
scene. In after lecture sessions the trainee can interact
with the system and its functionality. In front of an
immersive screen the trainee becomes part of the scene
and experiences the learn material more naturally.
Stereoscopic vision with depth perception enables new
pilots to easily asses complex 3D structures, aircraft
positions, etc. The “hands-on experience” helps to
deepen the understanding and motivate the trainee to
explore deeper into the learning material. Group
experience and group training can be enhanced with an
stereoscopic projection system. This might accelerate
memorization and pushes the later needed ability of
cooperative cockpit work through group experience.
VR Pilot Training
For today’s simulator training huge and expensive
simulators are needed. Each training hour costs up to
$5,000 and existing training facilities are currently at the
limits of their capacities. Additional to the later on
described training lessons, the system is also fit for use,
to train some real flight tasks. Therefore, the Virtual
Cockpit (VC) based on the above described technique
(HMD plus Stereoscopic Projection Screen) can be used.
In addition to the CBT approach, a simplified outside
visual is added to generate an immersive flight
simulation. The viewing distance has to be reduced to
approximately 20km in order to ensure sufficient
rendering performance (15–20Hz). A virtual Primary
Flight Display (PFD), Navigation Display (ND), and a
virtual stereoscopic Head Up Display (HUD) are used in
a first approach.21
Figure 5: Primary flight display and navigation display
These virtual displays show basic information necessary
to perform a controlled flight and allow basic
performance analysis with the system.
Aircraft System Lesson
Aircraft system knowledge is one of the main training
parts during theoretical pilot education. Simple system
diagrams are used to give the trainees an overview of the
whole system. Relations between aircraft subsystems
and how these systems work together will be explained
in the same way. This is not a very intuitive way of
learning. The best way to learn is to visualize them. The
visual channel is the most significant way to comprehend
The main advantage of VR-systems is the possibility to
display trainees a 3D geometry of an object and a
simulation of the real behavior. As an example the
behavior of gear, flaps, and rudder on an input from the
pilot is shown. Therefore, a complete aircraft model is
shown to the trainee (Figure 6). The model shows the
reaction of the aircraft and it is possible for the trainee to
zoom in different subsystems.
Figure 7: Gear view
In this example (Figure 7) the trainee can observe the
kinematics of the gear. He can imagine the 3D behavior
of the actuators during gear up and down procedure. So,
in case of a system failure, he can imagine what is
happen and where the errors can be. The system
knowledge increases because of the 3 dimensional form
of presentation.
Engine Lesson
During pilot education engines are visualized by
explosion sketches or vertical cuts through an engine.
This creates a complex visualization of the parts. For
instance, explaining the turbine turn rate at N1 and N2 is
rather difficult. Either the graphical representation is
showing too much or too view detail, forcing the
instructor to switch between several images.
Figure 8: Conventional visualization with a
cut through an engine
Figure 6: Aircraft outside view above the pedestal
Therefore, a lecture was developed that allows to
dismantle the engine from a full blown representation
down to the necessary key elements such as fan, turbine
stators, turbine shaft, and combustion chamber. Trainees
observe the animated engine and can position themselves
on arbitrary positions.
viewing among two pilots simulated in the same IVS is
Figure 10: Stereoscopic projection screen rendering
scene visible to the pilot
Figure 9: 3D engine inside view
Alternatively the they can follow a prerecorded flight
path through the scene. During the flight they can
arbitrary change the viewing direction. They can always
stop and move towards any object to get a closer look.
To increase realism and a feeling for object sizes, a
complete aircraft is rendered as well.
Stereoscopic vision enables trainees to be immersed into
the environment generating a closer and better
impression of the turbine. It enables the trainees to
achieve a feeling for real part and turbine sizes. With
ordinary paper sketches this is impossible. As an
enhancement, observers can be re-scaled to small sizes
allowing to closely observe small turbine parts and their
Based on the stereoscopic vision, instrument locations
and attached functionality can be memorized by
generating a mental map of the cockpit.
Force Feedback/Vision
It was determined that lacking force feedback in pure
IVS is a major usability limitation.6 Therefore, some
devices are physically available such as sidestick, pedals,
and thrust-lever. All others are replaced by simple plastic
panels to generate force feedback to the pilot (Seating
Buck).1 The Seating Buck can be easily reconfigured to
simulate arbitrary cockpit configurations.
With a Seating Buck force feedback device the
interaction time is reduced significantly providing a
more natural haptical feedback to the pilot.6
For the success of a VR CBT enhancement a large FOV
of more than 80° is needed.21 Above a 60° FOV pilots
can assess all visible information and geometry in the
cockpit. Above a 80° FOV also orientation and cross-
Figure 11: Seating buck to simulate force feedback
VR CBT and PT system can help to reduce education
cost by reducing expensive simulator hours for
familiarization and principle interaction training.
Otherwise, it can serve as an extension to already proven
CBT training concepts. The VR technology and
projection screen technology is mature enough to fulfil
these tasks. HMD with the requested FOV, force
feedback devices, and computers with sufficient graphics
power exist. However, real verification of training
transfer has to be further investigated in the future.
Flight Simulation
All ordinary software simulation modules known from
conventional flight simulation such as physical input
devices, virtual input devices, flight mechanics, traffic,
and rendering run in a distributed environment on
different high end graphics work stations. As simulation
module an Airbus A300 flight mechanics, ground
collision , weather, and sound modules are available. All
the modules are taken from a conventional flight
simulator available at the Institute for Flight Mechanics
and Control.23 It can be also used for comparing the VCS
with a real flight simulation.
The used VR equipment is critical for the success of
VCS. In principal most system components such as force
feedback generation, HMD FOV and resolution are
proven to be usable. Tracker lack has to be reduced
significantly. It is untested which influence the
combination of VR equipment has on the overall
performance. The usability of VCS to enhance simulator
training is unproven.
Even after optimization of all VR equipment it seems
unfeasible to completely replace flight simulators by
VCS. In a first step the understanding of the HMI
presented by intrusive, heavy, inconvenient Virtual
Reality gear has to be further investigated. Also, a
complete Virtual Reality simulation theory is missing.
On the first impression large projection screens have less
negative impact on the HMI.
Figure 12: Conventional flight simulator mock-up at
the FMRT
Motion Base
In addition, to the current approach of a fixed based
“Seating Buck”, the entire system can be mounted to a
motion base. This would increase the level of
immersiveness by aircraft motion. To simulate a civil
aircraft the performance of small two-seater (300kg)
motion bases would be sufficient. The increase in
realism and immersion is untested and need to be further
Tracker and System Lag/HMD Limitations
One of the key limitations to VC is today’s tracker
latency. The entire VCS has currently a latency of about
100ms.24 From tests it was deducted that 150ms are
sufficient for orientation tasks within the cockpit.24
Maximal lag for flying a VCS should be well below
Another limitation is the currently used HMD with a
FOV of maximal 56°. As stated above, this HMD
reduces the FOV in a way that disables flying and
orientation tasks within the cockpit. However, this is not
a general concept limitation because HMD vendors
already sell 120° equipment.
HMD resolution is very critical for the usability of VCS.
Research results indicate that with a hires HMD of
1280x1024 pixels and cockpit displays (PFD, ND, and
HUD) of 8 inch rendered at a distance of approximately
85cm (standard pilot-display distance) offers sufficient
resolution.21 Therefore, resolution with modern HMD is
no more limitation to Cockpit-IVS.
Future Work
The Institute for Flight Mechanics and Control will
further investigate the usability of VR-CBT and VCS.
The research will be focused on the human machine
interface generated through virtual simulations. The goal
is to prove the usability (especially of CBT) for real
usage in today’s flight training.
It is assumed that the introduction of large stereoscopic
projection screens into today’s pilot training will be a
natural step. The equipment is already usable for
classroom training applications and CBT.
Available Equipment
At the Institute for Flight Mechanics and Control a
variety of equipment can be used. A front projection
system (Ampro Projector) with a curved screen and four
shutter emitters is installed. For the system seven shutter
glasses (Christal Eye) are available. For IVS a nVision
Datavisor 10× (1280×1024), Kaiser ViSim500,
Polhemus Fastrack, two 18 sensor Cyber-Gloves, and a
triple Pipe Onyx (IR) can be used.
We would like to thank the Fraunhofer Institute for
Computer Graphics (IGD) and their software distributor
VRcom in Darmstadt for supporting us with their Virtual
Reality software package “Virtual Design II”.26
J. Schiefele et al., “Virtual Cockpit Simulation with
Force Feedback for Prototyping and Training”,
Society for Image Generation, Scottsdale, JS1-JS9,
McCarthy et al., “A Virtual Cockpit for a Distributed
Interactive Simulation”, IEEE Computer Graphics &
Applications, pp. 49-54, 1994
M. Stytz et al., “The photo realistic Virtual Cockpit”,
SPIE/SCS Joint 1996, Government & Aerospace
Simulation Conference, pp. 71-76, New Orleans,
M. Stytz et al., “Issues in the development of rapidly
reconfigurable immersive human-operated systems
for distributed virtual environments”, SPIE vol. 3085,
pp. 162-172, Orlando, 1997
J. Schiefele et al., “IFR Cockpit Simulation in a
Distributed Virtual Environment”, SPIE vol. 3067,
Orlando, 1998
J. Schiefele et al., “Simple Force Feedback for Virtual
Cockpit Environments”, SPIE vol. 3067, Orlando,
R.B. Loftin, “Virtual Reality for Aerospace Training”,
VR Systems Magazine, 1996
DoD, “Virtual Environment for Submarine Ship
Handling”, Res. and Eval Program 6.2, Virtual
Environment and Training, DoD, 1997
J. Schiefele et al., “Stereoscopic Projection Screens and
Virtual Cockpit Simulation for Pilot Training”,
Immersive Projection Screen Technology (IPT),
pp.211-222, Stuttgart, 1999
Cruz-Neira et al., “Surround Screen Projection based
VR. The Design and Implementation of the CAVE”,
SIGGRAPH, pp. 135-142, 1993
M. Bolas, “Human Factors in the Design of an
Immersive Display”, IEEE Computer Graphics &
Applications, pp. 55-59, 1994
Kozak, “Transfer of Training from Virtual Reality”,
Ergon 36, 1993
R. Kenyan, M. Afenya, “Training in Virtual and Real
Environments”, Technical Publication, Univ. of
Illinois, Chicago, 1997
B. Knerr et al., “Training in Virtual Reality: Human
Performance, Training Transfer, and Side Effects”,
Society for Image Generation, Scottsdale, 1996
E. Kasper, “Effects of in flight field of view restriction
on rotorcraft pilot head movement”, SPIE, Vol. 3058,
Orlando, 1997
Z. Szoboszlay et al., “Predicting Usable FOV Limits
for Future Rotorcraft Helmet Mounted Displays”,
DERA, UK, 1997
K. Arthur, “Effects of FOV on Task Performance with
Head Mounted Displays”, University of North
Carolina, 1997
S. Rogers et al., “Effects of System lag on head
tracked cursor control”, SPIE, Vol. 3058, Orlando,
B. Witmer, “Measuring Presence in Virtual
Environments: A presence questionnaire”, Presence,
Vol. 7, MIT-Press, pp. 225-240, 1998
M. Snow, “Charting Presence in Virtual Environments
and it’s Effects on the Performance”, Dissertation,
Virginia Polytec Institute and State University, 1996
J. Schiefele et al., “Evaluation of required HMD
resolution and FOV for Virtual Cockpit Simulation”,
SPIE, Vol. 3689, Orlando 1999
K. Stanney et al., “Human Factor Issues in Virtual
Environments”, Presence, Vol. 4, MIT-Press, pp.
327-351, 1998
O. Albert et al., “Das Forschungscockpit der TUDarmstadt, ein Werkzeug zur Untersuchung neuer
Cockpitkonzepte”, DGLR Fachausschußsitzung für
Luft- und Raumfahrt, 1998
A. Eichler, “Untersuchung des Einflusses der Tracking
Totzeit einer virtuellen Simulation auf die
Leistungsfähigkeit der Benutzer im Hinblick auf
künftige VR-Cockpitsimulationen”, Study Thesis,
TU-Darmstadt, 1999
F. Brooks, “What is real about Virtual Reality?”,
IEEE-VR, 1999
P. Astheimer et al., “Virtual Design II: An advanced
VR System for Industrial Applications”, Virtual
Reality World, Stuttgart, pp. 337-363, 1995
UAV Operations using Virtual Environments
Jan B.F. van Erp & Leo van Breda1
TNO Human Factors
Kampweg 5
3769 DE Soesterberg
The Netherlands
In virtual environments (VE), the limited field of view,
the lack of information on viewing direction, and
possible transmission delays may be considered as
potential problems in developing and maintaining a good
sense of situation awareness. Enabling unmanned air
vehicle (UAV) operators to use high quality (proprioceptive) information on (changes in) viewing direction
by introducing a head-slaved camera system with
head-slaved display (HMD) may improve situation
awareness, compared to using a joystick and a fixed
monitor. However, HMDs may degrade comfort and the
dynamics of head movements. Furthermore, time delays
and zoomed-in images induce a non-steady presentation
of the environment, and may impede adequate mapping
of spatial information. This paper reports an exploratory
study into the applicability of a head-slaved camera
system in unmanned platform applications. To overcome
the possible drawbacks of HMDs, we compared an
HMD with a head-slaved dome projection in a simulator
experiment. To overcome the possible drawbacks of
transmission delay, we introduced a new method to
compensate for the spatial distortions. This technique,
called delay-handling, preserves the correct spatial
relation between the viewing direction of the camera and
operator by presenting incoming images in the camera
viewing direction, and not in the actual viewing direction
of the operator.
The experimental results showed that delay-handling is
successful in supporting the perception of correct spatial
relations, i.e., it improves situation awareness. No
differences in task performance were found between the
actual HMD and the dome projection.
In operating a Maritime Unmanned Aerial Vehicle
(MUAV) the flow of information is very poor as
compared to real flying. If a human operator was
physically present at the remote site and performs
manipulations directly, he would receive a variety of
information on the result of his manipulations, such as
visual, auditory, tactile, and force feedback. However,
when the human is physically separated from the task
space, the feedback of the control actions has to be
artificially transmitted back to him.
The man-machine interface determines the extent to
which the operator can sense the remote environment
and consequently control the platform. Thus, the display
and controls in the operator environment should be
designed in such a way that the operator receives task
specific information and sufficient feedback. The images
provided by an on board camera is the main source of
information on the outside world for MUAV operators.
Because of the inherent characteristics of a cameramonitor system, and the restricted data link between the
remote site and the operator, these images are of
degraded quality, which may affect steering and control
performance and the operator’s situation awareness
Image degradation may come in different forms, e.g. a
reduced field of view, a zoomed-in image, decreased
information about the camera viewpoint and viewing
direction, a time delay between the control input and the
consequent feedback, and reduced spatial and temporal
resolution. It is plausible that the degradation of some
aspects of the feedback is more detrimental for operator
performance or the sense of SA than others; some
information may be redundant or of only secondary
value. In order to identify the limitations that may
become critical for the sense of SA when the operator
manually controls MUAV and/or camera movements we
first reflect on the concept of SA. Next, regarding
MUAV operators, the main issues that affect SA will be
discussed. Finally, we establish which principles of
interface design may support the operator in developing
a good sense of SA.
In teleoperation, situation awareness may be defined as
the operator’s ability to perceive, comprehend, and
predict the spatial layout of the elements in the
environment. SA is not a static phenomenon, but is
composed of a variety of changing facts, interpretations
and predictions in the context of task requirements.
Although operator performance undoubtedly depends on
SA, their exact relationship is not clear. Actually, there is
still disagreement among researchers as to just what
constitutes SA. However, the elements of SA are well
known and include such familiar human functions as
perception, information processing, decision-making,
memory, learning, and action-taking, performed within a
dynamic set of environmental circumstances and
SA is important in a wide variety of environments.
Acquiring and maintaining SA becomes increasingly
difficult as the complexity and dynamics of the
environment increase. Under some circumstances, many
decisions are required within a fairly narrow time span,
and task performance requires an up-to-date analysis of
the environment. Because the state of the environment is
For contact with authors: [email protected] and [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
constantly changing (often in complex ways) a major
portion of the operator’s job becomes that of obtaining
and maintaining good SA.
Barfield, Rosenberg and Furness (1995) describe the
main components of situation awareness: spatial, status,
and overall situation awareness. Spatial or navigational
awareness deals with the three-dimensional geometry of
the environment and refers to the operator’s mental
model of the vehicle’s position. What is my position and
how does this relate to the position of other objects? The
state of the platform, e.g. the amount of remaining fuel,
the position of the flaps, is represented in the status
component of awareness. The combination of spatial and
status awareness enables an overall awareness of the
total flight environment.
Endsley (1995) gives a more elaborated model of SA
with three components. Level one in this model refers to
the perception of the elements in the environment and
their relationship to other points of reference (i.e.
internal model). At this level, relevant characteristics
(colour, size, speed and location) and the dynamics of
the objects in the environment are represented. This
aspect is similar to what Barfield et al. (1995) termed
spatial awareness. Level two of SA goes beyond simply
being aware of the elements that are present, and
includes an understanding of the significance of the
elements. Based on level one knowledge, the operator
forms a holistic picture of the environment, comprehending the significance of objects and events. Thus, the
integration of various level one data elements at level
two of SA is crucial for the comprehension of the
situation. Level two of SA can be highly spatial in an
operating context. The relevance of different objects for
the operator’s action planning will depend on their
location and speed. Finally, the ability to project the
future actions of the elements in the environment forms
the third and highest level of SA. For example, in traffic,
knowledge of the status and dynamics, and the
comprehension of the situation, allows a driver to predict
the future actions of other drivers in order to prevent
Another aspect of SA should be mentioned at this point.
Although SA has been defined as a person’s knowledge
of the environment at a given point in time, it is highly
temporal in nature. That is, some aspects, like the
knowledge about the dynamics of the environment and
path prediction, are acquirable only over time.
Smolensky (1993) discusses the work of Stein, who
showed that controller’s eye fixation locations, which
had varied widely in the initial 10 to 15 minutes of an air
traffic simulation, decreased significantly beyond that
point in time. Anecdotally, Stein’s subjects reported that
the initial 10 to 15 minutes of a controllers shift is the
period of time during which he acquires the ‘big picture’,
or, SA. Another temporal aspect of SA relates to the
variations in relevance of elements across time. Some
elements are not of equal importance at all times,
although they should not fall out of consideration
completely. At least some SA on all elements is needed.
SA, therefore, is based on far more than simply the
information perceived about the environment. It is
related to a model of human information processing in
which attention and long-term memory enable
comprehending the meaning of information in an
integrated form. Memory does not only serve to direct
attention effectively, but also serves to interpret the
information that is perceived and to develop accurate
projections of future events.
SA in teleoperation
In teleoperation, an intervening system senses, mediates,
and presents information to the human operator. In this
process, a loss of information can occur, which may be
relevant to all three levels of SA.
At the lowest level, the system may fail to present
certain information that is important for SA in the
assigned task. First, systems may only present information of one modality (e.g. only visual information), based
on technological limitations and the designer’s understanding of what is required. Second, the information
that is presented may lack important cues; e.g. no
stereoscopic depth cues when a single camera is used.
Another major issue in teleoperation is the transmission
speed and capacity. Intervening communication systems
like satellites reduce transmission speed, resulting in
delayed feedback to the operator about his manipulations.
For level two SA, the information displayed by the
system must be integrated, and related to a mental model
to obtain a holistic picture, and to determine which cues
are actually relevant to the established goals. When no
model exists at all, level two SA must be developed in
memory. The absence of sufficient level one SA, the
inability to develop a sufficient mental model or the
inability to properly integrate or comprehend the
meaning of presented data, can lead to inaccurate or
incomplete level two SA. This may be caused by
incomplete or inaccurate presentation of data to the
human operator, or by a mismatch between information
presentation and perceptual, attentional, and working
memory characteristics of the operator.
Finally, level three SA may be lacking or incorrect. Even
if the mental model is sufficient for level two SA, and
the actual situation is clearly understood, it may be
difficult to accurately project future dynamics. Lack of
highly developed mental model and attention and
memory limitations may account for this. Furthermore,
some people are simply not good at mental simulation.
Regarding the control of unmanned platforms, loss of
SA is already present at level one of SA, causing
degraded sense of SA on level two and three as well. The
inability to assess basic properties as position, direction
and speed also hampers the operator in developing a
correct mental model (level two), and in making
adequate predictions about future states of the objects
(level three). Part of the problems are probably related to
the poor information flow specific in MUAV
applications, due to the following reasons:
A small field of view. A limited field of view suppresses
the use of peripheral visual information. The peripheral
area of the retina differs anatomically and functionally
from the foveal area (Schneider, 1969; Trevarthen,
1968), and is used to generate our sense of spatial
orientation (Ungerleider & Mishkin, 1982; Jeannerod,
1997). For example, a human operator’s performance in
a disturbance nulling task with only a central field of
view display can be dramatically improved if the field of
view is expanded to cover the peripheral retina (Kenyon
& Kneller, 1992).
Furthermore, a small field of view requires a higher
degree of integration of spatial information to build up a
representation of the spatial environment. That is, rather
than having a large field of spatial information in which
several objects (and terrain features) are localised, a
smaller field of view affords less spatial information at
any instant, which forces operators to integrate these
small ‘pieces’ of spatial information in time. The results
of a search and replace experiment using an HMD
(Venturino & Kunze, 1989) indicated that the field size
affects one’s ability to acquire spatial information.
However, an important observation in this experiment
was also that once the spatial information has been
mapped into spatial memory, humans could use that
information independently of the size of their ‘window’
to the world. This phenomenon is also found by
Thompson (1983), who asked subjects to walk with
closed eyes to previously viewed targets, and Tyrell et
al. (1993) who asked visually occluded subjects to
position a point of light at the location of a previously
viewed target.
A zoomed-in image. Often, the small field of view is
combined with a zoomed-in camera image. The
zoom-factor of the camera disturbs the normal relation
between rotational speed of the camera and translational
flow in the camera image. For example, Van Erp,
Korteling and Kappé (1995) found that operators largely
overestimate camera rotations when viewing a
zoomed-in camera image.
Few points of reference at sea. The lack of reference
points at sea may hinder the operator in developing a
good model of the position of objects in the remote
environment and their relations.
Low update rate. Update rates lower than 4 Hz limit the
perception of the direction and speed of objects, platform
and camera.
Transmission delays. Transmission delays will mainly
lead to degraded performance of the operator when
manually controlling the camera. Eventually, the
operator will develop a go-and-wait strategy, which will
hamper developing a sense of SA.
Degraded information on (changes in) the viewing
direction. Controlling the viewing direction of the
camera by means of a joystick while the images are
presented on a stationary monitor, withhold the operator
of proprioceptive feedback on viewing direction.
Normally this information is provided by muscle
spindles of neck and eyes, and therefore allows
automatic mapping of visual information on a mental
model. Since the viewing direction can not be directly
deduced from the camera images, it is usually presented
via additional indicators. However, this information
requires the operator to perform some kind of cognitive
processing in order to build a mental model, and it is not
intuitive and therefore slow.
In previous research, it was shown that introducing high
quality synthetic visual information can partly cancel out
problems regarding the zoomed-in camera image, the
lack of reference points, the low update rate and the
transmission delay, which all have an important camera
control component (Van Erp, Kappé & Korteling, 1996).
Field size and information on viewing direction may be
considered as the most important factors related to SA in
unmanned platform applications. Moreover, both factors
probably interact strongly. Although spatial information
can be used effectively regardless of the size of the
‘window’ to the world once it is stored in spatial
memory; the lack of information about the viewing
direction of the camera hinders the building of a mental
representation, and the integration of new information.
Head-slaved camera control
A possibility to convey high quality information about
camera viewing direction is the use of a head-slaved
camera system. When the viewing direction of the
camera is coupled to the viewing direction of the
operator, proprioceptive information is available, which
can be interpreted automatically. Automatic processing
tends to be fast, autonomous, effortless, and unavailable
to conscious awareness in that it can occur without
attention. It is hypothesised that system designs that
support automatic processing of information directly
benefit performance.
Applying a head-slaved camera system also requires a
head coupled image presentation (i.e. a head mounted
display, HMD) instead of a fixed monitor, see Kappé,
Van Erp and Korteling (in press). However, the use of
head-slaved camera control in combination with an
HMD also has two potential drawbacks.
First, HMDs may influence comfort and control
behaviour of the operator. Kotulak and Morse (1995)
discuss a survey of 58 aviators by Behar, who found that
51% had visual discomfort, 35% had headache, and 21%
had blurred vision. These symptoms could have a
common origin: eye-head co-ordination could be
affected by HMD characteristics, and smaller field sizes
place heavy demands on head movements, since subjects
must move their heads to sample the environment rather
than using the more effortless joystick control. A study
by Gauthier, Martin and Stark (1986) suggests that the
greater head inertia associated with HMDs may induce a
decrease in the amplitude-velocity relationship of head
movements, i.e. slowing of head movement and small
changes in head amplitude. Further, eye movements may
change secondary to these changes in head velocity. Eye
movement maximum amplitude and velocity increase
with increasing inertia. Gauthier et al. (1986) studied
these effects of added head inertia and discuss that
oscillopsia (continuous displacement or instability of the
visual world) was prominent and consistent in perceptual
reports of their subjects.
Second, transmission delays may distort the correct
relation between the external environment and the
perceived visual array. Because the images on an HMD
are presented in the actual viewing direction of the
operator, a transmission delay introduces a discrepancy
between the viewing direction of the camera at the
moment the images were recorded at the remote site, and
the viewing direction of the operator at the moment the
images are presented. This results in the operator
perceiving the world as unstable when he moves his
head. For example, when the operator has a steady image
of an object, moving his head will ‘drag’ it across the
environment during the transmission delay. Therefore,
transmission delays will probably impede adequate
spatial mapping of the visual information.
A possibility to reduce the first drawback (comfort) is to
project the images in a moving window projected onto a
dome, instead of on an HMD. A possibility to prevent
the second drawback (delay) is to display the images in
the viewing direction of the camera at the moment of
recording, and not in the actual viewing direction of the
operator (called delay-handling throughout the paper).
This results in an image location which corresponds with
the image content, and follows the actual viewing
direction of the operator with a delay, instead of an
image location which corresponds with the actual
viewing direction, but not with the image content.
In case the field of view on the environment has the
same size as the field of presentation (which is defined
as the size of the display on which the view on the
environment can be presented, e.g. the size of the dome),
the principle of delay-handling will lead to image loss on
the side contra-laterally to the direction of motion.
Therefore, the field of presentation must preferably have
spare space to overcome this loss. In this respect, domes
are preferable. The size of this spare space and the
transmission delay determine the maximum speed the
camera can rotate without image loss.
The present exploratory experiment was used to
investigate the possibilities of head-slaved camera
control for unmanned platforms. To elaborate on the
possible drawbacks mentioned above, we used two
presentation modes: a head-mounted display, and a
moving window on a dome; and we introduced different
transmission delays and tested the principle of delayhandling. To test the effect on the operator’s sense of
SA, we developed an experimental task, which included
level one, two and three of SA as defined by Endsley
Seven college-educated, right-handed male subjects
(age: 20 to 27 years) participated in the experiments. All
subjects had normal or corrected to normal vision, were
paid for their participation, and had no experience with
similar operator tasks.
All images were generated by a three-channel Evans and
Sutherland ESIG 2000 image generator (30 Hz update
rate). The images were presented via a head mounted
display (N-Vision, 41.5° × 34.5°, 800×600 pixels H×V),
or via a projection screen (a Seos PRODAS HiView S600 projection system, consisting of a spherical dome
and three video projectors; radius 2.9 m, 150° × 42°,
2400×600 pixels H×V). The subject’s head was
positioned in the centre of the dome. Head orientation
(horizontal and vertical) was registered by a Polhemus
Fastrack head-tracker (resolution 0.15°, 30 Hz), with the
sensor coil either mounted on the HMD or on a
lightweight plastic helmet (weight < 0.1 kg). Minimum
delay between head-tracking and displaying was about
60 ms. Head tracker data was used as input for the
mathematical model (ran with 30 Hz on a 486-based
PC), which calculated the motions of the simulated
(head-slaved) camera and the objects in the database.
The mathematical model also simulated the transmission
delay between the camera and the operator, by using a
pipeline with a size of 30 times the transmission delay
(s). A second 486-based PC was used for scenario
generation and data storage (30 Hz sampling frequency).
The presented view on the environment (window) had a
size of 13.3° × 10.0°, and could be projected in the
actual viewing direction, or in the viewing direction of
the camera for which the images were generated. Note
that with a transmission delay this resulted in a delayed
image content and a delayed image location,
The subject was seated in a chair with a right armrest, on
which a spring-loaded joystick was mounted. A response
button was mounted on top of the joystick (Figure 1).
Figure 1: An overview of the TNO MUAVsimulator facility
The camera-platform remained at a fixed position and
orientation throughout the experiment, altitude of 500
feet. The virtual environment depicted by the camera
image consisted of a textured sea, twelve ships, and six
square so called oil-rigs. The oil-rigs were arranged
along imaginary gridlines, such that they enclosed an
area defined by parallel and perpendicular lines between
the rigs (Figure 2). This area was defined as forbidden
for target ships. The distance between the platforms was
1000–2000 feet.
Six moving ships of equal type were defined as targets;
the other six ships were distracters, were of a different,
smaller type and had to be neglected. The targets moved
at 45 feet/s along a winding route that was unknown to
the subject, and had a maximum turn rate of 3°/s. The
ships headed for an end position within the forbidden
Overall task instruction was to give a signal when a
target ship entered the forbidden area, which actually
consists of the following parts:
• determine the form and location of the forbidden area
by detecting the position of the oil-rigs, and drawing
imaginary borders,
• detect and monitor the position and track of the target
• give a signal whenever a target ship enters the
forbidden area.
This experimental task was designed to implement the
different levels of SA as introduced by Endsley (1995).
Level one refers to the position of the oil-rigs and the
ships, their attributes, and their spatial relations in the
environment. Level two refers to comprehending the
significance of the different elements: which ships are
targets, and which targets are heading for the forbidden
area. Level three refers to the need to predict the future
position of targets, e.g. assess which of the targets will
reach the forbidden area first.
Figure 2: Illustration of a possible alignment
of the six oil-rigs
At the time that one of the targets actually crossed a
border (marked target position in Figure 2), subjects had
to keep the ship’s stern in the centre of the camera image
and push the button on the joystick. The target ship
disappeared when it was held within 2° of the centre of
the image at the time of the response. When the subject
did not give a response, the target ship automatically
disappeared when it reached a predefined end position
within the forbidden area. Whenever a target ship
disappeared, a new target ship was placed at a different
position in the environment to keep the number of ships
to be monitored constant during a run. A run was
completed when six target ships had disappeared.
During the run, performance was recorded in order to
calculate objective performance measures afterwards.
Furthermore, after the completion of a session, subjects
were given a post-test to ascertain that they had
memorised the alignment of the oil-rigs, i.e. if they
developed a mental model of the world during a run. A
forced-choice procedure was used, in which the subjects
had to choose the actual alignment of the oil-rigs out of
the six drawings (bird’s eye view) of possible
Independent variables
Three independent variables were manipulated in a full
factorial within subjects design: presentation mode
(HMD and dome projection), delay-handling (absent,
present), and transmission delay (0, 0.5, 1.0, 2.0, and 4.0
s), resulting in twenty conditions.
Dependent variables
The following performance measures were used:
• Time to locate the oil rigs (s). The measure was
defined as the time it took a subject to locate all six
oil-rigs, i.e. the time until the camera had been
pointed at all of the six platforms at least once.
• Time to border crossing (s). The measure “time to
border crossing” for each target was calculated as the
time that a target was away from the border to be
crossed at the moment of the response of the
participant. Time to border crossing was taken over
all targets signalled by the participant (between 1 and
6). This measure reflects the accuracy of the subjects
in estimating the position, course and speed of the
target ship relative to the oil-rigs, i.e. their accuracy
in the perception and prediction of spatial relations.
• SD heading (°). The measure “SD heading” is
defined as the standard deviation of the heading of
the viewing direction during a single run, and is a
measure of viewing behaviour.
• SD pitch (°). The measure “SD pitch” is defined as
the standard deviation of the pitch of the viewing
direction during a single run, and is a measure of
viewing behaviour.
• Multiple choice on platform orientation. This
measure was calculated as the number of correct
choices of the alignment of the six oil-rigs (summed
over the levels of transmission delay).
Statistical design
The experiment was completed in sessions consisting of
the five transmission delay levels for a combination of
presentation mode and delay-handling. These blocks of
five runs were, although not completely, order-balanced
across the subjects. Within each block, the order of
transmission delay was randomised. For each subject,
the twenty scenarios were randomly assigned to the
conditions, with the restriction that each combination of
condition and scenario occurred only once throughout
the experiment.
Each dependent variable was checked for outliers (scores
that deviated by more than 3 SD from the overall mean)
and sphericity. Incidentally, a large score on the time to
border crossing was found. Target ships could approach
a border until they were at a short distance from it, but
because of the winding route they moved along, not
actually cross the border. Therefore, values greater than
20 s were removed from the analysis. No other outliers
were found.
Results of the performance measures “time to locate the
oil-rigs”, “time to border crossing”, “SD heading”, and
“SD pitch” were analysed by a within-subjects design
with three factors: presentation mode (2) × delayhandling (2) × transmission delay (5) with the statistical
package STATISTICA 5.0. Significant results were further
analysed by a post-hoc Tukey test. Results of the
multiple choice question (only one observation per
session of five runs) were analysed by a within-subjects
design with two factors: presentation mode (2) × delayhandling (2).
First, subjects received a brief written explanation about
the general nature and procedures of the experiment. The
instructor then showed the projection dome, chair, the
plastic helmet and the HMD, and explained the purpose
and task in more detail. The subjects came in pairs: one
subject performed a session of five runs, preceded by a
practice run, while the other subject rested. The practice
run was with no transmission delay, was not registered,
and performed with a scenario not used during the
experiment. After a session the subject was instructed to
perform the multiple-choice task in a room near the
room in which the dome was situated.
Presentation mode. On the basis of experimental
observations (see Gauthier et al., 1986) and the smaller
field of presentation, a disadvantage of the HMD was
expected. However, none of the performance measures
showed a significant effect of presentation mode.
Delay-handling. Two dependent variables showed a
main positive effect of delay-handling. Time to border
crossing showed a performance increase of 15% with the
presence of delay-handling [means 5.8 s and 4.9 s,
F(1,6)=23,91, p<.01]. The mean number of correct
answers on the multiple choice task increases with 40%
(means 2.4 and 3.4) with delay-handling present, F(1,6)=
21.00, p<.01. Delay-handling showed no significant
Transmission delay. Three performance measures
showed a main effect of transmission delay. The time
needed to locate the oil-rigs [F(4,24)=20.72, p<.01], the
time to border crossing [F(4,24)=7.75, p<.01], and SD
pitch [F(4,24)=6.39, p< .01]. All effects showed
performance decline with increasing transmission delay.
The post hoc tests indicated that performance on the
former two variables was degraded for delays larger than
0.5 s, on the latter only for a delay of 4 s.
The present study concentrates on the concept of
situation awareness (SA) in relation to camera control of
unmanned platforms using virtual environment (VE)
techniques. In the introduction, it was hypothesised that
inherent characteristics of the man-machine interface,
like the limited field of view and the time delay between
image recording at the remote site and image
presentation, may hamper the operator in developing a
good sense of SA. Providing the operator with high
quality information on (changes in) viewing direction by
introducing a head-slaved camera system with headslaved display may support the operator and improve
SA. However, literature also shows that such systems
may degrade other aspects, e.g. comfort, control strategy,
and the spatial relation between viewing direction of
camera and operator as a result of transmission delays.
The present experiment focussed on the applicability of
head-slaved camera systems in MUAV applications. To
overcome possible drawbacks of HMDs, we compared a
head mounted display with a head slaved dome
projection and to overcome the possible drawbacks of
transmission delay. We introduced a mechanism of
delay-handling which preserves the correct spatial
relation between viewing direction of the camera and the
operator by presenting incoming images in the camera
viewing direction, and not in the actual viewing direction
of the operator. A new experimental task was introduced
to include the different levels of SA as discerned by
Endsley (1995).
The results show no significant effect of presentation
mode. Although mean values on SD heading and SD
pitch showed higher values with dome projection over
the HMD, the effects did not reach significance (p=.16
and p=.10, respectively).
The results indicated a positive main effect of the
principle of delay-handling (depicting the delayed
images in the camera, not in the actual head direction).
Both the results of the time to border crossing and the
multiple choice task show performance improvement
when delay-handling is applied. Time to locate all oilrigs and control behaviour did not differ with delayhandling absent or present. This indicates that delayhandling is especially useful for developing higher levels
of SA, i.e. in determining the exact spatial relation
between the oil-rigs and the imaginary borders and the
The main effect of transmission delay shows that this
variable both degrades the development of the sense of
SA at all levels, and the control behaviour of the
Because delay-handling results in a window moving
with a delay, the available field of presentation must be
larger than the field of view. This may be a disadvantage
for the HMD mode of presentation, because HMDs have
a restricted field of presentation. However, the lack of an
interaction presentation mode × delay-handling shows
that the field of presentation of the presently used HMD
was sufficient.
We also expected an interaction between delay-handling
and transmission delay. Increasing transmission delays
will disturb the spatial relations more for the same
control signals, and was therefore expected to increase
the positive effects of delay-handling. Even a third order
interaction (presentation mode × delay-handling ×
transmission delay) might have been present.
Transmission delays were supposed to be compensated
by presenting the images in the spatially correct viewing
direction. This method requires a field of presentation,
which is larger than the size of the camera images, and
must be increased with increasing time delays. Since the
field of presentation of the HMD is restricted, an
additional advantage of the dome projection was
expected for larger transmission delays. However, none
of the interactions was found.
It is recommended to perform human factors research
aimed at further improving operator performance by
optimising interface design. Areas of interest include the
• Directly compare the effects of joystick versus
head-coupled camera control on the sense of SA and
camera control performance.
• Investigate the effects of a zoomed-in camera image
on head-coupled camera control. The zoomed-in
camera image disturbs the relation between head
rotations and translational flow in the image, which
may be confusing and uncomfortable to the operator.
• Further explore the applicability of the method of
delay-handling in, for example, situations in which
the camera translates through the remote environment, or in which the camera image is zoomed-in.
• Investigate the relation between man-machine
interface characteristics and the different levels of
SA, and develop specific operator support. An
example is adding high quality visual information to
the camera image to provide the visual information
that is lost in some situations, e.g. as a consequence
of the low update rate of the image (by presenting
visual motion information), a zoomed-in image (by
presenting correct translational flow for camera
rotations), and transmission delays (by introducing a
predictive display).
Barfield, W., Rosenberg, C. & Furness III, T.A. (1995).
Situation awareness as a function of frame of
reference, computer-graphics eyepoint elevation, and
geometric field of view. International Journal of
Aviation Psychology, 5 (3), 233-256.
Endsley, M.R. (1995). Toward a theory of situation
awareness in dynamic systems. Human Factors,
37 (1), 32-64.
Gauthier, G.M., Martin, B.J. & Stark, L.W. (1986).
Adapted head- and eye-movement responses to
Environmental Medicine, 57, 336-342.
Jeannerod, M. (1997). The cognitive neuroscience of
action. Cambridge, UK: Blackwell Publishers Inc.
Kappé, B., Van Erp, J.B.F. & Korteling, J.E. (in press).
Effects of head-slaved and peripheral displays on
lane-keeping performance and spatial orientation.
Accepted by Human Factors.
Kenyon, R.V. & Kneller, E.W. (1992). Human
performance and field of view. 1992 SID
International Symposium Digest of Technical Papers
(pp. 290-293), Playa del Rey, CA: Society for
Information Display.
Kotulak, J.C. & Morse, S.E. (1995). Oculomotor
responses with aviator helmet-mounted displays and
their relation to in-flight symptoms. Human Factors,
37 (4), 699-710.
Schneider, G.E. (1969). Two visual systems. Science,
163, 895-902.
Smolensky, M.W. (1993). Toward the physiological
measurement of situation awareness: the case for eye
movement measurements. Proceedings of the Human
Factors and Ergonomics Society 37th Annual
Meeting (pp. 41-42).
Trevarthen, C.B. (1968). Two mechanisms of vision in
primates. Psychologische Forschung, 31, 299-337.
Thompson, J.A. (1983). Is continuous visual monitoring
necessary in visually guided locomotion? Journal of
Experimental Psychology: Human Perception and
Performance, 9, 427-443.
Tyrell, R.A., Rudolph, K.K., Eggers, B.G. & Leibowitz,
H.W. (1993). Evidence for persistence of visual
guidance information. Perception and Psychophysics, 54 (4), 431-438.
Ungerleider, L. & Mishkin, M. (1982). Two cortical
visual systems. In D.J. Ingle, M.A. Goodale &
R.J.W. Mansfield (Eds.), Analysis of visual behavior
(pp. 549-586). Cambridge, MA: MIT Press.
Van Erp, J.B.F., Kappé, B. & Korteling, J.E. (1996).
Visual support in target search from a moving
unmanned aerial vehicle (Report TM-96-A002).
Soesterberg, The Netherlands: TNO Human Factors
Research Institute.
Van Erp, J.B.F., Korteling, J.E. & Kappé, B. (1995).
Visual support in camera control from a moving
unmanned aerial vehicle (Report TNO-TM 1995 A26). Soesterberg, The Netherlands: TNO Human
Factors Research Institute.
Venturino, M. & Kunze, R.J. (1989). Spatial awareness
with a helmet-mounted display. Proceedings of the
Human Factors Society 33rd Annual Meeting (pp.
The Dangerous Virtual Building,
an Example of the Use of Virtual Reality
for Training in Safety Procedures
Miguel Lozano1, Marcos Fernandez, Joaquín Casillas, Javier Fernández & Cristina Romero
Institute of Robotics, University of Valencia
Polígono de la Coma s/n
MailBox: 2085 – 46071 — Valencia, Spain
There is an ancient proverb that says “Tell me and I will
forget. Show me and I may remember. Involve me, and I
will learn”. This has been the main principle behind the
big raising of immersive technologies in the field of
training and education.
Here we explain our experience in using this kind of
technology in the area of work risk and incident
prevention. The high accident rate suffered by the
construction sector has been one of the reasons that have
moved us to develop the system that this article
describes. The objective of the system is the training of
the operator in safety procedures on the job. For this
reason a VR system has been created that on the one
hand reproduces a similar environment to the one
experienced by the operator in real life, and on the other
hand it provides for a number of operations to be
completed. These operations which are very usual for the
worker in real life imply a risk that must be understood
by the worker, e.g. walking around the construction
trenches carrying some type of load could cause a
loosening of the ground resulting in death. For the
complete training of the worker, the virtual environment
contains the three fundamental phases of the
construction of a building. Besides all of the general
tools of the job may or may not have a safety
component. So the number of dangerous operations that
the system provides for and monitors are encountered in
real life (working on a scaffolding, in trenches, on roofs,
on the various floors, crashes, falls, overloads, etc.) By
means of training and learning about the risks involved
in the operations (from the most simple) you will obtain
the best preparation in the sector, reducing therefore the
rate mentioned above.
Using the system the worker is really involved in the
task, and is able to understand the real risk that the task
carries out, because he is in front of a screen that shows
the object in its actual size and he has to make the proper
decision. The system do not intent to train him or her in
the skills of the task but in the safety way to proceed in
its development.
This is a case that can be port to other military or civil
areas where not only are important the skills but also is
necessary to observe a methodology that ensures a safety
We point out also in this paper how is possible using
low-cost equipment to produce a good degree of
immersive system. This is an important point in order to
extend the use of those systems to such a sector or when
the number of subjects to be involved in the training
process make necessary to use a elevate number of
simulation systems.
Introduction and capabilities of the system
Virtual environments are of major interest to computer
graphics researchers; this is due, in part, to their ability
to immerse the user in a computer-generated alternate
reality in which we can easily recreate scenarios which
are too dangerous, difficult or expensive to play in real
life (Bukowski, 1997).
In this paper, we present an approach to this kind of
system, the dangerous virtual building system (DVB) is
an application of visual simulation, oriented to worker’s
education in the field of civil construction (Alkoc, 1993).
One of the main goals to achieve by the system is the
training of the operator in safety procedures on the job,
and the second is to give us a measurement or an
evaluation of these safety tasks.
The DVB has been designed to simulate a finite number
of risky procedures that could occur in a real work
environment. Demonstrating these procedures and
evaluating the risks that each one implies, the workers
can learn or review the safety routines that are often
forbidden. Later the application will provide a
measurement of the learning of capabilities of each
worker in these safety procedures.
The main user in the DVB system, will be a student of a
course in safety tasks, who works in construction field.
Generally, the student doesn’t know how to use most of
the common tools utilised in computers such as a mouse
or a keyboard. So in order not to hinder the learning
process an instructor is needed to advise in the
management of the system and to explain the goals to be
achieved during the simulation.
A prototype of this system, based on SGI workstation,
was developed and tested (Lozano, 1999) and currently
we are developing a new open architecture focussed on
the DVB training system.
The system under development consists of a centralised
instructor control sever plus twelve simulation nodes
(based on PC architecture). Each one of the subjects is
immerse in his/her own simulation process and the
For contact with the author: [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
instructor can control the development of each exercise.
The system has been based on distributed standard
architecture (CORBA) and for the output of the
simulation three possibilities has been offer: Head
Mounted Display, flat monitor or 2×2 meters screen. The
core of simulation graphics has been developed using
low-cost graphics platforms with LINUX operative
system and Performer libraries. The whole system offers
enough graphic quality for both purposes, the training
phase and the instructor node.
This instructor, who knows the capabilities of the
system, will control the simulation, ordering some kinds
of tasks for each student and enabling or disabling the
proper conditions for that task. Later he will check the
results given by the DVB in the learning process and will
be able give us the level achieved for each student.
The way to show these procedures has been based on a
VR application, so that, we can reproduce the familiar
environment of the worker and he must interact with the
system in order to achieve his goals. The input device
used for the subject interact with the system has been
and standard joystick.
The main capabilities of the system are:
• Simulate a virtual building environment managed
from a subjective point of view (the camera) and
controlled from the Joystick position (Cabral, 1996).
• Simulate and control the worker-environment
interaction: The system simulates a number of risk
situations (defined below), and must control the
reactions and consequences.
• A number of elements (objects, tools, ... etc.) must be
created and the interaction must be controlled by one
specific module called the worker’s bag (Santonja,
• The system takes into account the legislation
regarding safety rules, and informs the worker if his
behaviour doesn’t comply with these rules.
In the rest of this article we will define each one of these
capabilities, exploring in this way the contents of the
The Object Interface
The interaction with the objects commonly used in the
building area is a very important element of the
application. The application must allow the worker to be
able to select objects and carry out an action with them.
For this purpose, an object interface has been designed
similar to the interfaces of adventure games. Whenever
we wish we could show at the bottom of the screen an
area composed of the next elements:
• The upper row is used to show the objects that the
worker wears at a given moment, such as a helmet,
gloves, etc.
• The middle row shows the objects that the worker
carries in his hand, his pockets, or work belt, such a
large hoe, a shovel, etc.
• The right area shows the objects that the worker
carries in the wheelbarrow, if the worker finds it
necessary to collect them. The wheelbarrow will then
appear in the middle row as a transported object.
• The lower row shows the different actions that can be
carried out with the currently selected object
(Figure 1).
Figure 1: The Object Interface
Pressing the right spaceball button shows the object
interface area. Once opened the object interface, can be
in one of two possible states:
• Object selection: it allows the free choice between
the objects that the worker wears, transport, or carries
in the wheelbarrow.
• Action selection: it allows free choice between the
possible actions for the currently selected object.
The verification of the action with the selected object is
one of the most important steps in the diagram shown
above. In order to carry out this verification, it is
necessary is to take into account the weight and
maximum volume the worker can support. Moreover,
there is also a necessity to verify the number of ‘spots’
that remain free for an object to be carried. These ‘spots’
are the pockets, the belt, the worker hands, etc.
Once an object and the action the worker wants to
perform on that object has been chosen, it will be
checked to see if such an action is feasible with that
object, and if it may carried out. The selection process
can be summarised in Figure 2.
In order to help to the verification of the actions, a mask
is assigned to each object with the possible action that
can be performed with that object. As actions are
performed over the objects, the mask will be modified to
update the future possible actions over the object. In this
manner the execution of an action over an object can be
completely controlled.
Normal State
Left button
Channel 0
Channel 1
Object selection
Left button
Action selection
Error Message
3D visual database
Perform action
Object interface area
Figure 2: Object interface state diagram
Figure 3: Channel Structure
In Figure 3 we can see an image of the application object
interface with some of the building area objects loaded.
This figure shows us the objects that the worker wears in
the first row of the object interface: the worker wears
work suit, work boots, a tool belt, a safety harness, a
helmet, and work gloves. In the row showing the
transported objects, the worker carries an anti-gas mask
and the wheelbarrow with the objects shown to the right
of the interface (a large hoe, a shovel, a brick and a
cement sack). The current selected object is marked with
a blue square, as we can see in the helmet icon. The four
actions possible over the selected object are shown in the
lower row: cancel, transport, leave, and put into the
The interface object area is a dedicated channel different
to the visual database scene channel, with its own visual
database which is composed of small plane (two
dimensional) objects with its texture applied (Rohlf,
1994). This structure is shown in Figure 3.
Figure 4: Arrangement of working areas
Danger situations
The main purpose of the application is training of the
construction workers in issues concerning safety
conditions (Lozano, 1998; Bukowski, 1997). This
embraces knowing the essential equipment for each kind
of task, and the right way of doing that task.
The stage has been divided in five areas and two access
points, one for the workers and the other for the vehicles.
The areas are arranged as shown in Figure 4.
The five working areas are as follows:
• Equipment barracks: There are seven barracks with
different equipment and clothes that the workers can
use. There are elements, which are suitable for
general working in the building area, such as boots
and gloves. However, there is more than one element
for each type of clothes. For example, there are rain
and anti-slide boots, and the worker must choose the
right equipment for the task he is going to do
according to the weather conditions.
• Storage space: This is an area where the building
elements are stored. The workers should leave things
like cement sacks, wheelbarrows, and general
working tools in this area.
• The ditch: There is a ditch with two propped up walls
and the other two unpropped. The worker can go
down to the ditch trough a ladder.
• One floor building: It is a small area with a building
that has one floor and is under construction. There is
a scaffold for the worker to use when working on the
facade and a ladder to go up to the first floor.
• Four floor building: It is the biggest building in the
building area, and is also under construction.
In addition, there are a couple of access points. The
workers must use the people access point; otherwise they
could suffer serious injury.
In the next images we can see situations corresponding
to the areas mentioned above:
altitude, weight, etc. Examples of these situations
− Jumping: if the worker jumps from a surface (a
scaffold), he could get injured if the altitude is
moderate (up to four meters) or even die if he
jumps from a higher site.
− If the worker accumulate too many objects or
materials in concrete areas, there is a risk of
terrain collapse in those areas. So the worker
must wear the necessary safety equipment in case
such a situation occurs.
− Collision with dangerous objects: if the worker
drops a dangerous object (such a large hoe), it
could cause injury to other workers. This will be
advised by a warning message.
• Situations that depend on the place or area which the
worker is working in: there are a lot of these
situations, so we will describe a few organised by
− Vehicle access area: if the worker goes into the
building area through the vehicle entrance, he
could be run over by a truck (Bayarri, 1996).
− Storage space: here, the worker must walk
carefully because there are dangerous objects in
the area, so he should not stay on this area for a
long time.
− Ditch: in this area, there is a risk of terrain
collapse if the worker walks in the zone that is not
yet secure. If the worker wears a safety belt, he
could be rescued in case of terrain collapse.
− One floor building: here there is a scaffold that
does not comply with to building regulations, so
there is a risk of falling. The worker must be
attached to the scaffold through the safety belt in
order to prevent an accident.
− Four floor building: in this area there are several
different situations.
There is a trench surrounding the building with
duckboards for the workers to go into. If the worker
jumps the trench, he could fall and be seriously injured.
So he must use the duckboards to access the building.
Figure 5: Working areas of the application
Training can be broadly defined as the learning or
acquisition of skills in order to enhance performance at a
given task or job (Burinston, 1995). In order to train, the
building workers have thirty-six different dangerous
situations that have been defined throughout the building
areas. There are several types of situations:
• Situations that do not depend on the area which the
worker is working in: this situations depends on time,
Walking under the building without a helmet is
dangerous, because some object dropped from a higher
floor may hit the worker.
There are gaps for the lifts that are surrounded by
wooden fences, but in some cases the fence is not
complete. In this case, the worker must pick up a wood
board and complete the fence to prevent a possible
accident (such as falling through the gap in the fence).
There are provisional ramps for the workers to go up to
higher floors. Some of these ramps lack bricks, so the
worker may slide and fall. In this case, the worker must
use the correct ramps, or wear suitable boots.
The following images show some of the above
Figure 6: Risk of sudden fall of a brick
from the roof
In the left-hand picture the worker is walking under a
protective cover, so there is not a risk of the object
falling. In the right hand picture, a brick is falling on to
the worker. As there is no protective cover, the brick will
hit the worker, and if he is not wearing a helmet he will
be badly injured.
Concerning to the first aspect pointed out previously, the
basic analysis of the queries performed to the workers
concluded that they were very excited with the use of
this technology. At the first step users saw the system as
a new experience and they were more active than in
other teaching media, like video.
However at this point some problems were detected
concerning to the navigation in the three dimensional
environment and its location aspects. A couple of motion
sickness cases were detected.
Focussing on the second point, a good learning
transference was detected taking into account the written
test performed after the exercises.
Nevertheless is important to notice that these tests were
only initial evaluations that they only try to evaluate the
convenience of starting the process of implementing
actually a profitable system.
One of the problems detected in the first prototype was
the high cost of the system. The current system has
almost the same capabilities and a cost twenty times
Figure 7: Risk of falling down
through the lift gap
The left-hand picture shows that the fence surrounding
the lift gap lacks an element. The worker must secure the
gap zone by placing the wooden board on the floor.
In summary, the first action that the worker should
undertake is to go to the equipment barracks and put on
the correct clothes, depending on the area that he is
going to work in. Then he will be prepared to begin his
task, and go to the corresponding area. When finished,
he must leave the elements, which he has worked on, in
the storage area, and the clothes in the equipment
barracks. In order to leave the building area; the worker
must use the people access. The simulation restarts
whenever the worker is killed by an accident, and the
worker must go again to the barracks. In this manner the
worker will learn the correct elements he must use in the
corresponding areas.
Evaluation and future works
The previous prototype version of this system, which
was running on a Silicon Graphics workstation, was
tested with more than forty workers.
We can summarise the main objectives of those tests in
two aspects: firstly, evaluate the degree of acceptance of
the system in a group of people that is not familiar with
this technology, and secondly evaluate transference of
learning when it was produced.
We have been also working in solving the problems of
navigation, basically making the use of the joystick more
intuitive and limiting in some ways the freedom of
movements that some times produced a problem of
The current system as we have explained before is being
developed with twelve simultaneous training nodes. The
idea is to make this system portable in order to be
installed into a forty-feed truck where the training
process will be developed directly at the building area.
By this way we will reduce the learning cost and will
increase the productivity, taking into account the number
of persons able to run the system.
The process of real evaluation of the system will start by
the end of the year when the whole system will be ready
to go to the real building areas.
Alkoc, C., “A computer Based Simulation of Concreting
Operations”. Development in Civil & Construction
Eng. Computing, 8. Publi. Civil Comp Press (UK).
Bayarri, S., Fernández, M. & Pérez, M., “Virtual Reality
for Driving Simulation”. Communications of ACM
(May 1996).
Bukowski, R. & Séquin, C., “Interactive Simulation of
Fire in Building Environments”, SIGGRAPH 94.
Burinston, A. & Taylor, P., “Generic Training analysis
methodology”, ITEC 95.
Cabral, B., “Graphics Techniques for walkthrough
Applications”, SIGGRAPH 96, Course Notes.
Lozano, M., Casillas, J. & Fernández, J., “A Training
System for Crane and Material Transport System
Operation in Building Areas”, ITEC 98. Laussane.
April 1998.
Lozano, M. Et al., “The dangerous virtual building, a VR
training system for construction workers”, ITEC99.
The Hague. April 1999
Santonja, J., Bayarri, S., Pérez, M. & Martinez, F.,
“Simplificación de mallas triangulares para su
visualización en tiempo real”, CEIG’96.
Rohlf, J. & Helman, J., “IRIS Performer: A High
Performance Multiprocessing Toolkit for Real-Time
3Dgraphics”, SIGGRAPH 94.
Visualisation of Geographic Data in
Virtual Environments
Thomas Alexander1
Research Establishment for Applied Science (FGAN)
Research Institute for Communication, Information Processing and Ergonomics
Dept. Ergonomics and Information Systems
Neuenahrer Straße 20
53343 Wachtberg-Werthhoven
Virtual Environments (VE) are characterised as a
computer-based generation of scenes of abstract or
realistic environments, which can be perceived consistently. The use of VE is very promising in several
areas, especially when visualisation of complex data in a
realistic and clearly understandable way is needed. For
military applications VE technology has potential in the
area of research and development, training, mission
support and mission rehearsal.
A further application is use in Command & Control (C²)systems due to upcoming demands in this area.
In future battlespace scenarios huge amounts of highly
dynamic information will be available due to the
technical development of sensor, communication and
information systems. Therefore advanced techniques for
supporting the military commander and displaying
complex tactical situation data in a clearly understandable way have to be developed and evaluated.
In this connection a concept for pre-processing and
visualising incoming tactical data and three-dimensional
geographical data has been developed. This “Electronic
Sandtable (ELSA)”, as described in this paper, uses VE
technology. The ELSA facilitates a plastic stereoscopic
visualisation of three-dimensional data. It has been
designed to simulate a sandtable as commonly used by
the Armed Forces for tactical education and training.
Therefore the visualisation of digital geographic data
(elevation (DTED) and feature (DFAD) data) is
This paper focuses on the stereoscopic visualisation of
geographic data. Therefore different stereoscopic projection models are described and compared to each other.
For the Electronic Sandtable a model with a window
projection was chosen and implemented. The baseline
concept and first results of this implementation are
referred to in this paper.
1. Introduction
Huge amounts of highly dynamic information will be
available in future battlespace scenarios because of the
technical development of sensor, communication and
information systems. Broad data acquisition, transfer,
and presentation will enable the military commander to
get a variety of diverse information about the battlefield
scenario. The accomplished information dominance is
more and more considered to be essential for a battlespace dominance.
But the massive quantity of information is also
hazardous. Especially in time-critical situations when
tactical decision making under stress is required, relevant
information may be overseen and a wrong mental model
of the tactical situation might be gained.
That overload is likely to be reduced by using new
technologies for data pre-processing and data presentation. Because data presentation is of critical importance
in the whole process of decision making, ergonomic
research is required to analyse the whole process of data
presentation, considering new displays and interaction
Especially using Virtual Environment (VE)-technology
is promising. It was found to have high potential in
presenting and interacting with complex amounts of
data. Therefore VE will increase the clearness and intelligibility of a complex tactical situation. The situation
scenario is not perceived as a complex of abstract
information but as a pseudo-realistic model landscape.
This is intensified by an intuitive, easy to learn
interaction with the included objects.
2. Command and Control (C²) Systems
Command and Control (C²) Systems have been designed
to support the military staff in co-ordinating defensive,
peace keeping and enforcing missions, exercises,
humanitarian aid and ministerial expertise. For this
reason diverse sensor information data and information
data of knowledge databanks are joined in these systems.
A part of C² is the output and presentation of tactical
information. It has large influence on the general
decision making process, because the commander’s
mental model of the battlespace situation is based upon
the information perceived.
The SHOR-model (Stimulus, Hypothesis, Option,
Response) of decision making introduced by Wohl
(1981) proves this. It describes the process of decision
making from data gathering to executing responses. The
available and pre-processed information of a C²-System
is displayed by the Tactical Situation Display (TSD).
2.1 Tactical Situation Displays (TSD) today
The basic function of TSD is to display the current
situation of own and reconnoitred enemy troops and
For contact with author: [email protected]; tel. +49 228 9435 480, fax +49 228 9435 508
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
facilities in the operation area to the commander of a
military unit.
Moreover the TSD is used for tactical planning of
intended future operations. Quantity and quality of
situation data are essential for an adequate operation
planning (Grandt et al., 1997).
Today’s conventional TSDs might not be able to meet
the demands of future battlespace scenarios and have to
be extended by new, innovative technology. The strike
forces today uses two basically different types of TSDs.
The first one, shown in Figure 1, is a command post in
the field. The TSD used here works by means of paper &
pencil. Actual information is transmitted by radio or
field telephone and drawn into a map.
Figure 1: TSD at command posts “in the field”
It is obvious that in time-critical processes with large
amounts of rapid changing information this leads to an
overload of the operators. Moreover, the display may not
show valid or actual information and causes errors in
decision-making. However, it brings along the advantage
that the commander is in the field: He gains high
situational awareness, experiences the terrain, cover,
weather, etc. and knows “what is really going on” at that
On the other hand there are TSDs at operation centres.
Tactical situation data is pre-processed and computers
are used to visualise the results.
The advantages of these computer-based TSDs are:
Actuality of data, provided that the communication
infrastructure is fast enough; and different views of
levels of data aggregation and possibilities to include
additional battlespace information.
But the flood of information may lead to an information
overload and data representation is still limited to two
dimensions and techniques of interaction with data have
to be learnt.
The approach of using VE as TSD first expands the twodimensional visualisation to three dimensions. This
means that height information can easily be perceived.
Additional elevation aids, like elevation profiles or
colour texturing, can be skipped and replaced by others
(e.g. reconnaissance photos, weather data, etc.
The more important thing is that general interaction with
data is simplified and happens more intuitively. This
facilitates an experience of the tactical situation and the
generation of a correct mental model. In an ideal VEsystem the computer is not realised as an active entity,
but becomes an invisible assistant which knows about
user intentions and supports him (Alexander et al.,
1997). Therefore operator workload is supposed to be
reduced and situational awareness to be increased.
2.2 Application of VE-Technology in C²-Systems
The amount of studies and applications in the area of VE
and VE-technology has increased rapidly recently. But
whereas VE is close to become applicable in research
and development and for single training applications,
studies considering the specific use of VE in C² have just
begun. Therefore knowledge in this area is limited and a
lot of projects are in a conceptual phase.
Most research studies and projects in this area have been
started in the past two years. Because of ongoing
development in this area this is only a brief overview.
Detailed information is given in Alexander et al. (1999).
Generally speaking, the approaches can be divided into
two groups. The first group consists of concepts and
long-term programs including VE-components. This is a
top-down approach, which is at high political level and
typically application-oriented. The second group is
characterised by specific VE-projects and laboratories.
Consequently it follows a bottom-up approach and is
presentation- and technology-oriented. Fortunately, there
are links between both so that they meet and synergetic
effects exist.
The Swedish ROLF (Mobile Joint Command and
Control System 2010) is a long-term program. Its goal is
to determine new possibilities for military commanders
of using VE-Technology in mobile command posts.
ROLF describes requirements for situational awareness,
decision-making and support, work methodology and
organisation of military crew and staff. The main idea is
to use modern methods and technology to help a group
of operators in difficult situations with complex, timecritical decision making. ROLF includes the Aquarium
as TSD, which is a semi-immersive VE-system. The
TSD is used to visualise positions of own and enemy
troops, positions of important institutions, terrain and
weather data in different views. Data pre-processing is
used to select the data displayed and ensures that only
important information is visible (Sundin, 1996).
Especially the realisation that in future battle scenarios
all actions of the military commander will be in an
unclear, vague environment and the importance of an
information dominance led to the development of the
Command Post of the Future Program (CPoF) of
DARPA (1998). The program’s goal is to accelerate the
decision making process with ongoing reduction of the
staff. Therefore new technology is needed to make
maximum use of the whole human perceptory system in
order to transmit maximum amount of information. This
includes an interactive, three-dimensional visualisation,
three-dimensional interaction with computer-generated
objects, presentation of inaccuracy and probability,
integration of dynamic factors, three-dimensional sym-
bolic, integration of natural language processing and
integration of knowledge-based systems.
The second, more technology-oriented group of
approaches is larger. Institutions and laboratories
working in this area use different VE-technology. The
technology is often reconfigured to be used in different
research projects and experiments.
The US Battle Command Battle Lab (BCBL) performs
conceptional studies as well as experimental analysis in a
VR-laboratory. One goal is to develop a technology for
multi-media, scene-based application in education and
training for organisation and staff functions. This system
shall be connected to the Internet to increase the range of
application (Heredia, 1999).
At the US Naval Research Laboratory (NRL) an
advanced battle planing and management system has
been developed. The system works with a semiimmersive display and enables multi-modal interaction.
It was found to be very suitable for virtual-prototyping,
interactive mission planing and increasing situational
awareness (NRL, 1997).
Similar approaches, like Mirage of the Army Research
Lab (ARL) (IST, 1997), the Visualisation Architecture
Technology (VAT) of the Crewstation Technology
Laboratory (CTS) (Achille, 1998) or the Electronic Sand
Table of MITRE Corp. (MITRE, 1998) also use a semiimmersive VE-technology, as described further on.
Other approaches use full immersive VE or desktop-VE
respectively (Dockery & Hill, 1996; Morgenthaler et al,
2.3 The Idea of an Electronic Sandtable
The Electronic Sandtable at FGAN/FKIE has been
developed as an advanced display for tactical information in mission planing, control and rehearsal. The
concept is based on the sandtable metaphor. The military
sandtable, as shown in Figure 2, consists of a sandy
model landscape with simplified objects representing
woods, buildings, points of interest or military units. It is
broadly used in military education and training.
manually. Moreover the accuracy for representing real
geographic data is poor.
It is intended to model the sandtable by means of a VEsystem. This way the system becomes capable of
presenting dynamics, enabling real-time interaction and
changes of the point-of-view while benefits of the real
sandtable remain.
For this purpose geographic data and tactic data have to
be visualised stereoscopically. It is intended to create a
model landscape, in which dynamic battle scenario is
Furthermore additional functionality can be added, e.g.
visibility, range of weapon systems, etc. The implementation of this idea will be described in detail in
chapter 5.
3. Virtual Environments (VE)
The basic idea of generating and using a computergenerated artificial reality was mentioned first in science
fiction literature at the middle of the 20th century. Due to
rapid development of computer technology in the second
half of the century, a partly realisation of this idea
became possible. Nowadays these VE-Systems are
commercially available and starting to be used for a
broad range of applications (Alexander et al, 1999).
According to Bullinger et al (1997), Virtual Environments (VE) describe the computer-based generation of
an intuitively perceivable and experiencable scene of a
natural or an abstract environment. It is characterised by
capacities for multi-modal, three-dimensional modelling
and simulation of objects and situations. A further
characteristic is the close interaction of the human
operator with the system.
In this connection, Virtual Reality (VR) has been defined
by NATO HFM-021 (nn.) as:
“... the experience of being in a synthetic
environment and the perceiving and interacting
through sensors and effectors, actively and
passively, with it and the objects in it, as if they
were real. Virtual Reality technology allows the
user to perceive and experience sensory contact
and interact dynamically with such contact in any
or all modalities.”
This definition of VR, which is often used as a synonym
to VE, overlaps with VE. But whereas VE is application
oriented, VR describes, strictly speaking, a total model
of the reality, including all manifold facets of it. As this
is not possible today and may not be possible in future,
the further article will be use the term VE.
VE-systems are on their way of becoming to be used for
different applications. The main applications have found
to be research and development, training, telemanipulation and teleoperations, mission support, and mission
rehearsal. Further information about military applications is given in Alexander et al. (1999).
Figure 2: Sandtable in military education
But the traditional sandtable is static; all changes of
deployment have to be done manually. Each change of
region is very time-consuming and has also to be done
4. Geographic Data
Geography is the science of analysis of the surface of the
earth and the earth-human ecosystem. The historic roots
reach back to the antique world when geography was
used for the description of land, coasts and harbours.
Still the description of the surface of the earth, called
cartography, is one of the largest domains of geography.
However, today geography is not limited to physics
(geomorphology, climate, hydrographics, soil science,
and geography of vegetation and animal), but includes
political, social, economic and cultural aspects as well.
The structure of geographic databanks depends on the
kind of application the data is intended for. Usually
offices for land register and military offices are the main
principals and users.
Data for military cartography has to be as exact,
complete and actual as possible. This means a complete
collection of data about all kinds of objects and the exact
registration of their geographic co-ordinates is main
criteria for structure of the referring databank.
The geographic data available is divided into (Helmuth,
• Raster data, which describes a subset of pixel data,
like scanned paper maps of different scales.
Assignment to other geographic data requires georeferencing by means of the determined values for
the map’s corners.
• Picture data, which comprises geo-referenced or
non-referenced aerial or satellite photos. Equalising
reference points or procedures of aerial triangulation
do geo-referencing.
• Vector data, which includes pre-processed data of
surfaces (e.g. woods, lakes), lines (e.g. streets, rivers)
or points (e.g. power poles, points of interests,
bridges, towers) and the positions of their bases and
attributes. Vector data is usually two-dimensional
feature data and has to be merged with elevation data
from other sources. For visualisation vector data is
linked to detail objects.
• Matrix data, which describes terrain data structured
and saved in matrix format. Usually, terrain data is
organised like this.
All categories differ from each other in quality,
resolution and actuality. Generally, data is available in
scales between 1:25.000 and 1:250.000. The most
common data-format is summarised in Figure 3.
Data Type Name
aerial photos
satellite photos
satellite photos
Resolution / Scale
(dep. on region)
1:50.000 - 1:2.000.000
1:50.000 - 1:2.000.000
1:50.000 - 1:1.000.000
1:32.000 & 1:70.000
10 m X 10 m
30 m X 30 m
1:25.00 - 1:1.000.000
1:1.000.000 & 1:2.000.000
1:50.000 - 1:250.000
30 m X 30 m
90 m X 90 m
90 m X 90 m
Figure 3: Different Formats of Geographic Data
(Helmuth, 1996)
With growing demand on realistic education and training
and ongoing technical development of displays new
requirements for geographic data are emerged. In the
future the main needs will be higher resolution and
realistic texturing.
However, it cannot be taken as granted that all data
required is available in the format, resolution and quality
needed for the application. For this reason, an extension
of one databank by different other databanks has to be
done. This may lead to inaccuracies and inconsistency
making further data processing necessary.
5. Electronic Sandtable (ELSA)
The Electronic Sandtable has been implemented as a
testbed at the Research Institute for Communication,
Information Processing and Ergonomics (FKIE). The
structure and implementation of the semi-immersive VEsystem is described in this chapter.
5.1 Baseline Structure
Because of the large size of geographic databanks and
the need for real-time interaction, the underlying
structure has been arranged in two stages (Alexander et
al., 1997). A draft of this subdivision of the structure is
given in Figure 4.
Graphic Objects
Tactical Situation
of the
of the
& optimizing
Phase I
- online -
Control of
Phase II
- online -
Figure 4: Structure of the Electronic Sandtable
The first stage is executed offline. In this stage the scene
graph is determined. The scene graph is a hierarchically
ordered databank of all polygons included in the visible
In a semi-automatic process data and objects are
selected, integrated and re-ordered with respect to
maximum rendering performance. This re-ordered
polygon-databank is called the scene graph. Afterwards
the scene graph stays constant without any changes of its
In the second stage additional data is constantly added
and the scene graph is visualised online. The additional
data, i.e. tactical situation data and data from external
data sources, is linked to objects of the scene graph.
Additional input of external data using different
protocols (DIS, HLA) shall also become possible in
future. The incoming data controls position and status of
military units. Additional data like actual situation
videos or information of knowledge databanks can, also
be included.
After that the rendering subsystem selects the visible
subset of the scene graph. Out of this two separate
projections are calculated and written into two-frame
buffer. Then both frame buffers are visualised alternately
on a horizontal plane.
The human operator interacts with the scene by means of
different interaction devices. The inputs serve as
commands, which affect the objects of the scene graph.
They are logged for later analysis.
The operator is able to select different visible areas for
navigation. The borders of the area serve as one input
variable of the rendering subsystem. Additionally each
of the operator’s movements is tracked by a headtracker. The position output of the tracker is another
input variable of the rendering subsystem for new
projection calculation.
5.2 Data Processing and Visualisation
For visualisation the geographic data has to be
transferred into the scene graph to be visualised. The
process is executed offline and done semi-automatically.
It is divided into data selection, pre-processing and
optimising phase.
In the first step an area of interest is selected and the
relating terrain (DTED) and feature (DFAD) data is
extracted. Additionally, links between features and
geometric objects are defined. Afterwards the selected
data is saved in a temporary buffer, which has to be preprocessed, and optimised for visualisation.
Geometric objects include the geometric description of
the object (e.g. tanks, aeroplane) and additional
information (e.g. unit status, damage reports, etc.). At the
stage of real-time visualisation they are shown at the
position given either by the geographic data or the
tactical situation data.
The following steps of pre-processing and optimising are
necessary because terrain and feature data are generated
from geographic databanks. These databanks were
designed with regard to different requirements, which
makes them unsuitable for a real-time, realistic
Pre-processing takes into account that consistency and
integrity are highly important criteria for databanks. If
datasets of more than one databank are merged,
contradicting data might emerge and cause errors. Those
errors are based on errors or inaccuracies in the original
databanks, different data resolution or different actuality
of data acquisition.
As soon as consistency and integrity is proved, the
process of merging terrain and feature data starts.
Geometric objects are appended and, if necessary,
adjusted to ground level.
Finally the triangulation process starts and determines
polygons for visualisation.
For real-time visualisation an optimising process has to
be performed to keep the amount of rendered polygons
minimal. Therefore the databank system transfers only
information about the visual subset. Non-visible parts
outside the field of view are clipped.
For further reduction the databank is re-organised and
the scene graph is tiled. In the visualisation process the
distance to the point of view sets the level of complexity
for each tile.
Different levels of complexity called levels of detail
(LOD) are another technique to reduce polygons. LOD
means more than one representation of different levels of
complexity (different amount of polygons) for the same
subset. This means, if a subset gets closer to the point of
view, a higher LOD with more polygons is visualised.
Using these techniques, data is re-organised with regard
to visualisation issues. The output of this process is the
scene graph, which can be visualised in real-time on the
5.3 Concept of Semi-immersive Display Technology
The display technology used for three-dimensional
visualisation is a semi-immersive virtual workbench.
Krüger & Fröhlich (1992) have originally developed this
concept. The baseline concept is shown in Figure 5.
Today it is used for various applications.
A projector projects two computer-generated, timealternated pictures onto a mirror. The mirror reflects
them to a horizontal focussing screen. By using shutter
glasses, i.e. LCD-glasses shading each side alternately
synchronous to the projection, the operators perceive two
separate pictures for the right and the left eye. The
synchronisation works by an emitter sending infrared
signals synchronously to the picture projected.
Operators wearing
Virtual Workbench
Figure 5: Principle of a Semi-Immersive Virtual
Finally, both pictures perceived are fusioned by the
cerebrum to a single, three-dimensional model.
6. Stereoscopic Visualisation
The design of the user interface of VE-systems has been
found to be one of the main criteria of quality for its
application. The Electronic Sandtable (ELSA) serves as
the interface between the real environment on the one
hand and the virtual scene on the other hand. Moreover it
uses a different metaphor than the desktop-metaphor
used in various computer applications. Therefore new
interaction techniques and procedures have to be
developed, analysed and optimised according to a high
performance of the human-computer-system (Alexander,
A realistic, three-dimensional visualisation of terrain
data has to consider the physiological procedures of
visual depth perceiving. These procedures have been
studied extensively, and several different hypothesis for
depth perceiving exist.
Each hypothesis postulates the existence of depth cues.
The classic depth cues will be summarised later in this
chapter. Of those especially the stereoscopic disparity
and parallax are of critical importance for the application
of the Electronic Sandtable.
A computer-based visualisation has to take into account
different depth cues. For stereoscopic visualisation
different viewing models exist. The common models will
be presented in this chapter as well.
6.1 Process of Visual Perception
The physiological visual system consists of the eye as
sense organ for stimulus acquisition, the optic nerve for
stimulus transfer and the optic centre of the cerebrum for
stimulus processing.
According to Schmidt & Thews (1995) the human eye
can be divided into two subsystems:
• Subsystem 1 performs the refraction of incoming
light. Its main components are Iris (control of
incoming light intensity), lens (refraction), vitreous
body (stability) and diverse muscles (adjustment).
• Subsystem 2, jointly with the central nervous system,
transfers the light to stimulus signals of nerve cells. It
consists of the retina with its two different light
The stimuli are transferred via the optic nerve to the
optic centres of the cerebrum. Here the optic sensing and
recognition takes place.
Visual perception is generally based on three stages of
perception (Kelle, 1994):
The first stage is an egocentric perception of the own
person. This allows a separation of objects of the own
body and other objects, making possible to determine the
own position with regard to other objects and an
absolute depth perception.
The next step is a comparison of the objects in the
environment, allowing a relative depth perception.
Finally memory, experience and internal processing
mechanism lead to depth cues being fundamental for
spatial perception.
6.2 Depth Cues
Depth cues are visual system cues, which enable
perceiving of spatial dependencies (Hodges, 1992;
Schmidt & Thews, 1995). They can be divided into
monocular and binocular cues.
Monocular cues are valid for perception with one eye
The main monocular cues are:
• perspective: The projection of three-dimensional
environment onto a two-dimensional display surface
has large influence on the subjective depth perception. Most common projection is the linear projection
characterised by parallel lines meeting at a single
vanishing point.
• difference in size: If same objects are shown at
different sizes, the larger object seems to be closer
than the smaller one. This criterion is basically a
consequence of the perspective depth cue.
• known dimensions of objects: Know sizes of objects
are also influencing the subjective depth perception.
• shading: Occluding and covering enables a perception of relative position of several objects with
regards to each other. The object shown with a closed
shapes is perceived as closer than the other.
• light and shadow: The shadow within an object
makes conclusions about its spatial structure
possible. Position and size of the outer shadow gives
information about the kind of object (mountain or
valley) and its size.
• accommodation: Examining and focussing an object
requires an adjustment of the refraction of the optical
lens to get a sharp picture on the retina. This is called
The binocular depth cues require the total binocular eye
system. They influence the perception of short to
medium distances.
Traditional binocular depth cues are:
• convergence: For examining and fixation of a point
with both eyes the eyeballs have to be counterrotated,
until both lines of sights meet at the fixated point.
Only if this happens the object is pictured at identical
points of both retinas and a further processing of the
stimulus is possible.
• disparity and parallax: If one object is focussed in
space, other objects are represented at non-corresponding retina areas, causing two different pictures
for the right and left eye. The disparity is defined as
the distance between both single pictures. Because of
the importance for the Electronic Sandtable, this
depth cue will be described in detail in chapter 6.3.
Additionally to these static cues further dynamic cues
exist which have large influences on the depth
perception for medium distances (17–29 m) (Kelle,
1994). Because they are of no relevance for the semiimmersive display technology, they will not be described
in this paper.
6.3 Disparity and Parallax
Disparity and parallax have a large influence on depth
perception and are the main depth cues for stereoscopic
visualisation. Therefore they are described more
The distance between both eyes leads to different
representations of an object on the retina of the right and
the left eye. Both eyes perceive the object with a
different perspective. The distance between both pictures
is described by disparity.
If an object is looked at, it is represented at the fovea of
both eyes. A round spatial surface exists (horopter),
representing all objects on it on corresponding retina
aerials. Objects at positions different from the horopter
are represented at non-corresponding retina areas. If the
distance from the horopter is not too large, the cerebrum
fusions the right and left picture to a three dimensional
model. If it is too large, disturbing double pictures are
perceived (Schmidt & Thews, 1995).
Disparity is a mathematical dimension and cannot be
determined practically. Therefore the dimension of the
stereoscopic parallax has been introduced. For this a
reference level has been used which is parallel to the
eyes’ level and runs through the fixation point.
Parallax has been defined as (Helmholtz, 1910, ref. in:
Kelle, 1994):
p = ba × a ×
e * t + e2
p = parallax
ba = inter ocular distance
a = distance eyes / reference level
e = distance reference level / object
t = distance eyes / object ( =a+e )
Parallax is also a dimension for depth separation and
depth perception. Therefore it is deduced that depth
perception decreases with square distance. Furthermore
it increases linearly with inter ocular distance.
According to Kelle (1994), stereoscopic disparity and
parallax has been found to be useful only for near and
medium distance (maximum of 6–9 m).
Visualisation of geographic data of large scale means a
large distance between eye point and surface. It can be
concluded that exact modelling means that parallax and
stereoscopic depth perception will be very low.
Consequently, an exclusive use of real values for the
model parameters (e.g. depth scale) would lead to no
stereoscopic depth perception and the scene would be
perceived flatly. On the other hand, too large values e.g.
for depth scale would make the terrain more
mountainous and may cause a wrong mental model of
the terrain. For an ideal depth perception these
parameters have to be adapted so that operators perceive
the terrain structure subjectively correctly. Therefore a
dynamic adaptation of the interocular distance of
operators and depth scaling is needed.
Pilot experiments for determining optimum inter- ocular
distance and depth scale have just started.
6.4 Stereoscopic Projection
For three-dimensional stereoscopic visualisation three
different projection models are commonly used. Their
baseline geometry is illustrated in Figure 6.
In Computer Aided Design (CAD), aerial photo analysis
and for head-mounted-displays (HMD) projection
models with parallel line of sights are used, as shown in
Figure 6 (a). They are based on the assumption of a
centre eye-point perpendicular to the projection plane.
Right and left projections are calculated by using offset
values and parallel shifting the projection right and left.
The disadvantage of this model is that the scene can only
be visualised underneath the projection plane. This is
inconvenient for the concept of the Electronic Sandtable,
because the scene would always be located beyond hand
range. Another disadvantage is clipping at the borders of
the display as missing visual information for either right
or left eye appears. Especially at large displays this is
very irritating for operators.
Figure 6 (b) shows the geometry of a projection model
using rotated line-of-sights. Here the projections are
rotated in the way that both lines-of-sight meet in the
projection plane. The lines-of-sight do not stay
perpendicular to the projection plane. It enables a
visualisation underneath and as well as above the
projection plane. There are no irritating effects on the
borders of the display either. But because of the special
geometry, an error of vertical parallax occurs. It can be
observed at the borders of the display, where both lines
meet at a point, which is above the projection plane. This
leads to a “winding”-effect and the scene seems to be
projected on a cylinder rather than a plane. Vertical
parallax has found to be irritating especially on large
The last projection model uses window projection, which
means that two windows are introduced through which
the virtual scene is perceived. The windows are
positioned in the same level as the projection plane. Both
lines-of-sight meet at the projection plane and remain
perpendicular to it. In this model, stereoscopic parallax is
only dependent on the distance to the display and no
vertical parallax is introduced.
ey e p oints
pro jec tion
Figure 6: 3 projection models: (a) parallel projection,
(b) rotated projection, (c) window projection
This model is used for the Electronic Sandtable. As
shown in Figure 7, an asymmetric pyramid describes the
model for each eye. This means, the perpendicular line
through the top does not meet the centre of the pyramid
For each projection six parameters are used to identify
the pyramid. They include the values for front, back, top,
bottom, left and right clipping plane. These values are
calculated by x,y,z-position of both eye-points, scale
factor and the display size as input.
Pilot experiments have shown good results for this
projection model. Only little perspective error due to
tracking of real eye position was determined. In future,
this error will be minimised by calibrating the tracking
VE-systems. The key criteria for answering this question
will be performance of the human-VE system.
For this reason human performance metrics will have to
be introduced, formulated and analysed. They should be
as fundamental as possible, but still take into account the
characteristics of the application.
Jointly with other basic research studies they will be the
key issues of future research in this area.
Figure 7: Right and left asymmetric projection pyramid
and boundary surfaces (clipping planes)
7. Conclusion and Future Research
In this paper the baseline concept of using semiimmersive VE-technology as advanced TSD has been
described. The approach has been shown to be promising
and advantageous.
It has been emphasised that human factors and
ergonomics are the main issues for reasonable
VE-application. In this paper some research issues were
introduced and results of ongoing research studies in the
area of visualisation were presented.
So far only real-size shapes have been visualised. In
future geographic data of different scales will be used.
To evoke a stereoscopic depth perception, an adaptation
of the scale factor for elevation as well as the dimension
of inter ocular distance is necessary.
Another research topic is the maximum vertical range of
the display. The display technique causes contradicting
depth information, because both eyes accommodate on
the projection plane, but fixate an object closer or more
far away. However, if the virtual scene is too close,
parallax becomes too large and the cerebrum cannot
fusion both pictures. Therefore another research topic
will be to determine the maximum useful vertical display
range and the variability of human sense perceiving.
Pilot experiments in the visualisation area have been
started and are currently going on.
Other important areas with high influence on the
applicability of VE in C² are interaction and cooperation.
Interaction with the databank means navigation in the
scene and manipulation of virtual objects. Procedures
(software) and interaction devices (hardware) have to be
designed, evaluated and analysed according to the
application for both subgroups.
The concept of the Electronic Sandtable has been
designed to enable multiple operators working in the
virtual scene. It has to include co-operation concepts. In
contrast to full-immersive VE, in semi-immersive VE all
operators are present at the same location.
Communication and inter-operator interaction work the
natural way. Therefore mainly human-computer
interaction issues have to be analysed. These main issues
and problems are the development of a general concept
for co-operation and co-operation procedures.
But even if in future the system works as it is supposed
to be, one question to be answered still remains: The
question for quantification of the profit and gain of using
Achille, L.B. (1998): Naval Air Warefare Center
Laboratory. CSERIAC, Vol. IX: 2 (1998). WrightPatterson AFB: FRL/HEC/CSERIAC, USA.
Alexander, T. (1999): Virtual Reality als Koordinationswerkzeug bei Planungsaufgaben. In: Gärtner (ed.):
Fahrzeug- und Prozeßführung. DGLR-Bericht 99-02.
Bonn: Deutsche Gesellschaft für Luft- und
Raumfahrt e.V., pp 51-63.
Alexander, T.; Grandt, M.; Gärtner, K.-P. (1999):
Virtuelle Umgebungen für die Gestaltung zukünftiger
Lagedarstellungssysteme. FKIE-Bericht Nr. 4.
Wachtberg-Werthhoven: FGAN-FKIE.
Alexander, T.; Wappenschmidt, A.; Gärtner, K.-P.
(1997): Ein Verfahren zur Darstellung von Geodaten
in Echtzeit. In: Möller (ed.): Tagungsband 5.
Workshop Sichtsysteme – Visualisierung in der
Simulationstechnik. Aachen: Shaker Verlag.
Bullinger, H.-J.; Brauer, W.; Braun, M. (1997): Virtual
Environments. In: Salvendy (ed.): Human Factors
and Ergonomics, pp 1725-1759. New York: John
Wiley & Sons.
DARPA (1998): Project „Command Post of the
Future“. http://www.darpa.mil/iso/cpof/main.htm
Dockery, J.T.; Hill, J.M. (1996): Virtual Command
Center. Proceedings of the 1996 Command and
Control Research and Technology Symposium, Sn.
61-68, Washington, D.C.: ACTIS.
Dockery, J.T.; Hill, J.M. (1996): Virtual Command
Center. Proceedings of the ICCRTS 96. Washington
Grandt, M.; Alexander, T.; Gärtner, K.-P. (1997): Virtual
Environments for the Simulation of a Tactical
Situation Display. In: Holzhausen (ed.): Advances in
Multimedia and Simulation. Human-MachineInterface Implication. Proceedings of the Europe
Chapter of the Human Factors and Ergonomics
Society Annual Conference. Bochum: University of
Applied Sciences.
Helmholtz (1910): „Handbuch der physiologischen
Optik“. Verlag von Leopold Voss, Hamburg.
Helmuth (1996): Überblick über die verfügbaren
MilGeo-Daten. Wehrtechnisches Symposium „Die
digitale Karte“. Mannheim: Bundesakademie für
Wehrverwaltung und Wehrtechnik.
Heredia, M.D. (1999): Battle Command Battle
Laboratory Fort Leavenworth, KS: Overview. http://
Hodges, L.F. (1992) : Time multiplexed stereoscopic
computer graphics. IEEE Computer Graphics &
Applications, pp. 20.
Institute for Simulation and Training (IST) (1997):
„Mirage - A New Way o View the Virtual World“.
Kelle, O. (1994): Dynamische Tiefenwahrnehmungskriterien in computergenerierten interaktiven Szenen
und virtuellen Simulationsumgebungen. Düsseldorf:
Krüger, W.; Fröhlich, B. (1992): The responsive
workbench (virtual work environment). IEEE
Computer Graphics and Applications, Vol. 14(3), pp
MITRE (1998): The Electronic Sand Table.
Morgenthaler, M.; Steiner, G.; Mayk, I. (1998): The
virtual command post. Proceedings of the 1998
Command and Control Research and Technology
Symposium. Monterey: Naval Postgraduate School.
NATO HFM-021 (n.n.): Proceedings of the Workshop
on Virtual Reality and Military Applications., Orlando, USA. Neuilly-sur-Seine: NATO
Naval Research Laboratory (NRL) (1997): „Virtual
Reality Responsive Workbench for Sea Dragon“.
Schmidt & Thews (1995): Physiologie des Menschen.
Springer-Verlag, Berlin, Heidelberg, New-York, pp
278-315 (Gesichtssinn und Okulomotorik).
Sedgwick, H.A. (1986): Space and Motion perception.
In: Handbook of Perception and Human
Performance, Vol. 1. New York: John Wiley & Sons.
Sundin, C. (1996): ROLF — Mobile Joint C2 System for
the year 2010. A vision under development and test.
Proceedings of the 1996 Command and Control
Research and Technology Symposium, Sn. 176-187,
Washington, D.C.: ACTIS.
Wohl, J.G. (1981): Force Management Decision
Requirements for Air Force Tactical Command and
Control. IEEE Transactions on systems, man, and
cybernetics, Vol. SMC-11 (9), pp 618-639.
Acquiring Distance Knowledge in Virtual Environments
Prof. Dr. Edgar Heineken1 & Frank P. Schulte2
Gerhard Mercator University
Lotharstraße 65
47048 Duisburg
Experimental results on the perception and cognition of
distances in virtual environments are reported. These
results show differences in the accuracy of distance
perception depending on whether they are presented in
desktop- or HMD-VR. In addition, they show that
distance cognition in virtual environments is based on
online-judgements (perception based) or on inferential
judgements (memory based) depending on the subject’s
goal when navigating through the environment. Without
an explicit goal to learn distances (incidental learning
condition) the estimated length of routes in a virtual
environment is inferred by the number of features
(feature-accumulation-hypothesis) experienced on the
respective route, just like in natural environments.
1. Introduction
A spatial environment can be explored directly or by
means of a map. A number of studies dealing with the
acquisition and representation of and the access to spatial
information have documented differences in spatial
learning associated with different modes of experience.
Direct and map experience lead to a different
understanding of the environment. Navigating through
an environment enables subjects to estimate route
distances and route orientations (route knowledge),
whereas Euclidean distances and locations of landmarks
(survey knowledge) are easier to estimate if the
environment is presented using a map (Thorndyke &
Hayes-Roth, 1982; Giraudo & Pailhou, 1994; Taylor &
Tversky, 1996).
Spatial cognition research is becoming increasingly
interested in the use of virtual environments as
experimental settings: virtual reality technology provides
both an economical and flexible design of realistic
environments as well as a reliable registration of the
subjects interactions with the environment.
The results of spatial cognition research are of practical
interest when virtual environments are used as visualisation or training tools. Thus the question arises, whether
there are differences in processing spatial knowledge
(landmark-, route- or survey-knowledge) in natural and
virtual environments (Wilson, 1997; Witmer, Bailey,
Knerr & Parsons, 1996; Ruddle, Payne & Jones, 1997;
Rossano, West, Robertson, Wayne & Chase, 1999).
The paper refers to the acquisition of distance-knowledge in virtual environments, and to the perception and
cognition of distances.
In chapter 2 an experiment designed to compare desk-top
and HMD-VR with respect to supporting distance
perception is presented.
In chapter 3 a series of experiments on distance
cognition in virtual environments are reported.
Chapter 4 summarises and discusses the results
2. Distance Perception and Perceived Depth in
Virtual Reality
There are different kinds of psychological spaces. A
vista space means a space up to 30 m, explored by
looking ahead without locomotion. This kind of
psychological space can be contrasted with the
environmental space (the entire space is not visible from
the starting position, it can be explored only by
locomotion), and the geographical space (the space is so
large, that it can be explored only by means of a map).
When designing vista spaces in virtual reality factors
determining human space-perception have to be
considered. There are nine different sources of information the human visual system uses as depth cues:
occlusion, height in the visual field, relative size, relative
density, aerial perspective, binocular disparities, accommodation, convergence and motion perspective (see
Cutting, 1997). It is of interest, however, whether the
perceiver’s kind of interaction with the virtual
environment (e.g. whether the view’s orientation
changes depending on the user’s head movements, or
not) may also affect their spatial sensitivity and thus
their perception of distances in the space.
The hypothesis that distance perception in a virtual vistaspace is more accurate in HMD-VR than in desktop VR
is tested in a bisection-experiment.
A total of 18 subjects (7 male and 11 female)
participated in the experiment. Their average age was 26
years, ranging from 20 to 36 years. The environment
used in this experiment was created and presented using
Superscape VRT 5.50 software, running on a 500 MHz
Pentium III PC equipped with 196 MB RAM and a 32
MB Matrox G400 graphic accelerator card.
The environment showed a small forest through which a
path led. The whole scene consisted of 8000 facets, and
the maximum frame rate was limited to 20 frames per
second to avoid lag differences. The environment was
rendered in 640 to 480 pixel resolution.
Half of the group of subjects experienced the virtual
environment by means of a head-tracked HMD
For contact with Prof. Heineken: [email protected]; tel. +49 203 379 2541, fax +49 203 379 1846
For contact with Schulte: [email protected]; tel. +49 203 379 2519, fax as above
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
(Virtuality Visette Pro combined with a Polhemus
InsideTrak), the other half viewed the environment as a
video projection (JVC DLA 10 SXGA). It was made
sure that the FOV in both VR conditions was identical.
In both conditions the subjects remained in a standing
The subjects were instructed to bisect a route presented
to them in the virtual environment by moving a marker
to the mid of the route. Figure 1 shows the virtual
Figure 1: Starting point (circle) and end point (bar) of
the presented route. The marker (triangle) has to be
moved to the mid of the route. (note: ground texture has
been deleted for printing reasons)
The presented routes differed with regard to their length
— short (approximately 150 cm) or long (approximately
600 cm) — and with respect to the starting position of
the marker (above the mid and below the mid). Each
route hat to be bisected four times by each subject, twice
in ascending and twice in descending order with respect
to the initial position of the marker. The subjects stood
400 cm away from the route’s starting point.
The participants in each experimental group were given
different amounts of time to explore the virtual environment before their bisection task (30 seconds, 60 seconds
and nor opportunity).
significant (F(1,12)=4,92, p<.05). The factors “route
length” and “experience with the virtual environment”
did not affect the bisection error.
The results are showing that immersion improves depth
perception and facilitates the judgement of distances in a
virtual vista space.
3. Distance Cognition in Virtual Environments
There are two conflicting theories which try to predict
the cognition of distances experienced in environmental
spaces: the Feature-Accumulation-Theory (Sadalla &
Magel, 1980) and the Route-Segmentation-Theory
(Allen, 1988). According to the first theory, the
cognitive distance of a route is positively correlated with
the number of features experienced on the route, whereas
the second theory proposes a positive correlation
between estimated distance and the number of segmentations of the route.
Within the scope of those theories on distance cognition
a series of experiments have been realised by Petra
Jansen-Osmann in our institute. In the following section
we report the main results of her doctoral thesis (JansenOsmann, 1999).
3.1 Number of route turns and estimated route length
The length of a route with more turns is estimated longer
than a route with fewer turns. This result of an experiment carried out by Sadalla & Magel (1980) was
replicated in a virtual maze. 20 subjects navigated
successively through two mazes. The routes were of
same length but differed in the number of turns (2 turns,
7 turns). Afterwards, they had to travel on a straight
route until the distance covered seemed equal to the
route travelled in the maze.
The covered distance was significantly de-pendent on
the number of turns (t(1,14)=3,56, p<.005). The route
with more turns was estimated longer (Figure 3).
Figure 3: Cognitive distance (magnitude estimation)
for routes with different numbers of turns
Figure 2: Bisection error (deviation from the real mid
in percent) for the short and the long route under
different VR-conditions (HMD, desktop)
The error of bisection was calculated as the absolute
mean differences between the estimated mid and the real
mid. Figure 2 shows that the bisection error is greater
when the line is presented in desktop VR than when it is
presented in HMD-VR. The difference is statistically
The result corresponds with both hypotheses on distance
cognition because turns can on the one hand be regarded
as features and on the other hand as borders of route
segments, i.e. as segmentations.
3.2 Feature accumulation or route segmentation as
determinants of distance cognition
In an experiment with 30 subjects distance estimations
for segmented routes, routes enriched with features and
empty routes were compared.
A street-scene was used in desktop-VR. On each side of
the street 9 identical looking houses were presented. The
number of houses and the number of crossways could
not seen by the subjects from the starting point
(Figure 4).
(active navigation), the other half experienced the street
without joystick (passive navigation). Both groups
experienced the environment three times successively.
Afterwards the subjects had to collocate the 9 houses on
a vertical line on a sheet of paper with respect to their
respective distances (Figure 6) and is consequently
Figure 4: User’s view at the starting point (note:
crossways cannot be perceived at the starting point)
The spacing between the houses (Figure 5) as well as the
location of crossways was varied. The subjects had to
estimate six different distances between houses.
Figure 6: Protocol sheet: Collocation of the houses
with respect to their distances
The results show that both the segmented and the filled
sections were estimated equally longer than the empty
sections of the route, and that this difference was more
pronounced when the subjects had actively explored the
virtual environment (Figure 7).
Figure 5: Empty sections (1), filled sections (2) and
segmented sections (3) of different length in a survey
map of the street scene used in the experiment
The whole route could be broken down in empty
sections (1), sections filled with a house (2) or sections
segmented through a crossways (3). Each kind of section
could be short or long. Half of the group of subjects
navigated through the virtual street using a joystick
Figure 7: Cognitive distances (estimated length in
relation to the total route length) of empty, filled and
segmented route sections experienced actively or
Only the effect of the route design (empty, filled,
segmented) on the distance estimations (F(2,56)=12,16,
p<.001) was significant, showing that feature accumulation as well as route segmentation determine distance
cognition in virtual environments.
3.3 Distance cognition based on the presentation of a
survey map
Survey maps of the virtual streets used in the last
experiment were presented to 15 subjects on a monitor
for 1 minute after they had actively explored the virtual
environment. Their distance judgements were clearly
dependent only on route segmentation and not on feature
accumulation (Figure 8).
Figure 8: Cognitive distances (estimated length in
relation to the total route length in percent) of empty,
filled and segmented route sections experienced in
a map
The route design significantly influences the distance
estimation if the environment is presented on a map
(F(2,28)=8,73, p<.01) but only the route segmentation —
and not the feature accumulation — determines the
perceptual organisation of the map and as a consequence
the distance cognition (see Figure 9). When the street
scene is presented as simultaneous structure the distance
between two houses segregated by the crossways is
perceptually strengthened.
Figure 9: Cognitive distances (estimated length in
relation to the total route length in percent) of empty,
filled and segmented route sections in the case of
incidental and intentional learning
3.4 Online- vs. inference-based distance judgement
The learner’s goal when navigating through a virtual
environment is a crucial encoding-factor in the
processing of distances. In an experiment 30 subjects
navigated through the same virtual environment used in
the last two experiments. Half of them were instructed
that afterwards they would have to estimate distances,
whereas the other half was not explicitly instructed to
focus on the distances.
There is a systematic interaction between the factors
“route-design” and “kind of learning” (F(2,56)=11,36,
p<.01) indicating that route-segmentation or featureaccumulation determine distance cognition only in case
of incidental learning. If distances are learned
intentionally, which means that the subjects encode
distance directly, features and segmentations have no
effect on the distance estimation: the distance estimation
is based on the perceived distances (online judgement).
In the case of incidental learning distances are not
encoded directly, they are inferred afterwards using
houses or crossways as heuristics (inference-based
4. Concluding Remarks
Accurate distance perception and distance cognition are
necessary for applying VE in the field of training and are
therefor a prerequisite for its validity as a training tool.
There are differences in the accuracy of distance perception depending on whether the environments are
presented in desktop- or HMD-VR: immersion improves
depth perception and facilitates the judgement of
distances in a virtual vista space. Obviously the perceiver’s sensumotoric interaction with the virtual
environment provided by the tracking system enhances
his spatial sensitivity.
Distance cognition in a virtual environmental space can
be based on online-judgements (perception based) or on
inferential judgements (memory based) depending on the
subject’s goal when navigating through the environment.
The learner’s goal is a crucial encoding-factor in the
processing of distance-information. It determines the
kind of spatial knowledge transferable from the virtual to
the natural environment.
When VE are applied in the training of real world skills
based on accurate distance perception and cognition, the
designer should be familiar with psychological factors
which determine the learner’s spatial encoding and
judgement of distances (e.g. the role of featureaccumulation). It was shown that without an explicit
goal to learn distances the learner stores general
information (features) when navigating through the
environment, and later on judges the distance of a route
by using the frequency of features experienced on the
route as a heuristic.
Cutting, J.E. (1997). How the eye measures reality and
virtual reality. Behavior Research Methods,
Instruments & Computers, 29 (1), 27-36.
Gillner, S. & Mallot, H.P. (1998). Navigation and
Acquisition of Spatial Knowledge in a Virtual Maze.
Journal of Cognitive Neuroscience, 19, 445-463.
Giraudo, M. & Pailhous, J. (1994). Distortions and
fluctuations in topographic memory. Memory and
Cognition, 22, 14-26.
Jansen-Osmann, P. (1999). Kognition von Distanzen —
laborexperimentelle Untersuchungen in virtuellen
(http://www.uniduisburg. de/diss/diss9906/)
Ruddle, R.A., Payne, S.J. & Jones, D.M. (1997).
environments: experimental investigations using
extended navigational experience. Journal of
Experimental Psychology: Applied, 3, 143-159.
Sadalla, E.K. & Magel, S.G. (1980). The perception of
traversed distance. Environment and Behavior, 12,
Taylor, H.A. & Tversky, B. (1996). Perspective in
spatial descriptions. Journal of Memory and
Language, 35, 371.391.
Thorndyke, P.W. & Hayes-Roth (1982). Differences in
spatial knowledge acquired from maps and
navigation. Cognitive Psychology, 14, 560-589.
Tlauka, M. & Wilson, P.N. (1996). Orientation-free
representation form navigation through a computersimulated environment. Environment and Behavior,
28, 647-664.
Wilson, P.N. (1997). Use of virtual reality computing in
spatial learning research. In N. Foreman & R. Gillet
(Eds.), A Handbook of Spatial Research Paradigms
and Methodologies. Vol. 1. East Sussex, U.K.:
Psychology Press, pp 181-206.
Witmer, B.G., Bailey, J.H., Knerr, B.W. & Parsons, K.C.
(1996). Virtual spaces and real world places: transfer
of route knowledge. International Journal of Human
Computer Studies, 45, 413-428.
Development of Virtual Auditory Interfaces
LCDR Russell D. Shilling1, MSC, USN
United States Air Force Academy
Colorado Springs, CO 80840
1. Introduction
The design of visual components in virtual environments
has shown rapid improvement and innovation. However,
the design of auditory interfaces has lagged behind.
Whereas visual scenes have become more compelling,
the auditory portions of VE remain rudimentary. This
disparity is perplexing since auditory cues play a crucial
role in our day-to-day lives. Imagine entering a meeting
with a room full of people. When you enter the room,
you realize that the speaker’s voice is emanating from all
points in the room, yet the room is totally anechoic. In
addition, you see other attendees moving in the room,
yet there are no additional noises in the room except the
speaker’s voice. Despite walking into a “real”
environment, your sense of reality would most probably
be challenged. In fact, it is generally believed that the
sense of presence is dependent upon auditory, visual, and
tactile fidelity (Sheriden, 1996). Although the sense of
realism in VE is also dependent on visual fidelity, virtual
or spatial sound has been shown to increase the sense of
“presence” (Hendrix, 1996). It stands to reason that
when we develop poor auditory interfaces in a VE, the
perceived quality of the entire VE is compromised
(Storms, 1998). The problem with audio is that our
normal auditory environment is “transparent”. We don’t
consciously process a sound in our environment unless
we NEED to attend to it. Yet, when slogging through
mud while on patrol, soldiers use auditory cues to keep
track of the people around them while scanning for
threats in front of them. They don’t need to keep looking
at the people around them. While not consciously
processing the sounds of their comrades, if someone
stops walking, they’ll recognize the lack of sound
2. Methods of Sound Presentation
There are a variety of ways to present sound in virtual
environments. The most traditional method is to use
speakers to present sound either monaurally, in stereo, or
in surround sound. Speaker systems are bulky, do not
typically provide elevation cues, and do not allow the
sound engineer to have complete control of the auditory
environment. Speaker systems DO allow for the
possibility of presenting auditory stimuli such that the
entire body is stimulated, especially when powerful
subwoofers are employed. On the other hand, using
headphones in conjunction with signal processing
techniques, it is possible to generate stereo signals that
contain most of the normal spatial cues available in the
real world. Spatialized audio uses actual pinna cues
Tomasz Letowski, Ph.D.
Army Research Laboratory
Aberdeen Proving Ground, MD 21005
stored as Head Related Transfer Functions (HRTFs) to
give the perception of auditory objects as completely
externalized in azimuth and elevation (Wightman &
Kistler, 1989; Begault & Wenzel, 1993). When coupled
with a headtracking device, spatialized audio provides a
true virtual auditory interface. Using a spatialized
auditory display, a variety of sound sources can be
presented simultaneously at different directions and
distances. One of the early criticisms of spatialized audio
was that it was expensive to implement, however, as
hardware and software solutions have proliferated, it has
become feasible to include spatialized audio in most
systems. Spatialized audio solutions can be fit into any
budget, depending on the desired resolution and number
of sound sources required. Most head-mounted displays
are currently outfitted with headphones of sufficient
quality to reproduce spatialized audio, making it
relatively easy to incorporate spatialized audio in an
immersive VR system. A complete lexicon for
understanding and developing auditory displays can be
found in Letowski, Vause, Shilling, Ballas, Brungert &
McKinley (2000).
3. Effects of Auditory Displays on Performance
Illustrating the importance of sound, research conducted
using spatialized auditory displays has demonstrated the
importance of spatialized auditory cueing for reducing
response time in cockpit applications. Spatialized
auditory threat and attack displays were designed and
implemented for both the pilot and co-pilot gunner in an
AH-64 simulator at the Army Research Institute at Fort
Rucker, Alabama (Shilling & Vause, 1999; Shilling,
Letowski, & Storms, 2000). In this application, a
ground-to-air missile display was supplemented with a
spatialized auditory cue corresponding to the actual
location of the missile relative to the pilot and co-pilot
gunner. Figure 1 shows the difference between
spatialized and normal displays for the response time to
make the first 5 degrees of turn away from an incoming
threat. Response time was reduced by approximately 350
msec. These data are consistent with previous research
which demonstrated that response time to visual targets
was significantly reduced when paired with a spatialized
auditory stimulus (Perott et al., 1991) and the latency of
saccadic eye movements was reduced when using
spatialized auditory cues (Frens, Opstal & Willigen,
1995). In this same manner, auditory cueing can be used
to compensate for the effects of limited FOV HMDs
(Shilling, 1996). Applications can be further
supplemented by exaggerating normal auditory cues
For contact with author: [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
through so-called “supernormal localization” (Durlach,
Shinn-Cunningham et al., 1993). Finally, using
spatialized sound, speech intelligibility can be improved
when applied to multi-user virtual environments and
multi-channel radio communications (Haas, Gainer,
Wightman, Couch & Shilling, 1997).
Figure 1: Difference between spatialized and normal
4. Lessons from the Entertainment Industry
The entertainment industry has recognized the
importance of sound processing for over a century and
has learned many important lessons that can be applied
to problems in VE. At the beginning of the century, the
Edison Standard Phonograph represented the cutting
edge in audio technology. The method for cutting
grooves in the wax cylinders was robust and resistant to
the effects of scratches. However, consumers soon
abandonedwax cylinders with vertically etched grooves
for the less robust wax platter with horizontally etched
grooves, because the platters were easier to store. Today,
even though we have the technology to create astounding
audio when developing VE’s, it is more convenient to
ignore the auditory interface because customer’s aren’t
“requiring” high quality audio, software applications are
not typically easy to implement, and the contributions of
high quality sound are more subtle than for visual cues.
For instance, in motion pictures, sound has long been
recognized as playing a crucial role in the emotional
context of a film. Current efforts in my research are
focusing on applying lessons learned from the film
industry to problems associated with sound quality and
emotional content in VE. Much can be learned about
auditory special effects and sound system design from
Hollywood. The first real attempt at immersing the
audience in sound occurred with the production of
Disney’s “Fantasia” in 1939. Disney’s sound engineers
created a system called “Fantasound” which wrapped the
musical compositions and sound effects of the movie
around the audience. Though not a stereo production, the
effects were quite astounding. However, the system
required massive amounts of vacuum tube electronics
and 54 speakers spread around the theater at a cost of
$84,000 per theater. Virtually no theaters invested in the
system and “Fantasound” was never used again. Today,
we have a similar problem with applying sound in VE.
Although the cost of consumer audio equipment has
rapidly increased in quality and decreased in cost,
systems designed for VE’s are currently expensive and
the development software to implement them is limited.
Spatial audio sound servers, for example the AuSIM
Acoustetron and the Tucker-Davis Technologies PD-1,
typically cost in excess of $12,000. High cost and
limited software availability are clearly the result of a
lack of competition in audio products for VE.
5. Systematic Approach to Sound Design
On the practical side, the problem is not with the
software engineers as much as with the lack of a clear set
of requirements for implementing sound in VE. What is
needed is a systematic approach to rendering the
auditory environment necessary for any given
application. When we want to render visual scenes, we
rely on film as a reference. Unfortunately, when we
design auditory scenes, we typically rely only on
memory. In my laboratory, I am currently attempting to
develop a systematic approach to cataloging the auditory
environment to give the software engineer an objective
reference to compare the sound in the VE with the real
world experience.
One of the current efforts in my lab is to develop a
systematic approach for obtaining baseline data
concerning the content of an auditory environment. In
addition to cataloguing the different sounds in a real
environment, it is also important to systematically
measure the intensity of sounds being experienced by the
listener. In this manner, the VE developer has a highly
detailed reference with which to compare the real world
auditory environment with the virtual auditory
environment. Two systems are currently being evaluated.
The first system uses a portable Sony TCD-D8 DAT
recorder coupled with Sennheisser microphone capsules
(Figure 2). The microphone capsules will be inserted
into an observer’s auditory meatus (ear canal). In this
manner, a complete spatialized recording can be made of
the auditory environment, completely externalized with
azimuth and elevation cues. The second system (Figure
3) is more robust, using a larger set of microphones
produced by Core Sound which can clip to a set of
eyeglasses to produce a binaural recording, complete
with interaural time and intensity cues. Although, pinna
cues cannot be utilized, the advantage of the latter
system is that it would be more tolerant of extreme
conditions, especially if the recordings are made
outdoors. Both systems can be clipped to the belt and
will be used in conjunction with a real time logging and
event analyzer (CEL 593). The complete data set
including sound recordings and sound measurements
will be stored on CDROM for ease of use. The digital
recordings also allow for spectral analyses to be
conducted on specific auditory stimuli contained on the
tape so that synthesized versions of those stimuli can be
Figure 2: The used portable Sony TCD-D8 DAT
recorder coupled with Sennheisser microphone capsules
Begault, D.R. & Wenzel, E.M. (1993). Headphone
Localization of Speech. Human Factors, 35 (2),
Durlach, N.I., Shinn-Cunningham, B.G. et al. (1993).
Supernormal auditory localization. I. General
background. Presence, 2 (2), 89-103.
Frens, M.A., Van Opstal, A.J. & Van der Willigen,
R.F. (1995). Spatial and temporal factors determine
auditory-visual interactions in human saccadic eye
movements. Pereption & Psychophysics, 57 (6),
Haas, E. C., Gainer, C., Wightman, D., Couch, M. &
Shilling, R.D. (1997). Enhancing System Safety
with 3-D Audio Displays. Proceedings of the
Human Factors and Ergonomics Society 41st
Annual Meeting. 868-872, Albuquerque, NM.
Hendrix, C. & Barfield, W. (1996). The sense of
presence in auditory virtual environments.
Presence, 5 (3), 290-301.
Letowski, T., Vause, N., Shilling, R., Ballas, J.,
Brungert, D. & McKinley, R. (2000). Human
Factors Military Lexicon: Auditory Displays (ARL
Technical Report, ARL-TR-in print). Aberdeen
Proving Ground, MD: Army Research Laboratory.
Figure 3: The set of microphones produced
by Core Sound
Perott, D.R., Sadralodabai, T., Saberi, K. & Strybel,
T.Z. (1991). Aurally aided visual search in the
central visual field: effects of visual load and visual
enhancement of the target. Human Factors, 33,
Sheriden, T.B. (1996). Further Musings on the
Psychophysics of Presence. Presence 5(2), 241246.
Shilling, R.D. & Vause, N. (1999). Tri-service
cooperation, a key aspect of ARL-HRED spatial
audio research. Focus Army Research Laboratory,
5 (5).
Shilling, R.D., Letowski, T. & Storms R. (2000).
Spatial Auditory Displays for use within Attack
Rotary Wing Aircraft. Proceedings of the
International Conference on Auditory Displays,
April 2000, Atlanta, GA.
Storms, R.L. (1998). Auditory-visual cross-modal
perception phenomena. Doctoral Dissertation.
Naval Postgraduate School, Monterey, California.
Wightman, F.L. & Kistler, D.J. (1989). Headphone
simulation of free-field listening. II. Psychophysical
validation. Journal of the Acoustical Society of
America, 85, 868-878.
Educational Conditions for Successful Training
with Virtual Reality Technologies
Dr. Alexander von Baeyer & Dr. Hartmut Sommer
Einsteinstraße 20
85521 Ottobrunn
The paper focuses on those pedagogical conditions,
which should be met, in order to assure successful
training using virtual reality (VR) technologies.
Therefore, neither new technical inventions nor large
scale technical experiments are the issue of this paper.
Instead a systematic catalogue of pedagogical questions
will be proposed, which should be answered, before
virtual reality is planned for training purposes.
The pedagogical catalogue is derived from the basics of
educational psychology and media didactics. It
• a taxonomy of learning objects, which are most
suitable for virtual reality
• an analysis of training strategies and methods, as to
how well they are suited for training in an almost
entirely synthetic environment
• an analysis of the transfer of training, when virtual
reality is the major training medium
• and finally rules and basic cost data, which may help
to conduct cost effectiveness analyses.
In this paper I will try to give a short and comprehensive
overview on the basics of educational theory, which
should be applied to training with VR technologies. I
will do this in five statements. Each statement or thesis is
accompanied by explanations. I start with a new look on
a well known definition.
Probably everyone in this conference knows, what VR
is. Nevertheless, I will give my own add-on to a
commonly used definition and comment this definition. I
do this, because I want to define important educational
The common definition reads as follows:
VR is “a multi-dimensional human experience which is
totally or partially computer generated and can be
accepted by those experiencing the environment as
consistent” (NATO DRG Panel 8 on Human Sciences,
RSG 16).
My add-on is:
VR is a capability beyond life, virtual and constructive
simulation and of course much beyond Computer Based
Training systems, however it can be coupled with CBT.
VR can be created, in order to convey training objectives
and support training strategies.
Basic Statements
1. Statement
If training is the aim of VR, VR training programmes
must comply with the basics of social and educational
These basics do not differ from what should generally be
valid about training with constructive and virtual
simulation. VR is an other example that there should be
such things as simulation didactics. VR, however,
increases the pedagogical requirements to be considered.
These requirements concern mainly
• the distribution of learning material in a multisensory (multi-channel) experience (e.g. seeing,
hearing, feeling of one’s own body, feeling of
material properties, stress, decision making)
• the real experienced presence of an instructor and of
other students during the learning and exercising
process (social learning)
• the merging into VR and leaving the virtual
environment (e.g. different feeling of own security).
Related to these three general problems are the following
practical questions, which will partially be answered in
this paper:
• Are VR technologies justified by relevant training
• Do VR training programmes enhance the quality of
instruction and bring about better training strategies?
• Can the typical military crew and leadership
behaviour be preserved in VR, where this is
necessary for training?
• Are the offerings of VR accepted by experts of
training and operation as an environment that
facilitates learning?
• Will there be a chance to construct a consistent
training scenario with new synthetic elements of the
human environment?
These are the educational questions, which the VR
community is invited to discuss further.
2. Statement
If we take the classical taxonomy of learning objectives,
VR can be a relevant medium in complex psycho-motor
training, only for certain cognitive tasks, may be to
indoctrinate in the emotional and affective domain and
(as a still controversial matter) in a real social context.
In principal VR is useful for the following four types of
non-trivial application:
For contact with authors: [email protected] and [email protected], respectively; tel. +49 (0)89 6088 2127, fax +49 (0)89 6088 3612
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
• Perceptual-motor learning, where real images are
mixed with virtual components, e.g. the real hand
manipulating computer generated interfaces (this is
also called Augmented Reality),
• Perceptual cognitive training, when it becomes
necessary to build a “mental map” on the basis of
experience from various sense channels, not only
based on the visual system, e.g. complex assembly
tasks involving orientation in space, finding objects
and moving them from one place to an other,
discriminating different objects
• In general for team training in large scale exercises
like C² training, large staff exercises, disaster control,
but only as far as co-ordination skills and procedures
are concerned
• And finally the exploration of unknown
environments, provided that the data are up to date.
Examples for these types of application are
• Mission rehearsal, where all merits of VR are
• Reconnaissance, where VR however must have an
added value to conventional simulation and training.
valuable for articulated teaching and learning strategies.
These features are:
• a broader perceptual spectrum
• a higher degree of differentiation in the perceptions
(e.g. more depth cues)
• a higher degree of interactivity with the virtual
These three properties of a deeper immersion into the
artificial world offer the possibility, to differentiate and
structure learning activities in a more effective way.
The advantages of learning and teaching with VR
technologies are:
• more learning material can be presented to the
• part task and part function training can be applied to
a broader variety of learning tasks
• feedback control of learning success can become
more differentiated and apply to a broader spectrum
of tasks
• it may become easier to compose a set of part tasks to
a real world like whole task in a almost realistically
perceived learning environment.
The total immersion into a synthetic environment leads
to the exclusion of non-intended and disturbing
information. This fact can be used or better: misused for
indoctrination purposes. Sales promotions, radical
behaviour changes, rapid conveying of emotional
stimulus response patterns can be the objectives of such
techniques. This again leads to the question, if and how
much VR inhibits the ability of critical distance to the
learning of those tasks, which require a critical attitude,
e.g. all tasks comprising decision making between not
fully transparent alternatives.
The impact of total immersive VR technology on the
emotional behaviour is therefore a challenging new
research question.
However, VR requires a much more developed art of
constructing the curricula and of designing the learning
programmes and the learning aids. In short: VR makes
the training development much more demanding and
requires higher developmental qualifications.
Social learning is however not yet sufficiently
researched in fully immersive VR. The main problem
lies in the isolating effect of VR. This means that it is
still a not yet proven hypothesis, whether the acquisition
of interpersonal skills, even and especially if they are
interconnected with cognitive or procedural tasks can be
supported by those VR technologies, which isolate the
individual from direct personal contact with another
individual in the same learning group. There are,
however, semi-immersive VR-technologies like the
cave- technique or the virtual workbenches, where
individuals interact with each other “naturally”. These
techniques cover therefore in principle the all classes of
learning objectives.
3. Statement
Training Strategies in VR do not differ much from those
in virtual simulation and in CBT. However, they require
more dedicated analysis and development, because VR
offers more perceptual cues.
In comparison to constructive and virtual simulation VR
has some distinctive features, which make it particularly
4. Statement
The transfer of training into the operational situation
has to be carefully analysed, because VR represents
nevertheless only a part of “real reality”.
As we have already said, the social dimension of reality
is still hardly present in learning with VR technologies.
Along with this, decisive other aspects of the learner are
still drastically altered. These are
• the perception of the bodily self, which may be
necessary in many psycho-motor learning tasks
• the unnatural feeling of wearing a helmet or a glove,
which does either not resemble the normally worn
helmets and gloves, or is a totally unrealistic feeling
• the multi-sensory perception of the environment. e.g.
the not real feeling to walk a distance
• the apperception of the partner in the learning
process, whenever this may be required for the
acquisition of team building skills
• the apperception of the instructors, whenever this
may have a motivational effect on the learning
process or is a part of team building skills —
remember that in typical military tasks training and
personal example and leadership can not be
All this means that skill acquisition by means of VR
technologies puts the learner in sometimes extremely
artificial surroundings, encapsulates his consciousness
and lets him leave this virtual world with a repository of
artificial behaviours. The first thing after leaving the
artificial world of VR is to re-learn those behaviours,
which do not fully comply with the operational
environment, to de-condition the learner away from the
partially reduced and partially enriched experience
towards a normal interaction with the operational
environment. This again means, that although VR is an
expensive training and an often time valuable medium,
the transfer of training cannot be taken for granted and
must be ascertained with much effort. If the curricular
and didactic analysis has identified those tasks and skills
that cannot be trained with VR, the transfer of training of
the remaining VR-prone tasks can be evaluated without
too big problems.
5. Statement
Cost and effectiveness of training with VR must be
compared with training using virtual simulation.
Whenever virtual simulation is feasible, VR should be
analysed, whether it can produce better or cheaper
solutions than virtual simulation.
On the effectiveness side of the comparison cost
effectiveness analyses should consider the following
• The enhanced representation of new and extended
sensorial perceptions may increase the effectiveness.
• The possibility of mission rehearsal and procedural
training in extreme situations, where total immersion
is the only realistic experience, may also increase the
effectiveness (good example may be the training for
operations and maintenance in space or deep water).
• The reduced personal and interpersonal experience is
definitely a factor, which decreases the effectiveness
of VR in training.
On the cost side of the comparison the following issues
should be considered:
• The HMD technology is a cost decreasing factor.
• The software development is a drastically increasing
• Re-training and special transfer of training analyses
can become cost increasing factors.
Therefore, considering VR for training should always
start with cost effectiveness analyses based upon
thoroughly conducted training analyses. However, the
cost savings can reach several orders of magnitude, if
training using VR is correctly designed. Examples are
cargo handling skills or air drop skills, where the real
aeroplane would be too expensive and virtual simulation
is not giving the necessary depth cues.
To conclude this survey: What are the conditions of
success of VR in training?
1. For the time being a limitation to tasks, which do not
imply any personal proximity of other persons.
2. For the future more critical research into the
interpersonal and social impact of VR and how far
social interactions can be simulated in an total
immersive environment.
3. Always limitation to empirically researched and
proven simulation cues.
4. Always embedded in a well controlled transfer of
training evaluation.
5. Always planned on the basis of cost effectiveness
Entertainment Technology and Military Virtual Environments
Michael R. Macedonia1
US Army Simulation, Training, and
Instrumentation Command
12350 Research Parkway,
Orlando, FL 32826, USA
Paul Rosenbloom
University of Southern California
Information Sciences Institute and
Computer Science Department
4676 Admiralty Way
Marina del Rey, CA 90292, USA
…Knowing how to create compelling experiences; do low-cost, high-performance computing; support largescale network simulations; build graphics-modeling software is (or will be) [the entertainment community]
stock and trade. In these areas not only will it be futile for the Army to try to compete, but a waste of energy
and resources. Bran Ferren [1]
Bran Ferren makes a compelling argument that the
Entertainment Industry is driving the technology
advances needed for military virtual reality systems.
Moreover, the military virtual environment community
may actually be falling behind its civilian counterparts
by ignoring the rapid changes going on in entertainment
computing. These advances include low cost computer
graphics, agent technology, and the use of 3D audio.
In this paper we will explore some of the reasons how
and why the Entertainment Industry is advancing the
state of virtual reality (VR). We will also look at the
current problems of military simulation, particularly its
lack of story and emotion. Finally, this paper examines
how the US Army is trying to address these issues with
the establishment of the Institute for Creative
Technology (ICT).
The Entertainment Industry
The Entertainment Industry has in many ways grown far
beyond its military counterpart in influence, capabilities
and investments. For example, Microsoft alone expects
to increase R&D spending next year by 23 percent, to
$3.8 billion — compared to the US Army’s $ 1.2 billion
science and technology budget. The Interactive Digital
Software Association estimates that in 1998, interactive
entertainment businesses invested approximately $2
billion in new technology R&D, with an increase of
more than 20 percent. [2] This far outweighs current US
Army research and development for training and
simulation technology.
Moreover, the advances in the industry cannot be
ignored. Witness the rapid pace of development of the
graphics systems for game consoles and personal
computers  almost double performance every nine
months [3]. Compare this with the relatively slow gains
in “high-end” graphics platforms being used for the
According to Richard Weinberg at the University of
Southern California’s School of Cinema-Television,
Sony’s upcoming PlayStation 2 is an example of a
consumer-grade advanced technology gaming platform
that could revolutionize both the world of home gaming
as well as interactive training for the Army. The PS2 is
expected to have 34 times the power of the current
leading game system, the Sony PlayStation, and more
than twice the graphics performance of SGI’s (formerly
Silicon Graphics) high-end visualization system, the
Infinite Reality 2. Here is what Game Informer
Magazine (May 1999) says about the upcoming
Playstation 2: “PlayStation 2 could be a glimpse at
Hollywood of the 21st Century. Developers with this
kind of power in their hands could theoretically create
real-world environments, with living breathing
characters all affected by real-world physical attributes
such as gravity, friction and mass. Plus, PS2 can
accurately simulate different materials such as water,
wood, metal, and gas  real worlds that look like real
worlds. Full motion video that’s not full motion video,
but real-time game play with speaking characters, fluid
motions, and facial expressions.”
Playstation 2Graphics Synthesizer – Features and
General Specifications:
• GS Core: Parallel Rendering Processor with
embedded DRAM
• Clock Frequency: 150 MHz
• No. of Pixel Engines: 16 (in Parallel)
• Embedded DRAM: 4 MB of multi-port DRAM
(Synced at 150MHz)
• Total Memory Bandwidth: 48 gigabytes per second
• Combined Internal Data Bus Bandwidth: 2,560 bit
• Read: 1,024 bit
• Write: 1,024 bit
• Texture: 512 bit
• Display Color Depth: 32 bit (RGBA: 8 bits each)
• Z Buffering: 32 bit
• Rendering Functions: Texture Mapping, Bump
Mapping, Fogging, Alpha Blending, Bi- and TriLinear Filtering, MIPMAP, Anti-aliasing, Multipass Rendering
For contact with authors: [email protected], and [email protected]
Paper presented at the RTO HFM Workshop on “What Is Essential for Virtual Reality Systems to Meet Military
Human Performance Goals?”, held in The Hague, The Netherlands, 13-15 April 2000, and published in RTO MP-058.
Rendering Performance:
• Pixel Fill Rate: 2.4 giga pixel per second (with Z
buffer and Alphablend enabled), 1.2
• giga pixel per second (with Z buffer, Alpha and
• Particle Drawing Rate: 150 million/sec
• Polygon Drawing Rate: 75 million/sec (small
polygon), 50 million/sec (48 pixel quad with Z and
A), 30 million/sec (50 pixel triangle with Z and A),
25 million/sec (48 pixel quad with Z, A and T)
• Sprite Drawing Rate: 18.75 million (8 × 8 pixels)
Digital Output:
• Digital TV (DTV)
• VESA (maximum 1280 × 1024 pixels)
Other technical trends that will likely shape the military
training world will be digital cinema, the convergence of
television with the World Wide Web, and the continued
rapid growth of multiplayer Internet 3D games such as
Sony’s Everquest.
Weinberg also notes that, from a content perspective, the
computer game industry has considerable expertise in
games relevant to aspects of military training, with
significant interest in war games, simulations, and
military-like shooter games. For example, TalonSoft’s
The Operational Art of War II is expected to cover the
Vietnam War, Arab-Israeli wars, the Iran-Iraq conflict,
and Operation Desert Storm at the operational command
level, as well as several hypothetical conflict scenarios
ranging from India/Pakistan to a new Korean conflict.
Extreme Tactics, Warbreeds, and WarZone 2100 are but
a few examples of the war/strategy/shooter-style games
available. According to the May 15, 1999 issue of
Games Business, PC games by genre, ranked by unit
share from April 1998-March 1999 were comprised of
Strategy 21.8%, Simulation 13.4%, Adventure/role
playing 12.1% and Action 11.4%. 2
Even traditional flight simulation companies are taking
advantage of the emergence of commercial game
software for training. For example, Flight Safety
International re-markets a version of Microsoft Flight
Simulator and the Navy is experimenting with the game
for new pilot training.
What’s Wrong with Military VR?
Until recently, the military has led the way in developing
advanced virtual environments. We know the importance
of experiential learning through the development and use
of the National Training Center, Conduct of Fire
Trainers, Simnet, and flight simulators. The vision of the
military VR community has been to develop realistic
Weinberg was a key member in the development of the ICT
virtual environments to support training, mission
rehearsal, concept exploration and engineering design.
However, military simulations currently fall short of
enabling this vision of realism for a multitude of reasons.
First, the necessary technology does not yet exist, and
must be created. Our ability to immerse participants is
quite limited. For example, with respect to physical
immersion, it is currently possible to provide good
auditory, moderate visual, and primitive tactile/haptics
while essentially no olfactory or gustatory immersion is
possible. The ability to track full body motion, gesture
and expression is still nascent while virtual mobility is
limited to primitive two-dimensional approaches.
What technologies do exist for physical immersion tend
to be neither portable nor wireless. They also have
interoperability problems, fail to scale well to large
numbers of entities and have latency problems when it
comes to closely coupled interactions over long
distances. Defining (modeling), organizing and
distributing multimedia content also can be a problem.
Second, the stories and characters used in military
simulations are skeletal and rudimentary. A typical story
consists of a background briefing plus an event list. A
typical character is defined in terms of a role and a set of
scripted behaviors. Some degree of intellectual
immersion, to the extent of triggering some of the same
key decision making tasks that would occur in the real
world, is possible with such minimal story and character
definitions. However, rich story and engaging characters
can more fully engross the participant and provide a
more appropriate context for intellectual activity. (Note
that for peacekeeping training the US Army often hires
actors for live exercises at its Combat Training Centers.)
Lack of rich story and character also impairs emotional
immersion, as abstractions do not generally induce
intense emotions. Because emotions are powerful motivators, and can lead to significant shifts in both how the
world is interpreted and how decisions are made — in
the extreme, it can be a matter of decision making in a
life-or-death situation — this lack of emotional immersion is a major gap in making realistic simulations.
Emotional immersion is a particular strength of the
entertainment industry.
Third, the full set of necessary people to solve these
problems has been incomplete. Technical personnel
working with domain experts currently build military
simulations. This collaboration is critical, but creative
personnel  such as writers and cinematographers 
need to be added to the mix. The further advantages of
such a combination are that technical advances can open
up new creative realms, creative needs can drive new
research, and creative techniques can mask limitations in
Early in 1999, US Army leaders recognized a need for a
major transformation of our forces and the limitations of
our current simulation technologies. Furthermore, this
transformation required the ability to develop new
training and simulation systems for future conflicts that
leveraged the capabilities of both the entertainment
industry and academia.
The US Army and Department of Defense selected the
University of Southern California (USC) as a strategic
partner in the development of the Institute for Creative
Technologies (ICT) because of its unique confluence of
scientific capabilities and Entertainment Industry
relationships necessary for leadership in simulation.
The prime objective, as reaffirmed by Dr. Michael
Andrews, Deputy Assistant Secretary of the Army for
Research and Technology, was to build a special
partnership with the entertainment industry and
academia. Furthermore it was to advance the state of the
Army's technology and transition it quickly to programs
such as the Future Combat System.
A University Affiliated Research Center (UARC) is a
strategic relationship, requiring both breadth and depth
in capabilities matched with industry partnership to
achieve major advancements in science and technology.
This model of research is not new. For example The
National Automotive Center (NAC) serves as the Army's
focal point for the development of dual-needs/dual-use
automotive technologies and their application to military
ground vehicles. The NAC identifies the common needs
of the Defense Department, automotive industry and
academia for the purpose of collaborative research and
Part of USC’s uniqueness arises from its location in Los
Angeles, at the hub of both the entertainment and
aerospace industries; part arises from its standing as a
leading private research university; and part arises from
the capabilities and stature of its component units, and
the working relationships they have developed with
USC’s top-ranked School of Cinema-Television grew up
with the entertainment industry and continues to
maintain uniquely close ties with it. USC’s School of
Engineering is ranked 12th in the nation. Its Information
Sciences Institute is home to leading academic research
groups in networking and artificial intelligence. USC's
top-ten (and in some rankings, top-five) ranked
Annenberg School for Communication leverages off of
the Los Angeles area's varied strengths in new
technology, telecommunications, film, television, radio,
newspapers and magazines, and policy and research
The Institute for Creative Technology
USC established ICT under the auspices of the US Army
Simulation, Training, and Instrumentation Command
(STRICOM) to focus on developing the art and
technology for synthetic experiences that are so
compelling participants will react as if they are real. That
is, ICT will bring verisimilitude — the quality or state of
appearing to be true — to synthetic experiences. This
will produce a revolution in how the military trains and
how it rehearses for upcoming missions; just to name
two quite specific, but highly critical, military needs.
However, more generally, it will provide a quantum leap
in helping the Army prepare for the world, soldier,
organization, weaponry, and mission of the future.
Beyond the military, ICT will also advance a compelling
new medium for (at least) entertainment, education, arts,
and travel.
From the start, ICT leverages heavily off of this dual-use
nature by actively engaging the Entertainment Industry
(comprising film, TV, interactive gaming, etc.) and
possibly other industries later. ICT will serve as a means
for the military to learn about, and benefit from, the
technologies that are being developed in the
Entertainment Industry, and for transferring technologies
from the Entertainment Industry into the military. ICT
will also work with creative talent from the
Entertainment Industry in order to adapt their concepts
of story and character to increasing the degree of
immersion experienced by participants in synthetic
experiences, and to improving the utility of the outcomes
of these experiences.
ICT will pursue a combination of basic and applied
research (plus some educational activities). Basic
research will cover six thrusts crucial to the kind of
verisimilitude that is the institute’s mission [4]:
1. Immersion — Providing compellingly realistic
2. Networking and Databases — Organizing, storing
and distributing content
3. Story — Providing compelling interactive narratives
that propel experiences
4. Characters — Replacing human participants with
automated ones
5. Setup — Authoring and initializing environments,
models and experiences
6. Direction
understanding experiences
Applied research will be organized around a small
number of long-term themes; for example, simulating
futuristic style forces. Within each theme, a set of key
projects will be identified, along with an integration
architecture that will eventually bring them all together
in a single system covering the theme. Projects will be
pursued via sequences of prototypes of increasing
functionality and level of integration. The Army and the
Entertainment Industry will be actively involved at each
step in helping to ensure that what is done meets their
Key elements associated with USC’s array of relevant
existing capabilities include:
• The Entertainment Technology Center, (ETC) which
is a research and development project of the School
of Cinema-Television. ETC’s mission is to discover,
research, develop and accelerate entertainment
technology. Steven Spielberg and George Lucas sit
on the ETC board.
• The Annenberg Center for Communication that
technologies through interdisciplinary research and
• The Integrated Media Systems Center, (IMSC), a
National Science Foundation (NSF) established
center providing multi-media technologies. USC
successfully outbid 117 other university competitors
in response to the 1996 NSF national competition for
an integrated media center.
• The Information Sciences Institute, which combines
world class research and development across a broad
range of computer science and engineering with a
strong relationship with the Department of Defense.
ICT Vision
The vision for the ICT is to develop the art and
technology for synthetic experiences that are so
compelling participants will react as if they are real.
Participants will be fully immersed physically,
intellectually, and emotionally. They will be capable of
full three-dimensional mobility. Their behavior will be
propelled through engrossing stories stocked with
engaging characters that may be either automated or
manned — the high quality of the automated characters
along with the provision of plug compatibility will make
it impossible to distinguish. They will interact with the
experiences as if they are real. In short, the ICT will
provide a new meaning for “high fidelity”:
Imagine the soldier of the not so distant future. It is
Sunday and he is at home in Los Angeles. He and his
best friend in Hong Kong are relaxing by immersing
themselves in the nostalgic world of the 1990s. They are
founding an Internet startup company during the heyday
of the speculative bubble, learning to deal with venture
capitalists, trying to fend off large predatory rivals, and
ultimately trying to steer their new company towards a
successful Initial Public Offering. However, just when
the story is getting really engrossing, a high priority
videomessage arrives from his commanding officer with
the news that he will be shipping out within a few days,
along with the five thousand or so other members of his
Strike Force.
The mission will be to help keep the peace in the latest
global hot spot, but there are not yet any details
concerning his unit’s specific mission or the volatile
situation that currently exists on the ground there. He
also knows nothing about the country’s history, culture
or language. Fortunately he has a long flight ahead of
him, and the Army is ready for him.
STEVE is an intelligent tutor developed by USC/ISI
for the Office of Naval Research. [6]
He begins the flight with a brief on-line course covering
the history and culture of the region. A virtual tutor helps
him make the best possible use of the very limited time
he has available. (See figure). He then dons his personal
immersion system and walks into a simulated market in
the capital city, where a helpful (computer generated)
shopkeeper introduces him to the basic aspects of the
language along with the range of interpersonal
interaction styles — both positive and negative —
common to the culture.
Next, he is briefed by his commanding officer on his
unit’s mission — to keep innocent civilians from being
hurt in factional violence while preventing, as much as
possible, new flare ups among the factions. By sharing
an immersive space with his commander and the rest of
his unit — even though in reality they are physically
dispersed across several transport aircraft — he is able to
join them for a quick tour of their area of responsibility,
followed by a session in which they are able to
familiarize themselves with the uniforms and weapons
used by the various factions. He can pick up the
uniforms and examine them as well as see them on
various models. He can try out the weapons himself, as
well as pull up specs and performance numbers on them.
At all times he can discuss what he sees and does with
his commander and the other members of his unit.
During his final few hours he is immersed in a sample
mission. The sights, sounds and smells of the city
immediately bombard him. There are people everywhere
going about their lives as best they can. He’s a bit scared
and hesitant at first, but fortunately the rest of his unit is
there in the street with him. There’s a second unit
nearby, however he is unaware that they — along with
all of the citizens with whom he is interacting — are
computer-generated characters.
He is in a large central plaza in the city. A bazaar is
located in one part of the plaza and throngs of people are
milling about bartering for various goods. The plaza is
ringed by several government buildings and at the far
end there is a large church. The scene is a rich and
confusing tapestry of life — our soldier struggles to
remember the identifying features of the various factions
as he attempts to make sense of the scene. Suddenly,
near the church, a large disruption occurs and reports
ring out, echoing off the buildings. What is going on? Is
one of the rebel factions trying to attack the government?
Rifles at the ready, he and other members of his squad
rush toward the disturbance, where they confront — a
wedding party leaving the church and a group of
celebrants setting off large firecrackers.
Switching the safety back on, he shoulders his rifle and
breathes a sigh of relief while a computer generated tutor
emphasizes the need to assess the situation before taking
action and points out that in this culture celebrations are
often accompanied by fireworks which can be mistaken
for gunfire. This kind of immediate feedback is enabled
through the use of computer agents as tutors. Because it
is provided in context, it can be much more effective
than an after action review, where there may be a
substantial delay between the exercise and the review.3
This scenario was orchestrated by the Director, another
computer agent that directs the behavior of the other
agents in the simulation and the environment. By
exercising control of these elements, the Director ensures
that the exercise follows the intended story line so that
the intended training goals can be achieved. In this case,
this scenario was intended to create a situation in which
the soldier would be confronted with an ambiguous but
potentially threatening situation where it would be
necessary to decide whether or not to act — and where
the wrong decision would have disastrous consequences.
Although the soldier in the exercise is free to make
choices, the Director manipulates the simulation so that
eventually he is forced to confront the intended dilemma,
thereby achieving the pedagogical goals for the
simulation. For example, if the soldier and his squad had
not noticed the initial disturbance, the wedding
celebration would have become louder and more
boisterous, until it could not be ignored. Furthermore,
the squad’s failure to recognize the disturbance in its
early stages would be an issue that the tutor would cover
during its in situ review of the exercise.
commonplace. Verisimilitude of this sort will require
combining the art of (interactive) storytelling with the art
and technology of transforming these stories into
compelling interactive experiences. It inherently
involves collaboration between the kinds of creative and
technical experts found in the entertainment industry and
the kinds of researchers and system builders found in the
academic, industrial and military R&D communities.
Fortunately, all of these necessary partners are either
already present at USC or linked closely with it.
We expect that by creating a true synthesis of art and
technology4 and of the capabilities of the entertainment
industry and the R&D community — all in service of
verisimilitude — military training and mission rehearsal
will be revolutionized by making it more effective in
terms of cost, time, the types of experiences that can be
trained or rehearsed, and the quality of the result. It will
also provide a new medium for entertainment, enabling
both individuals and groups to be fully immersed and
engaged in compelling experiences from their homes, or
wherever they happen to be located.
Beyond entertainment, verisimilitude will also provide
new media for (at least) both immersive distance
learning and the arts (particularly the performing arts). It
could also even support a new mode of virtual travel;
providing immersive presence in a remote location, and
augmenting the local populace (with whom direct
interaction may not be possible) with synthetic
characters with whom interaction is possible.
The computer and Internet revolutions have substantially
changed the direction of entertainment from delivery in a
mass medium such as television to a mass customized
experience via the Web and PC. However, the art of
entertainment still requires stories, characters and
direction to make the experience meaningful and
The US Army faces the same challenge of adapting to
the changes brought about through the mass marketing
of supercomputing (e.g. Playstation 2), low-cost
graphics, and the higher expectations of technically
savvy soldiers.
Moreover, a more fundamental need is to represent new
kinds of problems such as urban conflict, operations
other than war, and information operations that cannot
be simulated well in military virtual environments today.
As the vignette presented above demonstrated, there is
an urgent requirement to represent the human
dimensions of war and conflict to provide training for
the truly difficult decision-making problems our soldiers
This is just one of many possible examples of the kind of
experience that ICT will make possible and, in fact,
This vignette was partly developed by William Swartout, the
Technical Director for ICT.
Providing what Richard Lindheim, the Executive Director of
ICT, has referred to as Show Technology as a complement to
the more common combination of art and business as Show
must face. NATO’s experience in Kosovo is now a
common one for countries such as the United Kingdom
(e.g. Northern Ireland).
The establishment of the Institute for Creative
Technologies is just one of many steps needed to
providing the essence of verisimilitude into training and
virtual reality systems. The US Army will explore all
avenues of entertainment technology to keep pace with
the challenges presented to us, whether in application to
distributed learning or embedded training systems.
Ultimately, we want to prepare our soldiers for the future
by experiencing it.
[1] Bran Ferren, Some Observations on the Future of
Army Simulation, RDA Magazine, May 1999, US
Army, Washington, DC, pp. 31-37.
[2] Michael Macedonia, Why Digital Entertainment
Drives the Need for Speed, IEEE Computer,
February 2000, pp. 124-127.
[3] Michael Macedonia, Sony versus Wintel: Mortal
Combat, IEEE Computer, July 1999, p. 112.
[4] Michael Zyda, ed., Modeling and Simulation:
Linking Entertainment and Defense, National
Research Council, National Academy of Sciences
Press, Washington, DC 1997.
[5] Nat Durlach and Anne Mavor, eds., Committee on
Virtual Reality Research and Development, Virtual
Reality: Scientific and Technological Challenges,
National Research Council, National Academy of
Sciences Press, Washington, DC 1994.
[6] J. Rickel and W.L. Johnson. Animated Agents for
Procedural Training in Virtual Reality: Perception,
Cognition, and Motor Control. Applied Artificial
Intelligence pp. 13:343-382, 1999.
Paul Rosenbloom is currently Professor in the Computer
Science Department at the University of Southern
California (USC), New Directions at the Information
Sciences Institute (USC/ISI), and Deputy Director of the
Intelligent Systems Division at USC/ISI. Prior to coming
to USC in 1987, was an Assistant Professor of Computer
Science and Psychology at Stanford University from
1984 to1987, and a Research Computer Scientist at
Carnegie Mellon University from 1983 to 1984.
Received a B.S. degree in Mathematical Sciences from
Stanford University in 1976 and M.S. and Ph.D. degrees
in Computer Science from Carnegie Mellon University
in 1978 and 1983, respectively.
Michael Macedonia is chief scientist and technical
director of the US Army Simulation, Training, and
Instrumentation Command, Orlando, Fla. A graduate of
West Point, Macedonia served as an infantry officer in a
number of United States and overseas assignments
including Germany and the Middle East. He is a veteran
of Desert Storm. Following his military service,
Macedonia became the Vice-president of the non-profit
Fraunhofer Center for Research in Computer Graphics,
Inc. (CRCG) in Providence, Rhode Island. Macedonia
then joined the Institute for Defense Analyses in
Alexandria, Virginia. His M.S. is in Telecommunications
from the University of Pittsburgh and he has a Ph.D.
degree in Computer Science from the Naval
Postgraduate School.
Information on ICT
For more information on ICT, please contact the
Program Manager, Dr. James Blake, at STRICOM. His
email is [email protected] or +1 407 3843502.
1. Recipient’s Reference
2. Originator’s References
3. Further Reference
5. Originator
4. Security Classification
of Document
ISBN 92-837-1057-6
Research and Technology Organization
North Atlantic Treaty Organization
BP 25, 7 rue Ancelle, F-92201 Neuilly-sur-Seine Cedex, France
6. Title
What Is Essential for Virtual Reality Systems to Meet Military Human Performance
7. Presented at/sponsored by
the Workshop of the RTO Human Factors and Medicine Panel (HFM) held in
The Hague, The Netherlands, 13-15 April 2000.
8. Author(s)/Editor(s)
9. Date
March 2001
10. Author’s/Editor’s Address
11. Pages
12. Distribution Statement
There are no restrictions on the distribution of this document.
Information about the availability of this and other RTO
unclassified publications is given on the back cover.
13. Keywords/Descriptors
Military training
Human factors engineering
User needs
Research projects
Motion sickness
Man machine systems
Virtual reality
Virtual environments
14. Abstract
This workshop aimed to identify the functional requirements of potential military applications
of Virtual Reality (VR) technology, to report the state-of-the-art and projected capabilities of
VR technologies, and to propose future research requirements and directions for military
During the workshop discussions, forty participants from military organisations, academia and
industry put forward their opinions on the significant bottlenecks and opportunities in the
development of military VR applications. Presentations discussed visual, haptic, auditory and
motion feedback, navigation interfaces, and scenario generation, modelling software and
rendering hardware.
VR research transition opportunities include the domains of training, planning & mission
rehearsal, simulation supported operation, remotely operated systems and product design.
Critical bottlenecks are a lack of natural interfaces, a lack of technology standardisation and a
lack of behavioural models and team interaction tools in VR.
In general, better co-ordination between military organisations, industry and academia is
necessary in order to identify gaps in current knowledge and to co-ordinate research.
Suggestions for closing gaps are included.
This page has been deliberately left blank
Page intentionnellement blanche
Télécopie 0(1) • E-mail [email protected]
L’Organisation pour la recherche et la technologie de l’OTAN (RTO), détient un stock limité de certaines de ses publications récentes, ainsi
que de celles de l’ancien AGARD (Groupe consultatif pour la recherche et les réalisations aérospatiales de l’OTAN). Celles-ci pourront
éventuellement être obtenues sous forme de copie papier. Pour de plus amples renseignements concernant l’achat de ces ouvrages,
adressez-vous par lettre ou par télécopie à l’adresse indiquée ci-dessus. Veuillez ne pas téléphoner.
Des exemplaires supplémentaires peuvent parfois être obtenus auprès des centres nationaux de distribution indiqués ci-dessous. Si vous
souhaitez recevoir toutes les publications de la RTO, ou simplement celles qui concernent certains Panels, vous pouvez demander d’être
inclus sur la liste d’envoi de l’un de ces centres.
Les publications de la RTO et de l’AGARD sont en vente auprès des agences de vente indiquées ci-dessous, sous forme de photocopie ou
de microfiche. Certains originaux peuvent également être obtenus auprès de CASI.
Streitkräfteamt / Abteilung III
Fachinformationszentrum der
Bundeswehr, (FIZBw)
Friedrich-Ebert-Allee 34
D-53113 Bonn
O.N.E.R.A. (ISP)
29, Avenue de la Division Leclerc
BP 72, 92322 Châtillon Cedex
P.O. Box 20701
2500 ES Den Haag
Coordinateur RTO - VSL/RTO
Etat-Major de la Force Aérienne
Quartier Reine Elisabeth
Rue d’Evère, B-1140 Bruxelles
GRECE (Correspondant)
Hellenic Ministry of National
Defence Industry Research &
Technology General Directorate
Technological R&D Directorate
D.Soutsou 40, GR-11521, Athens
Chief of International Cooperation
Research & Development Department
218 Niepodleglosci Av.
00-911 Warsaw
Directeur - Recherche et développement Communications et gestion de
l’information - DRDCGI 3
Ministère de la Défense nationale
Ottawa, Ontario K1A 0K2
Department for Scientific
Institute of Military Technology
Ministry of Defence
H-1525 Budapest P O Box 26
Estado Maior da Força Aérea
SDFA - Centro de Documentação
P-2720 Amadora
Danish Defence Research Establishment
Ryvangs Allé 1, P.O. Box 2715
DK-2100 Copenhagen Ø
Director of Aviation
c/o Flugrad
Distribucnı́ a informacnı́ stredisko R&T
VTÚL a PVO Praha
Mladoboleslavská ul.
197 06 Praha 9-Kbely AFB
INTA (RTO/AGARD Publications)
Carretera de Torrejón a Ajalvir, Pk.4
28850 Torrejón de Ardoz - Madrid
Centro di Documentazione
Tecnico-Scientifica della Difesa
Via XX Settembre 123a
00187 Roma
Defence Research Information Centre
Kentigern House
65 Brown Street
Glasgow G2 8EX
Voir Belgique
Millı̂ Savunma Bas,kanli i (MSB)
ARGE Dairesi Bas,kanli i (MSB)
06650 Bakanliklar - Ankara
NASA Center for AeroSpace
Information (CASI)
Parkway Center
7121 Standard Drive
Hanover, MD 21076-1320
Norwegian Defence Research
Attn: Biblioteket
P.O. Box 25, NO-2007 Kjeller
NASA Center for AeroSpace
Information (CASI)
Parkway Center
7121 Standard Drive
Hanover, MD 21076-1320
The British Library Document
Supply Centre
Boston Spa, Wetherby
West Yorkshire LS23 7BQ
Canada Institute for Scientific and
Technical Information (CISTI)
National Research Council
Document Delivery
Montreal Road, Building M-55
Ottawa K1A 0S2, Canada
Les demandes de documents RTO ou AGARD doivent comporter la dénomination “RTO” ou “AGARD” selon le cas, suivie du
numéro de série (par exemple AGARD-AG-315). Des informations analogues, telles que le titre et la date de publication sont
souhaitables. Des références bibliographiques complètes ainsi que des résumés des publications RTO et AGARD figurent dans les
journaux suivants:
Scientific and Technical Aerospace Reports (STAR)
Government Reports Announcements & Index (GRA&I)
STAR peut être consulté en ligne au localisateur de
publié par le National Technical Information Service
ressources uniformes (URL) suivant:
Virginia 2216
STAR est édité par CASI dans le cadre du programme
NASA d’information scientifique et technique (STI)
(accessible également en mode interactif dans la base de
données bibliographiques en ligne du NTIS, et sur CD-ROM)
STI Program Office, MS 157A
NASA Langley Research Center
Hampton, Virginia 23681-0001
Imprimé par St-Joseph Ottawa/Hull
(Membre de la Corporation St-Joseph)
45, boul. Sacré-Cœur, Hull (Québec), Canada J8X 1C6
Telefax 0(1) • E-mail [email protected]
NATO’s Research and Technology Organization (RTO) holds limited quantities of some of its recent publications and those of the former
AGARD (Advisory Group for Aerospace Research & Development of NATO), and these may be available for purchase in hard copy form.
For more information, write or send a telefax to the address given above. Please do not telephone.
Further copies are sometimes available from the National Distribution Centres listed below. If you wish to receive all RTO publications, or
just those relating to one or more specific RTO Panels, they may be willing to include you (or your organisation) in their distribution.
RTO and AGARD publications may be purchased from the Sales Agencies listed below, in photocopy or microfiche form. Original copies
of some publications may be available from CASI.
Coordinateur RTO - VSL/RTO
Etat-Major de la Force Aérienne
Quartier Reine Elisabeth
Rue d’Evère, B-1140 Bruxelles
Director Research & Development
Communications & Information
Management - DRDCIM 3
Dept of National Defence
Ottawa, Ontario K1A 0K2
Distribucnı́ a informacnı́ stredisko R&T
VTÚL a PVO Praha
Mladoboleslavská ul.
197 06 Praha 9-Kbely AFB
Danish Defence Research
Ryvangs Allé 1, P.O. Box 2715
DK-2100 Copenhagen Ø
O.N.E.R.A. (ISP)
29 Avenue de la Division Leclerc
BP 72, 92322 Châtillon Cedex
Streitkräfteamt / Abteilung III
Fachinformationszentrum der
Bundeswehr, (FIZBw)
Friedrich-Ebert-Allee 34
D-53113 Bonn
GREECE (Point of Contact)
Hellenic Ministry of National
Defence Industry Research &
Technology General Directorate
Technological R&D Directorate
D.Soutsou 40, GR-11521, Athens
Chief of International Cooperation
Research & Development
218 Niepodleglosci Av.
00-911 Warsaw
Department for Scientific
Institute of Military Technology
Ministry of Defence
H-1525 Budapest P O Box 26
Estado Maior da Força Aérea
SDFA - Centro de Documentação
P-2720 Amadora
Director of Aviation
c/o Flugrad
Centro di Documentazione
Tecnico-Scientifica della Difesa
Via XX Settembre 123a
00187 Roma
See Belgium
P.O. Box 20701
2500 ES Den Haag
Norwegian Defence Research
Attn: Biblioteket
P.O. Box 25, NO-2007 Kjeller
INTA (RTO/AGARD Publications)
Carretera de Torrejón a Ajalvir, Pk.4
28850 Torrejón de Ardoz - Madrid
Millı̂ Savunma Bas,kanli i (MSB)
ARGE Dairesi Bas,kanli i (MSB)
06650 Bakanliklar - Ankara
Defence Research Information
Kentigern House
65 Brown Street
Glasgow G2 8EX
NASA Center for AeroSpace
Information (CASI)
Parkway Center
7121 Standard Drive
Hanover, MD 21076-1320
NASA Center for AeroSpace
The British Library Document
Canada Institute for Scientific and
Information (CASI)
Supply Centre
Technical Information (CISTI)
Parkway Center
Boston Spa, Wetherby
National Research Council
7121 Standard Drive
West Yorkshire LS23 7BQ
Document Delivery
Hanover, MD 21076-1320
United Kingdom
Montreal Road, Building M-55
United States
Ottawa K1A 0S2, Canada
Requests for RTO or AGARD documents should include the word ‘RTO’ or ‘AGARD’, as appropriate, followed by the serial
number (for example AGARD-AG-315). Collateral information such as title and publication date is desirable. Full bibliographical
references and abstracts of RTO and AGARD publications are given in the following journals:
Scientific and Technical Aerospace Reports (STAR)
Government Reports Announcements & Index (GRA&I)
STAR is available on-line at the following uniform
published by the National Technical Information Service
resource locator:
Virginia 22161
STAR is published by CASI for the NASA Scientific
United States
and Technical Information (STI) Program
(also available online in the NTIS Bibliographic
STI Program Office, MS 157A
Database or on CD-ROM)
NASA Langley Research Center
Hampton, Virginia 23681-0001
United States
Printed by St. Joseph Ottawa/Hull
(A St. Joseph Corporation Company)
45 Sacré-Cœur Blvd., Hull (Québec), Canada J8X 1C6
ISBN 92-837-1057-6
Fly UP