See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/336679007

A Driving Simulator Based on Video Game Technology

Conference Paper · November 2006

CITATIONS READS
19 2,687

3 authors, including:

Mikael Lebram Henrik Engström

University of Skövde University of Skövde
34 PUBLICATIONS 682 CITATIONS 82 PUBLICATIONS 1,079 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Henrik Engström on 20 October 2019.

The user has requested enhancement of the downloaded file.

 A Driving Simulator Based on Video Game Technology
Mikael Lebram∗ Henrik Engström† Henrik Gustavsson‡
University of Skövde University of Skövde University of Skövde

Abstract is that it enables generation of extreme situations or traffic environ-

ments not available in the student’s surroundings.
This paper presents the design and architecture of a mid-range driv-
ing simulator developed at the University of Skövde. The aim is to Using simulators for drivers’ education is different from pilot edu-
use the simulator as a platform for studies of serious games. The cation in many ways. The volume of students involved in driving
usage of video game technology and software has been a central education is larger than that of pilot training. The cost associated
design principle. The core of the simulator is a complete car sur- with a driving training program is also much lower. This implies
rounded by seven screens. Each screen is handled by a standard that simulators for drivers’ education need to be less expensive in
PC, typically used for computer games, and the projection on the order to be widely adopted.
screens is handled by budget LCD-projectors. The use of consumer
electronics, standard game technology and limited motion feedback In the video game area, driving and racing has been a central theme
makes this simulator relatively inexpensive. In addition, the archi- for a long time. Since Night Driver (1976) [Atari 2006], the origi-
tecture is scalable and allows for using commercial video games in nal first-person racing game, hundreds of racing titles have been re-
the simulator. leased. In addition, car driving is a central activity in games of other
genres as well. Over the last 30 years driving games has gone from
Observations from a set of experiments conducted in the simula-
relatively simple simulations in arcade-machines to highly realis-
tor are presented in this paper. In these experiments driving school
tic rally simulations that runs on an off-the-shelf personal computer
students were instructed to freely explore a driving game specifi-
(Figure 1).
cally designed for the simulator platform. The result shows that the
level of realism is sufficient and that the entertainment value was
considered to be high. This opens the possibilities to employ and
use driving simulators for a wider set of applications. Our current
research focuses on its use with serious games for traffic education.

CR Categories: I.3.2 [Computing Methodologies]: Computer

Graphics—Graphics Systems; I.6.3 [Computing Methodologies]:
Simulation and Modeling—Applications

Keywords: driving simulator, virtual reality, computer games, se-

rious games
Figure 1: 30 years of racing games. Night Driver [Atari 2006] (left)
from 1976 and GTR II [SimBin 2006] (right) from 2006

1 Introduction
Over the years the graphical quality of computer games has in-
creased exponentially, the current level of detail of graphical mod-
The use of simulators for training is an old and well accepted
els is sufficient for most training simulator purposes. In addition,
method used in situations where training in real environments is
video game developers have high skills in producing entertaining,
difficult, dangerous and/or expensive. In particular, simulators for
immersive products that motivate their users to spend many hours
civil and military air pilot training are well established [Rolfe and
Staples 1988]. Computer based flight simulators have been used a week [ESA 2005]. The games are sold as consumer products in
since the 1960’s. In the light of this, the usage of car driving simu- large volumes which implies that the price is only a fraction of the
lators is less common. There exists advanced simulators for traffic price of a specialised simulator product.
safety research [Kuhl et al. 1995; ITS 2006; Östlund et al. 2006]
and there are some examples of simulators for driving education The goal for the simulator presented in this paper is to utilize the
[INRETS 2006], but for the vast majority of drivers the training is developments in the video game area to create an advanced driv-
solely conducted in real traffic environments. The potential advan- ing simulator using video game technology. This includes the use
tage of using driving simulators in, for example, traffic education of standard of-the shelf soft- and hardware infrastructure as well as
adaptation of commercial-of-the-shelf (COTS) games. The simu-
∗ e-mail: mikael.lebram@his.se lator is currently used to explore how serious games [LoPiccolo
† e-mail: henrik.engstrom@his.se 2004; Zyda 2005; Blackman 2005] can be developed and used in
‡ e-mail: henrik.gustavsson@his.se traffic education.

This paper is organized as follows: In Section 2 we give a back-

ground presentation of driving simulator technology. In Section 3
we present the architecture and design of our driving simulator fol-
lowed by Section 4 where we report experiences from using the
simulator in a set of experiments. In Section 5 we draw conclusions
and elaborate on the implications of our approach.

39
2 Background that are designed to produce a realistic driving experience. As an
example, the GTR racing game has been developed by a racing
team [SimBin 2006] with a goal to produce a highly realistic racing
Driving simulators are developed and used for research purposes experience. For example, the simulation is so good that it is used by
mainly within the traffic safety context. The French national insti- racing drivers to memorize courses [Björklund 2006]. The cost of
tute for transport and safety research presents a survey of simula- a driving simulator composed of a PC, some CRT:s, a racing wheel
tors [INRETS 2006] including 50 simulators for research purposes, and a game is typically less than 1 000 Euro.
a handful for corporate purposes (e.g. car industry) and 20 driving
training simulators. The research simulators are generally in the
high-end with large budgets while the training simulators are typi-
cally low-range products targeted at driving schools. 3 The Driving Simulator
A driving simulator is composed of a number of components, illus-
trated in Figure 2. The goal of the driving simulator developed at the University of
Skövde is to utilize the cost-effective and entertaining aspects from
gaming simulators. The system is composed of 8 standard game
Graphical environment PCs - 7 clients and a server. Each client is connected to a budget
LCD-projector projecting on screens surrounding a real car. As the
simulator is intended to be used for video games, the requirements
on realism are somewhat different compared to high-end simulators
Computer architecture used for traffic safety research. For games there is always an enter-
tainment requirement that has to be considered. By using of-the-
shelf hardware components it is possible to utilize game software
and technology. We have successfully modified a number of COTS
games to run on the simulator platform. In this section we present
the architecture and design of the simulator.

Software environment
3.1 The driver environment
Physical feedback systems
The driver environment is a complete Volvo S80 with authentic con-
The driver environment trols and instrumentation. Figure 3 shows the driver environment.

Figure 2: Components of a driving simulator

The driver environment may be more or less realistic ranging from

an authentic complete car to a steering wheel (or even mouse and
keyboard). The graphical environment renders the simulated world
to the driver. It may differ in graphical quality, the size and range of
the projection as well as the technology used to generate the view
(CRT displays, 2D projection, 3D shutter-glasses etc.). The phys-
ical feedback system is responsible for generating other inputs to
the driver, such as sound, motion and other haptic feedback. The
computer architecture ties the various systems together and consti-
tutes the platform for generating and monitoring the simulation by
the software. The software available for a simulator is governed by
the nature of the underlying systems. Specialised software may be
needed if the underlying hardware is tailor made.
The most notable difference between high-end and low-end sim- Figure 3: The driver environment
ulators is the physical feedback systems. The research simulators
have large mechanical systems that generates g-forces in different The use of a real car provides a great deal of realism to the simu-
directions. As an example, the Leeds Advanced Driving Simulator lator. Users of the system have no problems in understanding the
[ITS 2006] is a 4 million Euro project with a spherical projection functionality of the interface to the simulator. In addition the feeling
dome. The driver environment and the dome are appended to a mo- of being inside a car is a familiar situation which, for most people,
tion base that has 8 degrees of freedom. The simulator is hosted in is associated with a responsibility for the car and fellow road users.
a 14x12x7 meter hall. As a contrast S-4150, a basic training sim- This brings a sense of seriousness to the driving.
ulator from Simulator Systems International [SSI 2006], consists
of a steering wheel and a CRT display in front of the driver. The
simulator has clutch and brake and no physical feedback system. 3.2 Graphical environment
The cost of S-4150 is approximately 5 000 Euro.
In addition to the simulators mentioned above there exists a large The graphics generated in the simulator is projected on seven flat
number of ”driving simulators” in the shape of video games. We screens as illustrated in Figure 4. These screens cover the whole
refer to these as gaming simulators. A gaming simulator may be an field-of-view for the driver and the parts covered by the rear view
arcade machine designed to give an entertaining experience and not mirrors. In the design phase an alternative solution, to back-project
necessarily a realistic one. There however exists gaming simulators the graphics directly on the windshield, was rejected for a number

40
of reasons: Firstly this solution gives a very closed and flat projec- ternal surround system of the car. In addition a ”ButtKicker” [Gui-
tion where the external parts of the car are not visible. It is also tammer 2005] is used to generate vibrations in the body of the car
not possible for the driver or passengers to move their heads to get which are propagated to the whole car including the steering wheel.
a different perspective on the surroundings (e.g. if the windshield One important property relating to physical feedback in a driving
post is covering some part of the view). Another aspect to consider simulator is the haptics of the steering wheel. In a car with servo-
is the distance from the observer to the projected screen. The sim- steering there is not as much movements as when there is a direct
ulations are generated using window-projection [Cruz-Neira et al. connection between the steering wheel and the tyres. The most im-
1992] that is computed from the perspective of the driver which portant remaining physical property is that the wheels should strive
means the passengers will experience a distortion. With the chosen to return to their original position. In the simulator this has been
solution the distance to the screens is greater and the distortion for achieved by placing each front wheel on an axial ball bearing. Due
passengers is acceptable. An additional advantage by using screens to the steering axis inclination there will be a strive to return the
outside the car is that it enables the original rear mirrors to be used wheels to a parallel position. In addition, the movements of the
and it is possible for the driver to turn his head and look in the front wheels gives a notable movement of the car that can be con-
rear window. The latter is not possible when, for example, LCD- sidered to be a form of passive physical feedback.
displays are used as rear mirrors.
The physical feedback component that possibly contributes most to
the perception of speed in the simulator is the internal fan. It is
controlled by the simulation and the force of the fan is linear to the
speed of the car [Carraro et al. 1998]. When the driver is reaching
a high speed the wind and the substantial noise from the fan con-
tributes to create a high speed perception. It is well known that it
is difficult to get a good perception of speed in computer generated
simulations [Godley and Fildes ; Östlund et al. 2006]. The use of
a fan is a simple but effective way to increase the perceived speed.

3.4 Computer Architecture

The computer architecture in the simulator consists of 8 standard

gaming PCs equipped with a mid-range graphics card. One com-
puter is acting as server while the other are clients each responsible
for one screen. The clients and server are typically running iden-
tical simulations with the only difference that the server is send-
ing synchronize messages to the clients. The clients differ only in
the camera position used when rendering. The computers are con-
nected in an Ethernet LAN. All hardware components are standard
consumer products. The only tailored component of the simulator
is the interface with the car [Mine 1995]. The movements of the
steering wheel and other controls are monitored by microcontrollers
that communicate with the server via a USB game control protocol.
Figure 4: The simulator car surrounded by screens (1-7) In this way the car can be seen as a highly specialised joystick. The
advantage with this approach is that the simulator can be used with
The projection on the screens is similar to that used in a Cave any computer game that supports joysticks.
[Cruz-Neira et al. 1992]. The choice of rectilinear projection in-
stead of cylindrical or spherical is mainly economical. Each screen
is handled by a budget LCD-projector and as the screens are not
3.5 Software Environment
projected seamlessly there are low requirements on the calibration.
This also makes it possible to use a large number of screens and
hence cover a larger fraction of the field-of-view, than is common As mentioned above, almost any computer game can be played on
in mid-range simulators. The forward visual field-of-view is 220 by the simulator platform. In most situations it will however be lim-
30 degrees, and 60 by 30 degrees in the rear direction. As a com- ited to use only one screen. To utilize all 7 screens the software
parison, the high-end simulator used in [Peters and Östlund 2005] has to support multiple clients with adjustable camera positioning.
has a forward visual field-of-view of 120 by 30. The extensions required are hence very small and we have success-
fully managed to adjust several commercial games to be used in the
simulator using multiple screens. In addition to using COTS games
3.3 Physical feedback systems we have also developed an infrastructure based on an open source
game engine. This allows for custom made simulation application
and games.
The generation of physical feedback in a driving simulator may
be extremely complex. The simulator at the university of Skövde
adopts a fixed-based approach which means that no g-forces are
generated. This is in total contrast with the mid-range simulator 4 Experiments
presented by Huang and Chen [Huang and Chihsiuh 2003] which
emphasises on the motion system in favour of the graphical system
and the driver environment. The simulator has been used in an experimental study with 24 driv-
ing school students as subjects. The gaming background of the stu-
The illusion of movement in our simulator is generated by the use of dents varied from inexperienced (13) to experienced players (5).
sound, vibrations and the car’s fan. The sound is generated in the in- The experimental setup was such that the subjects were offered as

41
much time they wanted (up to a maximum of 30 minutes) playing tion only considers the relative position within the lane (irrespective
and exploring a game. This was followed by a number of evaluative of what lane the car was in). The tails in the histogram are due to
tests where they were instructed to perform certain tasks, followed lane changes and the gameplay is such that frequent lane-changes
by a questionnaire. All simulations were monitored and logged. are required to succeed. In fact, the drivers changed lane on aver-
age every 15 seconds. The most notable property of the histogram
The game used in these experiments is relatively simple. The player is the large bar in the middle. Despite the frequent lane changes the
is driving on a five-lane motorway following an ambulance. The drivers spent almost 25% of the time exactly centred in the middle
difficulty of the game increases by the intensity of the traffic and the of the lane. The central bar is moreover more than double the size
behaviour of fellow road users. Although we had other main goals of the surrounding bars. We interpret this as the drivers have intu-
with these experiments, they have also provided some feedback on itively managed to position the car very close to the centre of the
the performance of the simulation environment. lane. This implies that the visual representation gives a realistic im-
First of all, the subjects where extremely positive concerning the pression of the position of the car. The rectilinear projection hence
entertainment value of the simulator. In the questionnaire subjects seems to work very well.
were asked to specify how they agreed to the statement ”it was
The use of original rear view mirrors also seems to be efficient.
fun to drive”, on a 5-graded Likert scale where 1=fully disagree
The subjects’ use of the mirrors was monitored during experiments
and 5=fully agree. The average for all subjects where 4.6 which is
and the result shows that they used both the internal as well as the
a very high result considering the relative simplicity of the game.
external mirrors frequently. In fact, the use of mirrors was more
This result may also be derived from the amount of time the sub-
frequent than the lane-changing. On average the subjects used the
jects spent in the simulator. They were explicitly instructed to de-
mirrors every 10 seconds compared to 15 seconds for lane changes.
cide themselves when to stop driving. The result was that experi-
enced gamers spent on average 29 minutes playing the game com- Simulator sickness (also termed cybersickness) is a well known
pared to 23 minutes for inexperience players. This is a statistically problem in simulators and is related to motion sickness [Harm
significant difference which is interesting as one may suspect that 2002; AGARD 1988]. As much as 30% of the users of simula-
experienced gamers would not appreciate a game that is far from a tors may experience symptoms severe enough to discontinue use
state-of-the-art racing game. One possible interpretation is that the [Harm 2002]. Simulator sickness is believed to be caused by con-
simulator platform itself contributed to the positive experience, in fusion between the perceived motion and the actual motion [Bertin
particular for gamers. et al. 2004]. The problem seems to be difficult to totally eliminate,
Concerning the realism of the game and simulator the average of even for high-end simulators [Peters and Östlund 2005].
the subjects was 3.6 for the statement ”the driving was realistic” Since the simulator presented in this paper is a fixed-based system,
(using a 5-gradet Likert scale). This is clearly above average which problems with simulator sickness was not unexpected. These prob-
indicates that the simulator is efficient. Some users commented on lems were however minor in the experiments. Four subjects (17%)
initial problems with the control of the car. These initial problems reported sickness as one of the reasons they decided to stop playing
do not seem to have had any negative impact on the overall expe- the game. The average playing time of these four subjects was 21
rience and performance of the drivers. This can be confirmed from minutes compared to 25 minutes for those that did not report any
analysis of how the drivers managed to position their car in the lane sickness problems. The relatively small difference in time makes
(lateral position). Lateral position is commonly used for validation us believe these subjects did not experience severe problems with
of driving simulators [Green 2005; de Waard et al. 2005]. Fig- simulator sickness.
ure 5, illustrates the relative lateral position of the car during all
experiments.

30 5 Discussion
25
In this paper we have presented a driving simulator based on video
game technology. Our approach has been to use relatively inex-
Percentage of time

20
pensive hardware components to create a graphical system that sur-
15 rounds a real car whose instrumentation has been adopted to be used
as a game control. A main difference to high-end simulators is the
10 modest physical feedback system. The presented simulator uses
a fan, vibrations and sound in addition to the graphical feedback.
5 The driving simulator has successfully been used in an experimen-
tal study. Observations from this study indicate that the simulator is
0 efficient in that it creates a realistic and entertaining experience to
-1

-0.5

0.5

the users. The absence of physical feedback does not seem to incur
Relative position
serious problems with simulator sickness. In addition the rectilinear
projection gives a realistic perception of the simulated environment.
This has been shown by analysing the lateral positioning of the car.
Figure 5: Histogram of the relative lateral position of the car
When developing a simulator one goal is to create a realistic experi-
The total driving time for all subjects was almost 12 hours. The ence. Realism comes to a price and with a limited budget the benefit
position of the car was sampled at 10Hz. The histogram in Figure 5 has to be balanced with its price. In our approach we have decided
is based on all logs from all experiments (413 973 samples). The to sacrifice the physical movement realism in favour of the realism
relative position of the car in the lane was divided into 21 discrete of having a real car as the driver environment. We believe that the
intervals. The histogram was created by summarizing the number use of a real car is one of the key benefits of the presented simula-
of times the car was positioned in respective interval and then divide tor. The smell and touch of a car gets the driver in the mind-set of
it with the total number of samples. Note that the recorded informa- driving. In addition, our simulator allows for passengers, which is

42
 a typical driving property which is neglected in many other simula- DE WAARD , D., S TEYVERS , F., AND B ROOKHUIS , K. 2005. How
tors. The driving task is, for example, much harder to handle when much visual road information is needed to drive safely and com-
there are two fighting children in the backseat. fortably? Safety Science 42, 7, 639–655.
The simulator architecture presented in this paper is flexible and ESA, 2005. Entertainment software association -
scalable. For example, the number of screens used can easily be game player data. Retrieved 2005-04-13, from
extended by adding a projector and a PC for each screen. The server http://www.theesa.com/facts/gamer data.php.
can broadcast messages to all involved clients which mean that the
total load of the system is in practice independent of the number of G ODLEY, S. T., AND F ILDES , B. N. Driving simulator validation
screens. Each PC handles the rendering of one screen which differs for speed research. Accident analysis & prevention 34.
only in their camera positioning. The flexibility of the architecture G REEN , P., 2005. How driving simulator data quality can be im-
is illustrated by the fact that we have successfully modified several proved. DSC 2005, Orlando.
COTS games to be playable on the simulator platform. The ability
to adopt and use commercial software is important as the cost of G UITAMMER, 2005. Buttkicker. Retrieved 2006-10-02, from
software development may be huge. In future studies we plan to http://thebuttkicker.com/.
use COTS racing games whose graphical quality require budgets H ARM , D., 2002. Motion sickness neurophysiology, physiological
way beyond that of the simulator hardware. correlates, and treatment. in Handbook of Virtual Environments,
To summarize, the contribution of the presented work is that we Stanney (edt.).
have combined the quality and cost-effectiveness of the gaming H UANG , A., AND C HIHSIUH , C. 2003. A low-cost driving sim-
technology with the extensiveness of the mid-range to high-end ulator for full vehicle dynamics simulation. IEEE Transactions
simulators. We estimate that the hardware cost of the presented on Vehicular Technology 52, 1, 162–172.
simulator is less than 20 000 Euro excluding the cost of the car.
INRETS, 2006. Driving simulators. Retrieved 2006-10-02, from
In our ongoing and future research we will use the simulator to http://www.inrets.fr/ur/sara/Pg simus e.html.
explore the potential benefit of using computer games in traffic ed-
ucation. We will test whether a serious driving game designed with ITS, 2006. Leeds advanced driving simulator. Retrieved 2006-10-
the specific purpose of enhancing certain traffic safety variables is 02, from http://www.its.leeds.ac.uk/facilities/lads/.
effective. The general idea is to combine the strengths of traditional
K UHL , J., E VANS , D., PAPELIS , Y., ROMANO , R., AND WAT-
simulators with the fun of games.
SON , G. 1995. The iowa driving simulator: An immersive re-
search environment. Computer 28, 7, 35–41.

Acknowledgements L O P ICCOLO , P. 2004. Serious games. Computer Graphics World

27, 2.

This work has been sponsored by Länsförsäkringsbolagens Forskn- M INE , M. R. 1995. Virtual environment interaction techniques.
ingsfond, Skövde Kommun, Tillväxt Skaraborg, and Volvo Cars. Tech. rep., Chapel Hill, NC, USA.
Ö STLUND , J., N ILSSON , L., T ÖRNROS , J., AND F ORSMAN , A.,
2006. Effects of cognitive and visual load in real and simulated
References driving. VTI rapport 533A-2006, Swedish Road and Transport
Research Institute (VTI), Linköping, Sweden.

AGARD, 1988. Motion cues in flight simulation and simulator P ETERS , B., AND Ö STLUND , J., 2005. Joystick controlled driving
induced sickness. Advisory Group for Aerospace Research & for drivers with disabilities. VTI rapport 506A-2005, Swedish
Development, Conference proceedings. Road and Transport Research Institute (VTI), Linköping, Swe-
den.
ATARI, 2006. Atari inc. Retrieved 2006-9-23, from
http://www.atari.com. ROLFE , J., AND S TAPLES , K., Eds. 1988. Flight Simulation. Cam-
bridge University Press.
B ERTIN , R., G UILLOT, A., C OLLET, C., V IENNE , F., E SPI É ,
S., AND G RAF, W., 2004. Objective measurement of simula- S IM B IN, 2006. Simbin development team. Retrieved 2006-9-23,
tor sickness and the role of visual-vestibular conflict situations: from http://www.simbin.se.
a study with vestibular-loss (a-reflexive) subjects. Neuroscience, SSI, 2006. Simulator systems international. Retrieved 2006-10-02,
San Diego, California. from http://www.simulatorsystems.com.
B J ÖRKLUND , P., 2006. Chief technical officer SimBin develop- Z YDA , M. 2005. From visual simulation to virtual reality to games.
ment team AB. Personal Communication. Computer 38, 9, 25–32.
B LACKMAN , S. 2005. Serious games...and less! SIGGRAPH
Computer Graphics 39, 1, 12–16.
C ARRARO , G. U., C ORTES , M., E DMARK , J. T., AND E NSOR ,
J. R. 1998. The peloton bicycling simulator. In VRML ’98:
Proceedings of the third symposium on Virtual reality modeling
language, ACM Press, New York, NY, USA, 63–70.
C RUZ -N EIRA , C., S ANDIN , D. J., D E FANTI , T. A., K ENYON ,
R. V., AND H ART, J. C. 1992. The cave: audio visual ex-
perience automatic virtual environment. Commun. ACM 35, 6,
64–72.

43

View publication stats