-
Overview of the UMTRI Simulator
UMTRI has a fixed-base driving simulator, that uses DriveSafety
software to schedule events to occur, create the associated scenes
(examples of which are below), and to sense and record driver reactions.
UMTRI's driving simulator currently runs version 1.6.2 of the DriveSafety
software. This is the third driving simulator UMTRI has had, with
the first 2 being internally developed starting in about 1990. The
DriveSafety software is found at most of the top tier universities
in the U.S., as well as at GM, Ford, Delphi, and Motorola.
|
|
Downtown
(Traffic, traffic signals are programmable) |
Rural with Fog |
|
|
Residential |
Semi-rural-industrial |
-
The simulator hardware and software is supplemented by a custom UMTRI-designed
audio-visual (AV) system. The DriveSafety simulator has been used
in studies of driving workload, in-vehicle devices (navigation systems,
cell phones), warning systems, and the effects of health on driving
(e.g., Alzheimer’s disease, driver age), and other topics.
The UMTRI Driving Simulator hardware consists of a cab, computers,
projectors, and miscellaneous AV equipment. To provide reliability,
all key simulator components have uninterruptible power supplies.
The cab is a full-size vehicle with a touch
screen center console, a unique, computer-controlled back-projected
LCD speedometer/tachometer cluster, operating foot controls, a torque
motor to provide realistic force feedback on the steering wheel, and
a haptic seat with 8 vibrating motors. The in-cab displays can be
controlled by Macintosh or Windows computers running RealBasic or
Visual Basic software. The haptic seat vibrators have been used in
studies of lane departure and forward collision warning. Beneath the
cab is a base shaker to provide Z-axis vibration.
 |
 |
-
Because several recent studies have concerned warning systems, audio
in the cab has received considerable attention. Sounds are presented
by a 10-speaker system from a Nissan Altima, supplemented by a 5-speaker
system for road sounds. The simulator software generates 3D audio,
including engine noise, wind noise, tire squeal, and other sounds.
There are also two speakers in the headrest of the driver’s
seat that have been used for warning sounds.
Road scenes are projected on 5 forward
screens (by 5 projectors) almost 16 feet from the driver (200 degree
field of view) and a rear channel 12 feet away (40 degree field of
view). Given the desire to simulate conditions in which far acuity
limits viewing the road scene, this is reasonably close to 20 feet,
and is greater than is the case in many academic simulators. Each
channel is 1024x768, is anti-aliased, and updates at 60 Hz. The center
channel will eventually be upgraded to a higher resolution. The wide
field of view is particular useful for studies of merging situations.
However, in some situations, to minimize opportunities for motion
discomfort, the side channels may be turned off. Graphics are generated
by 6 Linux computers (1/scene projector) whose actions are coordinated
by another Linux computer that in turn is controlled by a Windows
computer that provides the experimenter interface.
Simulated worlds are created using tiles (as in SimCity). There
are about 250 tiles in the library, including scenes from rural, urban,
residential, industrial, and expressway settings including intersections
with programmable traffic signals. All roads comply with AASHTO and
MUTCD standards. Scenes are currently daytime only, though bad weather
(fog, rain, snow) can be simulated. Below are some example images.
The following video (approx 4 minutes long) shows two situations typically
experienced in the simulator; an intersection, and highway driving.
The video shows multiple perspectives including inside the vehicle and
from the driver's perspective. It also has an overview showing the front
3 video channels from outside the vehicle. To view the video, Real Player 10.0 or greater must be installed on
your computer. To download the newest version of Real Player, please
click
HERE. Apple Users: Please hold the control button when clicking
the link above. This will save the file to the desktop and you can run
it from there.
-
Traffic is programmable, either following the general rules of the
road or as scripted (following at a certain distance, lead vehicle slowing
in front of the subject, side vehicles cutting off the subject, cars
pulling in front of the subject, etc). In some experiments, over 20
vehicles have been controlled, as well as several pedestrians. There
are many types of vehicles, including sedans, sports utility vehicles,
To accelerate scenario and experiment development,
UMTRI, with funding from GM, has developed software with a point and
click interface for developing scenarios for some expressways and urban
drives. That software allows for quickly setting what each object in
the world (each car and traffic light) does for each section road. Normally,
scenario development takes 1-3 months of engineering time to create
the desired world and write the needed scripts in TCL to control scenarios.
Where the UMTRI scenario tools can be used, development times should
be reduced to weeks.
 |
Typically, about 25 driver- and vehicle-performance measures (steering
wheel angle, throttle position, yaw angle, speed, lateral lane position,
time to collision, distance to the lead vehicle, etc.) are recorded at
up to 60 Hz by the main simulator computer. In addition, if there is an
in-vehicle task to perform such a entering a destination, keystrokes and
keystrokes times are recorded and time synched with the main simulator
computer. That data may be logged by an UMTRI-written simulation of a
real interface, or by sponsor-provided hardware and software. For sponsor-provided
systems, it is important to provide a TCP/IP port with access to user
input, which often is not the case. Typically, experiments generate 2.5Mb
of raw data to reduce and analyze per hour of subject time.
In addition, for some experiments, a Seeing Machines version 3.0 eye-fixation
system (2 cameras, 1 Windows computer) has been installed
in the driving simulator. This system generates a considerable amount
of data, and because of associated costs, is used selectively.
A particularly unique feature of the UMTRI driving simulator is the extensive
AV system for recording driver performance
and presenting video, of which the projectors to show road scenes are
a small part. Sample images follow. There are microphones in the cab to
capture drivers’ reactions, as well 7 video cameras (driver face,
hands, driver’s feet and foot controls, several views of the instrument
panel), as well as all projected images, the instrument panel and any
interior displays, making it possible to have a variety of views available
to the experimenter for recording on a quad-split image which is be recorded
onto a DVD for later reviewing. Control is achieved using 8x4 and 16x16
video switchers, and a 12x4 audio mixer. This system was the basis for
the AV system UMTRI built for a simulator at GM.
For further information see:
Anonymous (2002). New Driving Simulator, UMTRI Research Review, April-June,
33(2), 1-5.
|
-
When Are Driving Simulator Tests Favored?
In driving research, there is no one best tool. Sometimes one needs
a screwdriver and sometimes a hammer. In driving, the options are
often to conduct a paper and pencil evaluation, to use a simple mockup,
to collect data in a driving simulator, or to collect data on the
road. The “road” can refer to a test track or a public
road.
Often, the selection of a test context is between in a simulator
and on the road (in a real vehicle). The simulated situation provides
complete control of the test situation. The exact same conditions
of weather, road surface condition, ambient illumination, vehicle
positions, and speeds, etc., can be repeated time and time again,
and for every subject. In the real word, even with confederate vehicles,
there are some things beyond control, with the weather being one of
them.
Simulators are also favored where risk to drivers is an issue. People
cannot really die in a driving simulation, so they can be exposed
to crash provocative situations that a human subjects review board
would never accept for an on-the-road study.
However, simulations are just that, not the real world, only an approximation,
and ultimately, some time before production, concepts need to be assessed
in real vehicles on public roads.
At UMTRI, both approaches are used, with the choice depending on
the maturity of the solution being examined, the time of year, the
need to examine daylight vs. nighttime conditions, risk to subjects,
and so forth. In very large projects, it is often the case that concepts
are first explored in the driving simulator, and then when the alternatives
have been narrowed to a small set (and thought to be safe enough for
use by test subjects), explored on the road.
|
Can someone rent the driving simulator to teach my son/daughter
how to drive?
The simulator is a research tool, but could be used to develop an
educational program to help teenagers learn how to drive. At the present
time, UMTRI does not have such a program set up, or set of validated
scenarios to use for that purpose. UMTRI is very interested in developing
such, but a research project to develop and validate scenarios is
far beyond the means of all but the extremely wealthy (hundreds of
thousands of dollars).
Can someone rent the simulator to do research or testing?
Yes, but…Many have the mistaken impression that since they
can drive, they know how to conduct a driving study. A parallel claim
would be that knowing how to eat should provide the knowledge of how
to study eating and eating disorders. Knowing how to do a task and
how to study it are quite different.
In the case of the driving research, the issues are what should be
the dependent and independent measures, what should be the sampling
rates, how much time should be allocated for various conditions and
practice, and so forth. Furthermore, there are not existing worlds
or scenarios that are ready to use, so they need to be created for
each simulator experiment. For example, if one wanted an urban scenario,
the spacing between each intersection pair needs to be specified as
well as the timing of each traffic signal and the rules that cause
each signal to change. Also, the movement of each vehicle and each
pedestrian needs to be specified for each time period in each scenario.
Custom scenarios are created because there are not standards for such
and each investigator wants a world that that has been designed to
specifically answer their questions.
Depending on the project, creating the experiment requires at least
a few weeks of effort from the principal investigator and the simulator
engineer, and then a week to a few months to create the desired scenarios.
Because this requires intimate knowledge of the simulator, knowledge
gained from years of use, this is something an outsider could not
do efficiently. The new UMTRI scenario creator has the potential of
reducing scenario development times substantially.
When testing is in progress, the simulator engineer needs to start
up the simulator and load in the desired files, troubleshoot the system
when problems occur, operate the simulator during the experiment,
save the data from each block, and shut down the simulator at the
end of the day. Thus, even during the data collection phase, one needs
to add the burdened rate of the simulator engineer to the hourly charge
for simulator time.
Finally, once the data is collected, it needs to be reduced and filtered,
and then analyzed. Those tasks require researchers with prior experience
in processing driving simulator data.
Thus, prospective sponsors often think they can economize by just
paying for simulator time. However, since they are not intimately
familiar with the simulator, they do not know how to program scenarios,
operate the simulator, or process the resulting data. UMTRI’s
unique contribution to a project is primarily the expertise of its
research staff in conducting driving studies.
Furthermore, if UMTRI is not involved in the analysis of the data
and writing the final report, then the case for having UMTRI personnel
as the authors of resulting reports and articles is weak. Keep in
mind that the University evaluates UMTRI personnel (meaning job titles
are selected, salaries are set, and pay raises are determined) based
on the quantity of written output. The University wants to maximize
the intellectual contributions of its personnel.
Technical Details of the UMTRI Driving Simulator
Projectors |
Side channels
|
4 Epson Powerlite 703c |
Front channel |
1 Canon Realis SX50 |
Rear channel |
1 Epson Powerlite 82c |
Instrument panel |
1 Sharp XG-E850U |
Computers |
Host and side channels
|
5 Custom built PCs running Linux |
Forward and rear channels |
2 Dell Optiplex GX270 running Linux |
Experimenter interface |
1 Dell Optiplex 745 running Windows |
Instrument panel |
1 Mac G3 |
Secondary task |
1 Mac G4 |
Sight Distances (Subject to Screen) |
Center (16 ft)
Left (16 ft) Right (16 ft)
Far Left (13 ft) Far Right (13.5 ft)
Rear (12 ft) |
Screen Materials |
Front, rear, left and right screens |
4 permanent screens (drywall with special reflective paint) |
Far left and far right screens |
2 Custom built DaLite
pull-down matte white screens |
AV Equipment |
16x16 video switcher |
Knox 16x16 HB |
8x4 video switcher |
Sigma Electronics RC-840 |
|
