Go to previous article
Go to next article
Return to 2003 Table of Contents
David L. Jaffe
VA Palo Alto Health Care Systems
Palo Alto, CA
During the summer of 2002, the Paralyzed Veterans of America Bay Area and Western Chapter and the Spinal Cord Injury Service at the Veterans Affairs Palo Alto Health Care Systems sponsored a project under the Summer Scholars Program. This ten-week effort investigated the use of a driving simulator to assess the driving performance of veterans following a CNS injury.
Subjects for the pilot study were recruited from the Comprehensive Rehabilitation Center (CRC) population within the Physical Medicine and Rehabilitation Service. These subjects were either outpatients living in the local area or patients who were hospitalized and undergoing rehabilitation. The veteran subject population had a mean age of 58 years (range of 28 to 78) and consisted of 4 stroke, 3 TBI, and 3 SCI/D (spinal cord injury / disorder) subjects (1 female and 9 male). All of these ten participants had prior driving experience and were scheduled for an in-car driving evaluation. Only subjects who were able to operate a standard accelerator and brake were used, as hand controls could not be fitted to the driving simulator used for this study.
In addition, a control group consisting of four individuals without a disability was included in the study to provide normative data to establish a baseline of performance for comparison with patients with CNS injury.
The study protocol consisted of three items: a multi-part driving questionnaire, a driving simulator assessment, and an in-car assessment.
A brief three-part driving questionnaire was administered before and after driving the simulator and after the on-road evaluation. It assessed the subjects' driving history, physical condition, perception of how well they thought they would perform various driving tasks with the simulator or in a car, emotional state, and confidence in passing the DMV road test. Additional questions solicited the subjects' thoughts on the feel of the simulator and how well it represented their driving ability. The final section of the questionnaire asked subjects how well they thought they drove on the road and their thoughts about the role of a simulator in preparing them for driving a car.
For the pilot study, a low-cost driving simulator from Systems Technology Incorporated (STI - Hawthorne, CA), the STI version 8.16, was employed. This circa 1996 product consisted of a twenty-one inch color monitor for the simulator display, a steering wheel mounted on a base, throttle and brake pedals, and a DOS-based PC, monitor, and speakers. Its specialized software produces the visual scenes and auditory driving displays. Driving courses were scripted with commands in an 'event' file. The commands are elements of a simple scenario definition language that defines the appearance and actions of the driving situation including the road, other vehicles, signs, traffic lights, pedestrians, intersections, buildings, and scenery. The graphics on this system were rather crude - the images of other cars and buildings lacked detail, shading, and texture.
Road courses for three different difficulty levels were scripted to model typical driving environments one would encounter 1) around the VA Palo Alto Medical Center, 2) in residential areas, and 3) in commercial and freeway settings. Each course is representative of the different driving situations a veteran would be exposed to if he/she participated in a regular VA driving evaluation under the supervision of a certified VA driving instructor. Each course was designed to take approximately ten minutes to complete.
The first course was 10000 feet in length and resembled the perimeter road on the Palo Alto Medical Center's grounds. This simplest course was designed with only a few cars approaching from the opposite direction and few pedestrian crossings. The speed limit was set at 15 miles per hour.
The second simulator course resembled a residential area that included more simulated pedestrians, cars approaching the driver, and cars in the driver's lane. The addition of cars in the same lane adds to the course's difficulty as subjects must plan and maneuver across the centerline to pass slower cars. The length of the course was 16000 feet and the speed limit was 25 miles per hour.
The final course represented a driving environment that would be encountered in a commercial district and on a freeway. This four-lane course included traffic lights instead of the stop signs of previous courses and a higher traffic density. Subjects must react to changing traffic lights and maneuver between the two lanes to overtake slower vehicles. Sharper turns and a higher speed limit (45 and 55 mph) added difficulty on this 31000-foot stretch of simulated road.
The STI driving simulator collected a variety of data during operation: off-road accidents, collisions, lane position, steering wheel rate, road heading error, curvature error, throttle input, longitudinal acceleration, and longitudinal speed. The simulator recorded the average and standard deviation of most of these measures. Additional data recorded included: number of pedestrians hit, speed limit violations, number of traffic light stops, and simulation run time.
The subject's driving performance with the simulator was scored by an investigator using a form that itemized various driving aspects: speed, obey traffic signs and signals, safety, lane tracking, lane changes, turns, steering control, following distance, brake reaction time, throttle/brake coordination, merging into traffic, and speed / accuracy of decisions.
Veteran subjects with CNS disorders were identified from the in-patient and local outpatient populations. All of these individuals were scheduled to undergo an in-car evaluation. The pre-simulator portion of the driving questionnaire was administered. Next, a short practice session helped the subject get accustomed to the operation and feel of the driving simulator. Next, the three course simulations were driven. Then the second part of the questionnaire was then filled out.
After using the simulator, the Driving Program manager performed the behind-the-wheel driving assessment in one of the VA Driving Program vehicles. Their in-car driving performance was scored using the same evaluation form employed in the simulator assessment. The in-car assessment protocol was the same as is normally performed by the VA driving instructor.
Data from the subjective driving evaluation forms, questionnaires, and driving simulator were analyzed to determine a relationship between simulator performance and behind-the-wheel performance for all participants. The following hypotheses were tested:
1. Different driving performances are produced by control subjects, SCI subjects, and TBI subjects. This employs the simulator data.
2. SCI and TBI subjects' in-car performance can be predicted from their simulator performance. That is, patients who do poorly on the simulator also perform poorly with the in-car evaluation. This utilizes data from the subjective driving evaluation forms.
A comparison of overall performance differences between the control and subject groups on the simulator using the data collected by simulator was made. It shows a significant difference (P<0.05) between the two groups. This data validates Hypothesis 1: subjects' performance on the simulator was significantly different (and poorer) than the control group.
Differences in simulator performance by the three course levels by both groups using the subjective scoring data were also analyzed. There is a significant difference (P<0.05) between the performance on Levels 2 and 3, the two more difficult levels where higher driving speeds, more cars, more pedestrians, and sharper curves were presented.
Subjective scores between subjects with SCI, Stroke and SCI/D during simulator and on-road driving showed no statistical difference.
Finally a graph was made of simulator versus on-road performance using the subjective driving data for those subjects who were unable to complete their in-car evaluation due their unsafe driving performance. The statistical significance of this relationship (P<0.05) implies that unsafe on-road drivers can be identified from their poor simulator performance, validating Hypothesis 2.
The driving questionnaire responses were not statistically analyzed. However, prior to their in-car assessment, subjects almost universally reported a high level of confidence in their ability to drive a car despite their injuries. They also under-reported their physical limitations they thought would affect their driving. This creates a potentially dangerous situation: individuals with CNS injuries may consider themselves able to drive, when in fact they are unsafe.
The pilot study had several limitations. The simulator provided only a crude low-resolution image. Its software was unable to generate limit lines and pedestrian walkways at intersections and crosswalks. This made it difficult to estimate where to stop. In addition, the simulator's one monitor provided a single through-the-windshield view. State-of-the-art simulators employ three (or five) monitors for a wide-angle view encompassing the left and right side windows for a more realistic driving experience.
The short study length limited the number of subjects to 10, an insufficient number to determine if there was a statistical driving performance difference between groups of CNS injuries.
However, the pilot study did demonstrate that a driving simulator could be used to identify patients with driving problems before an in-car assessment.
Fredericks P, Using a Simulator to Assess the Driving Ability of Veterans following Stroke, TBI, and SCI, Bay Area and Western PVA/SCI Summer Scholars Program Report, August 2002.
Go to previous article
Go to next article
Return to 2003 Table of Contents
Return to Table of Proceedings