Where The Laboratory Meets The Road
Fine-tuning a highway driving simulator’s motion-base system goes a long way toward heightening a test subject’s perceptions, leading to a more realistic driving experience.
Understanding how drivers react to stimuli in the roadway environment is critical to improving the design of U.S. roads. Driving simulators are among the tools that human factors researchers use to assess motorist behavior. Driving simulators that are designed for research provide several advantages over the collection of data in real-world environments. Where the safety of a design is in question, a simulator can enable observation of a driver’s performance without the risk of injury. Driving simulators also enable control over traffic and weather, and can ensure the exact repeatability of conditions, which is important in isolating cause and effect.
For instance, researchers can specify roadway scenarios and events to simulate specific interactions between vehicles and pedestrians. They also can repeat simulations without waiting for those interactions to occur naturally in particular real-world locations.
Driving simulators used for highway research range from simple desktop units with a monitor, steering wheel, brake, and accelerator pedals to large domes that enclose a complete vehicle mounted on a complex, eight-degrees-of-freedom motion platform that offers a 360-degree field of view. Motion platforms add to the realism of the simulation by offering additional motion cues (proprioreceptive, vestibular) to better simulate real-world driving, immersing drivers in the experimental conditions.
However, presenting motion cues that are perceived as realistic to participants in driving simulator research presents numerous challenges. All motion platforms are limited in their range of motion before they must return to a neutral position. The goal is to impart to participants realistic perceptions of acceleration, braking, and turning, while staying within the platform’s limited range of motion.
Researchers have used driving simulators in a variety of research and development applications. According to Chris Monk, chief of the Human Factors/Engineering Integration Division at the National Highway Traffic Safety Administration, “driving simulators have been used for developing and visualizing infrastructure, testing advanced safety systems that warn drivers, driver distraction research, and analyzing driver behavior, among other applications.”
More specifically, at the Federal Highway Administration, transportation safety specialists and engineers can “drive” a simulated road before it is built to verify signage, pavement markings, lighting, emergency response access, and emergency traffic flow. FHWA’s Turner-Fairbank Highway Research Center is home to a motion-base highway driving simulator with a virtual 360-degree field of view generated by three ultra-high-definition projectors and supplemented with three liquid crystal display (LCD) monitors that simulate rearview mirrors. The simulator’s sound system provides engine, wind, and tire noise, as well as other environmental sounds. Three-dimensional digital image-generation software offers a scalable, flexible, and customizable environment for creating real-time, fully interactive driving simulations.
The FHWA highway driving simulator was recently upgraded with a new hexapod motion base with six degrees of freedom. The parameters of a six-degrees-of-freedom motion base are roll, pitch, yaw, x (linear forward/backward), y (linear left/right), and z (vertical). The visual system projects images within the driver’s field of view onto a cylindrical screen surrounding the vehicle.
The visual scene is synchronized with the motion base’s movement to enable the scenes to correlate exactly with driver controls, thus maximizing realistic perception. However, as part of the upgrades to the motion base, the laboratory staff needed to ensure that the motion-base system is finely tuned to provide the best perception for drivers.
“Tuning is critical to increase the realism and validity of the driving experience, while decreasing the likelihood of causing users to experience simulator sickness,” says David Yang, Human Factors team leader at FHWA. “That’s when test drivers feel discomfort caused by incompatible signals from visual--motion sensing--systems.” Aside from being unpleasant, simulator sickness can affect the reliability of measurements.
The team tuned the motion base, including all of the relevant system enhancements, to improve test drivers’ perception of the connectedness between the motion base and the visual system display. Here’s how they did it.
Tuning for Perception and Data Collection
The goal in tuning the simulator is to adjust the system functions systematically and methodically to achieve the most realistic driving experience feasible. Because the motion base was the major system to be significantly upgraded, it was the starting point for system tuning. However, other simulator functions fed into the overall solution. For instance, steering, acceleration, and braking are all key components in effective functioning of the highway driving simulator.
Creating a Roadway Layout
From a practical standpoint, the first step was for the researchers to employ a field research vehicle to drive real-world roadway layouts. This would help determine which layouts would enable starting, stopping, and turning maneuvers that would fall within the comfortable range of control for a typical driver. This roadway layout would serve as the baseline for tuning the simulator.
The field research vehicle, an instrumented sport utility vehicle capable of collecting data such as vehicle speed, acceleration, and other measures, was used to test proposed roadway layouts that would be modeled and programmed to test the motion base of the simulator. The field vehicle used the test track at the Federal Law Enforcement Training Center in Cheltenham, MD. If a proposed roadway layout was found to be uncomfortable to drive, for example, if turns could not be made comfortably at the desired speed, then that roadway layout would be dropped from further consideration. After screening out unrealistic layouts, the remaining layouts were then modeled and programmed into the simulator and used for perception testing of the motion base.
Simulator Perception Feedback
To begin the tuning process for the motion base system, which includes the hexapod apparatus that enables the vehicle to move with six degrees of freedom, a core group of simulation engineers and psychologists who research human factors performed the initial testing. This testing was an iterative process, and each change to the motion-base settings achieved a more refined level of realism for the driver. The team tested the perception of motion after each change to the motion base by driving the simulator. Each team member evaluated the perceptions of steering, acceleration, braking, and turning, and their feedback led to increased refinement of driver perception.
If more than one of the core members responded with similar negative feedback, the research team continued to seek a combination of settings that was acceptable to all. For instance, when braking, the motion system tilts the car forward to simulate the feeling produced when a car slows but inertia causes the driver’s body to attempt to maintain the current velocity. If the forward tilt is perceived as something other than a typical deceleration, that perception is a false cue. In extreme cases, a false deceleration cue can be perceived as spinning or sliding (as if on ice) rather than as a change in speed. This was resolved by filtering out small parameter changes to the motion base after the visual scene indicated a stop in motion.
Tuning and Testing
The configuration of the motion base includes three modes with 100 variable parameters total. The modes indicate the amount of “washout”--none, medium, or full--that occurs in the movement of the motion base. Washout is the ability of the software to anticipate what the motion base will do next by pre-positioning the motion base. For example, if the car is coming to a stop, the motion base will gently move to the rear-most location so that upon acceleration, the motion base can lunge forward and provide the perception of acceleration.
An example of a parameter for a mode is a variable that specifies the allowable distance that the motion base can travel in the x, y, or z directions. One hundred variables produces an extremely large number of possible combinations of settings for a tuning exercise. Therefore, a subject matter expert with extensive experience in motion-base driving simulators, equations of motion, and vehicle dynamics was brought in to select the initial settings by tuning the simulator to a baseline level. The expert was able to reduce the number of settings to be considered during tuning to one mode of operation and 20 parameters that were of the most significance.
For each track layout--a highway scenario and a driving maneuver (acceleration, braking, turning, and slalom)--the core testers conducted runs to ensure that the perceptibility was acceptable, that is, as realistic as possible. After the core team agreed upon the acceptable driving perceptions for all layouts and maneuvers, the research team recruited 10 volunteers, all of whom were FHWA employees and contractors, to evaluate the perception of motion during longitudinal acceleration and braking, turning, and negotiation of a slalom. The team then collected additional feedback to help validate the selected motion system settings.
For example, to evaluate the motion during acceleration and braking, the team developed a questionnaire to evaluate the quality of the perceived motion and whether any false cues were present. Acceleration was tested from 0 miles per hour (mi/h) (0 kilometers per hour [km/h]) to each of the following speeds: 10, 30, 40, and 50 mi/h (16, 48, 64, and 80 km/h). Two different rates of acceleration were rated: moderate and hard. Once at the assigned speed, the participants were asked the following series of questions:
- How realistic did the acceleration feel? Rate it on a scale from 1 to 7, where 1 represents “not at all realistic” and 7 is “very realistic.”
- While accelerating, did you ever feel like you were decelerating?
- Did you feel as though you were being tilted up or down?
- Overall, how acceptable was the motion cueing?
For the second and third questions, if the answer was “yes,” the participant was also asked to rate how strong the feeling was.
The researchers also tested deceleration from braking in a similar manner. In fact, after answering the acceleration questions for a particular speed, the participants were asked to brake either moderately or hard and then answer a similar set of questions. Analogous tests were conducted for turns of various radii and for a slalom course at two different speeds.
Perceptions of Acceleration and Braking
Participants gave ratings for the realism and acceptability of the simulator’s acceleration that were nearly identical (so only the acceptability ratings are presented here). For acceptability, the mean ratings ranged from 5.3 to 5.7 on the 7-point scale. The research team judged these ratings to be satisfactory. A few of the 10 participants reported a slight indication of false cueing. In all cases of reported false cues, the participants indicated that when they completed the acceleration, they felt a slight deceleration. In most cases, that feeling was rated a 3 on a scale from 1 to 7, where 1 represented “not very noticeable” and 7 represented “very noticeable.”
The researchers also deemed the participants’ acceptability ratings for motion cueing related to braking to be acceptable, with mean ratings from 5.0 to 5.8 on the 7-point scale. Most of the participants perceived tilting while braking, but the researchers attributed this feeling to the motion system dipping the nose of the car, as would be expected when decelerating, as well as to induce a perception of sustained deceleration. There were only three reports of false cues (for example, feelings of acceleration at some point during braking) across dozens of trials, and even in those cases, the participants judged the overall perception as acceptable.
Perceptions of Turning
The researchers asked participants to rate their perception of turns with turn radii of 131 and 984 feet (40 and 300 meters), each with deflection angles of 60 and 90 degrees. (The deflection angle is the change in compass heading, so that a 90-degree deflection would be a right angle change in direction.) All turns were made while holding the speed constant. The speeds tested ranged from 5 to 30 mi/h (8 to 48 km/h), except that the higher speeds were not used with the 131-foot (40-meter) radius and the lowest speeds were not used with the 984-foot (300-meter) radius because the researchers were testing normal driving behavior.
Mean acceptability ratings ranged from 5 to 7, where 1 represented “not at all acceptable” and 7 was “totally acceptable.” The participants provided a mean rating above 5 (on the same 7-point scale) for their overall perceptions of the turning maneuvers for all speeds, radii, and deflections.
Sustained turning, such as what a driver experiences traversing a roundabout, is challenging to simulate, but turns of shorter durations tend to be perceived as more realistic. “Turning is a difficult perception to simulate, compared to how a person would feel in the real world,” says Jason Williams, a senior software engineer with Textron Systems Support Solutions who works on highway driving simulators. “It is a difficult perception to simulate because there is a physical limit to the movements of a six-degrees-of-freedom motion base, which is truncated when longer, sustained forces are usually experienced by drivers in real-world turning maneuvers, such as when traversing a roundabout.” However, the results indicate that the motion base does simulate the initial force of turning and provides drivers with a realistic perception of shorter turns, such as lane changes.
Perceptions of Slalom Maneuvers
The participants drove through a simulated slalom course set up with large cones placed 100 feet (30.5meters) apart (distance chosen for reasonability to maneuver through each slalom based on normal driving). Participants were instructed to alternate passing to the right and left of the cones while maintaining a fixed speed. Three speeds were used: 15, 20, and 25 mi/h (24, 32, and 40 km/h). The researchers determined these speeds to be reasonable and safe based on speeds measured in the field. Following each pass through the slalom, the participants were asked questions such as these: How realistic was the feeling of moving through the slalom? How comfortable did you feel moving through the slalom? How controllable did the car feel while driving through the slalom? When asked about realism, comfort, and controllability, the participants’ mean ratings were all above 5 on the 7-point scale, where 1 was low (“not at all realistic,” “very uncomfortable,” and “uncontrollable,” respectively) and 7 was high (“very realistic,” “very comfortable,” and “easy to control,” respectively).
The researchers also asked the participants whether they had experienced any false motion cues while driving through the slalom, such as slipping, sliding, spinning, or moving in the wrong direction. All of the participants indicated that they had not experienced false motion cues.
Making It “Feel Like The Real Deal”
Producing a driving simulation that mimics the exact perceptions that a motorist experiences while driving a car on the road remains a challenge. For instance, it is difficult to create the exact acceleration, deceleration, and turning forces that a driver feels while driving. However, through systematic and methodical tuning of driving simulators, researchers can optimize all significant functions for the best possible driver perception.
According to Monique Evans, director of FHWA’s Office of Safety Research and Development, “ensuring that the driving simulator’s motion base is finely tuned in order to increase the realism and validity of test results is crucial for performing valid experimental studies.”
Although the results achieved for turning movements were not as favorable as those for acceleration, braking, and slalom testing, the researchers decided that the tuning process should focus mostly on acceleration and braking, because the next few planned experiments involve driving only on freeways. Further tuning of the turning motion was deferred until an experiment with navigation at lower speed roads is planned.
Another finding from this study is that the motion base is not always the root cause of false cues. The researchers found that modifying steering, accelerator, and braking algorithms also can help produce the best possible driver perceptions.
“Overall,” says FHWA’s Yang, “the systematic tuning process minimized the perception of false cues during simulated scenarios. And, this fine-tuning makes operating the highway driving simulator feel like the real deal.”
Barry Wallick, with Textron Systems Support Solutions, has been the program manager and principal systems engineer for FHWA’s highway driving simulator at the Turner-Fairbank Highway Research Center since 2002. Wallick received a B.S. in electrical engineering at the University of Maryland and an M.S. in computer science at Loyola University Maryland.
Michelle Arnold is a transportation specialist on the Human Factors Team in FHWA’s Office of Safety Research and Development. She received a B.S. in psychology and an M.A. and Ph.D. in applied behavior analysis from Western Michigan University.
For more information, please contact Michelle Arnold at 202–493–3990 or email@example.com.