Date of Award
12-2014
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Legacy Department
Mechanical Engineering
Committee Chair/Advisor
Dr. John Wagner
Committee Member
Dr. Timothy Burg
Committee Member
Dr. Todd Schweisinger
Committee Member
Dr. Ardalan Vahidi
Abstract
Traffic fatalities and injuries continue to demand the attention of researchers and governments across the world as they remain significant factors in public health and safety. Enhanced legislature along with vehicle and roadway technology has helped to reduce the impact of traffic crashes in many scenarios. However, one specifically troublesome area of traffic safety, which persists, is run-off-road (ROR) where a vehicle's wheels leave the paved portion of the roadway and begin traveling on the shoulder or side of the road. Large percentages of fatal and injury traffic crashes are attributable to ROR. One of the most critical reasons why ROR scenarios quickly evolve into serious crashes is poor driver performance. Drivers are unprepared to safely handle the situation and often execute dangerous maneuvers, such as overcorrection or sudden braking, which can lead to devastating results. Currently implemented ROR countermeasures such as roadway infrastructure modifications and vehicle safety systems have helped to mitigate some ROR events but remain limited in their approach. A complete solution must directly address the primary factor contributing to ROR crashes which is driver performance errors. Four vehicle safety control systems, based on sliding control, linear quadratic, state flow, and classical theories, were developed to autonomously recover a vehicle from ROR without driver intervention. The vehicle response was simulated for each controller under a variety of common road departure and return scenarios. The results showed that the linear quadratic and sliding control methodologies outperformed the other controllers in terms of overall stability. However, the linear quadratic controller was the only design to safely recover the vehicle in all of the simulation conditions examined. On average, it performed the recovery almost 50 percent faster and with 40 percent less lateral error than the sliding controller at the expense of higher yaw rates. The performance of the linear quadratic and sliding algorithms was investigated further to include more complex vehicle modeling, state estimation techniques, and sensor measurement noise. The two controllers were simulated amongst a variety of ROR conditions where typical driver performance was inadequate to safely operate the vehicle. The sliding controller recovered the fastest within the nominal conditions but exhibited large variability in performance amongst the more extreme ROR scenarios. Despite some small sacrifice in lateral error and yaw rate, the linear quadratic controller demonstrated a higher level of consistency and stability amongst the various conditions examined. Overall, the linear quadratic controller recovered the vehicle 25 percent faster than the sliding controller while using 70 percent less steering, which combined with its robust performance, indicates its high potential as an autonomous ROR countermeasure. The present status of autonomous vehicle control research for ROR remains premature for commercial implementation; in the meantime, another countermeasure which directly addresses driver performance is driver education and training. An automotive simulator based ROR training program was developed to instruct drivers on how to perform a safe and effective recovery from ROR. A pilot study, involving seventeen human subject participants, was conducted to evaluate the effectiveness of the training program and whether the participants' ROR recovery skills increased following the training. Based on specific evaluation criteria and a developed scoring system, it was shown that drivers did learn from the training program and were able to better utilize proper recovery methods. The pilot study also revealed that drivers improved their recovery scores by an average of 78 percent. Building on the success observed in the pilot study, a second human subject study was used to validate the simulator as an effective tool for replicating the ROR experience with the additional benefit of receiving insight into driver reactions to ROR. Analysis of variance results of subjective questionnaire data and objective performance evaluation parameters showed strong correlations to ROR crash data and previous ROR study conclusions. In particular, higher vehicle velocities, curved roads, and higher friction coefficient differences between the road and shoulder all negatively impacted drivers' recoveries from ROR. The only non-significant impact found was that of the roadway edge, indicating a possible limitation of the simulator system with respect to that particular environment variable. The validation study provides a foundation for further evaluation and development of a simulator based ROR recovery training program to help equip drivers with the skills to safely recognize and recover from this dangerous and often deadly scenario. Finally, building on the findings of the pilot study and validation study, a total of 75 individuals participated in a pre-post experiment to examine the effect of a training video on improving driver performance during a set of simulated ROR scenarios (e.g., on a high speed highway, a horizontal curve, and a residential rural road). In each scenario, the vehicle was unexpectedly forced into an ROR scenario for which the drivers were instructed to recover as safely as possible. The treatment group then watched a custom ROR training video while the control group viewed a placebo video. The participants then drove the same simulated ROR scenarios. The results suggest that the training video had a significant positive effect on drivers' steering response on all three roadway conditions as well as improvements in vehicle stability, subjectively rated demand on the driver, and self-evaluated performance in the highway scenario. Under the highway conditions, 84 percent of the treatment group and 52 percent of the control group recovered from the ROR events. In total, the treatment group recovered from the ROR events 58 percent of the time while the control group recovered 45 percent of the time. The results of this study suggest that even a short video about recovering from ROR events can significantly influence a driver's ability to recover. It is possible that additional training may have further benefits in recovering from ROR events.
Recommended Citation
Freeman, Paul, "Autonomous Control and Automotive Simulator Based Driver Training Methodologies for Vehicle Run-Off-Road and Recovery Events" (2014). All Dissertations. 1441.
https://tigerprints.clemson.edu/all_dissertations/1441