Evaluation and Validation of the Effect of Connected and Automated Vehicle Safety Applications on Driver Behavior - A Driving Simulator Approach
Loading...
Links to Files
Permanent Link
Collections
Author/Creator
Author/Creator ORCID
Date
2019-10-01
Type of Work
Department
Transportation
Program
Doctor of Philosophy
Citation of Original Publication
Rights
Subjects
Abstract
Background–Considering the rapid boom in information technology and people’s increasing dependence on mobile data, automotive manufacturers have started equipping vehicles with wireless communication capabilities, manufacturing what are commonly known as connected vehicles, and automated systems to assist drivers with certain driving tasks. These technological advances have led to fast tracking the deployment of connected and automated vehicles, and an increased momentum in implementing these applications, as the number of driving assistance systems pre-equipped in cars by automotive companies has witnessed a sharp increase during this decade. However, most of the new cars come pre-equipped with these applications, which means that the drivers’ reactions to such applications are not fully examined since most of the experiments involving these applications are done using microscopic simulations with the behavior of the drivers’ being assumed. Therefore, this rapid deployment and implementation has led to a lack of research in understanding the drivers’ reactions to such applications before they actually use them, which is an essential element in ensuring the effectiveness and successful implementations of such applications.
Objectives–To investigate driver behavior in terms of braking, steering and throttle control, change in speed and eye gaze movements in the presence of connected and automated vehicle applications using a driving simulator. Some of these driving simulator findings are validated using real world data.
Method–Four distinct analysis are conducted to evaluate five different connected and automated vehicle applications. A hazard-based duration model is used to evaluate driver braking behavior while a random forest model is used to rank the most important variables impacting change in speed and steering wheel and throttle take over reaction times. A heat map of eye gaze movements is created to identify objects of interest with the most fixations.
Data–The study consisted of 93 participants from diverse socio-economic backgrounds who drove in 186 experiments. These participants answered a pre-simulation socio-demographic questionnaire as well as a post simulation, driving simulation experience questionnaire. Two of the connected and automated vehicle applications in the driving simulator were validated using real world data obtained from University of Michigan Transportation Research Institute.
Results–The use of Pedestrian Collision Warning and Red-Light Violation Warning had a significant impact on participant braking behavior, where participants resorted to initial aggressive braking in the presence of these applications. The eye gaze heat map showed that majority of the participants glanced at the visual notification from these applications thus reinforcing the findings from the significant braking behavior. Forward Collision Warning had a positive influence on change in speed while Curve Speed Warning had no impact on speed. These findings were validated with the findings from the real-world data provided by University of Michigan Transportation Research Institute. Lastly, the steering wheel and throttle Take Over Reaction times in the post autonomous mode being 2.47 seconds and 2.98 seconds respectively, is greatly influenced by the annual miles driven, age, and familiarity with this technology. The gaze analysis shows that participants viewed the rearview mirror and speedometer more while also being involved in offscreen distractions, while in the autonomous mode.
Conclusions–This study shows that the CAV applications with the exception of the Curve Speed Warning (CSW) application, are effective in improving the safety and comfort of the drivers, confirming the hypothesis that these CAV applications impact the drivers’ performance positively. The contributions and implications of the study are discussed.