University of Virginia
Assessment of human driver safety at Dilemma Zones with automated vehicles through a virtual reality environment
Abstract: Ensuring the safety of mixed traffic environments, in which human drivers interact with autonomous vehicles, is an impending challenge. A virtual traffic environment provides a risk-free opportunity to let human drivers interact with autonomous vehicles, indicating how variability in traffic environments and human responses compromises safety. Analyzing the section of road preceding an intersection known as the Dilemma Zone helps identify risks, as the Dilemma Zone involves significant uncertainty in human response (ranging from 10% to 90% proceeding through the intersection versus stopping). The goal of this study is to discover vulnerabilities at the Dilemma Zone by implementing a virtual traffic environment, in which a human driver and autonomous vehicles share the road. The scenario of interest evaluates risks of rear-end collisions at the Dilemma Zone involving a human-driven vehicle following an autonomous vehicle. The virtual traffic environment was constructed in the Unity 3D platform and displayed through a Virtual Reality headset. The team tracked vehicle dynamics and physiological responses. Participants, recruited at the University of Virginia, were given pre- and post-experiment surveys, which collected information regarding driving experience and sentiment towards autonomous vehicles before and after driving in the virtual traffic environment. This study’s results demonstrated the use of virtual reality to expose human drivers to autonomous vehicles in low-risk situations. Rear-end collisions displayed improper manual driving behaviors, validating the Dilemma Zone’s risk. Decreasing comfort levels surrounding autonomous vehicles corresponded with increased following distances in the simulation. Conclusions revealed that incorporating autonomous vehicles onto roads with human drivers does not completely eliminate collisions, and that human-oriented factors are impediments to implementation. Developers should incorporate accurate autonomous vehicle algorithms into virtual reality to advance the technology.