Investigating the Safety of Human Drivers in Dilemma Zones with Autonomous Vehicles using Virtual Reality Simulations
In a recent study, participants from the University of Virginia were recruited to investigate the interaction between human drivers and autonomous vehicles (AVs) in critical road sections, particularly the Dilemma Zone. This zone, located near intersections, is where drivers must decide whether to stop or proceed, and becomes more complex when interacting with AVs due to their cautious or programmed responses.
The study, conducted in a virtual traffic environment constructed using the Unity 3D platform and presented through a Virtual Reality headset, aimed to discover vulnerabilities at the Dilemma Zone. Participants were given pre- and post-experiment surveys to provide information about their driving experience and sentiment towards autonomous vehicles.
Key findings revealed that AVs often follow strict collision avoidance and regulatory compliance strategies, leading to conservative behaviors like stop-and-go dynamics or slower reactions near obstacles or cyclists. This can cause discomfort or impatience in human drivers, as seen in the reduction of drivers’ "time efficiency" and comfort scores over time when following behind cautious AVs.
These human behavioral changes, such as impatience or abrupt decision-making prompted by uncomfortable or uncertain interactions with AVs, can increase the risk of rear-end collisions. Human drivers may respond awkwardly or unexpectedly, especially in DZ-like conditions where split-second decisions are required.
The simulation experiments also demonstrated that the decision-making logic of AVs in safety-critical, mixed-traffic scenarios can still have gaps leading to safety risk, including rear-end collisions, particularly when confronted with corner cases or aggressive behaviors from surrounding vehicles. Human drivers' uneasy reactions to AVs’ cautious or defensive driving may compound this problem.
While AVs are equipped with collision avoidance systems and adaptive cruise control designed to reduce crashes, their conservative driving patterns can sometimes trigger uncomfortable or unintended responses from human drivers, reducing traffic flow smoothness and increasing collision risk in zones like the Dilemma Zone.
In summary, human drivers' discomfort and response patterns towards autonomous vehicles’ cautious maneuvers in critical areas such as the Dilemma Zone can escalate rear-end collision risk, especially in simulated environments modeling mixed traffic. This effect arises due to a complex interplay where AVs' conservative and regulation-compliant driving causes human drivers to adjust behavior in less predictable or comfortable ways, sometimes increasing collision likelihood.
The study suggests that improving AV controller design to better consider human driver comfort and behavioral tendencies is key to reducing rear-end collisions around the Dilemma Zone. It also highlights the importance of incorporating accurate autonomous vehicle algorithms into virtual reality environments to advance the technology and ensure safer interactions between human drivers and AVs.
Technology plays a significant role in simulating real-world scenarios for investigating human-autonomous vehicle interactions in critical zones. The study uses Unity 3D platform and Virtual Reality headset to create a virtual traffic environment, aiming to improve AV controller design for better human comfort and safer interactions.
