Trust and Acceptance of Adaptive and Conventional Collision Avoidance Systems

Abstract

A simulator driving experiment was conducted to test the implementation of an adaptive driver support system for avoiding collisions caused by hazardous lane changes. Conventional (nonadaptive) collision avoidance systems with fixed function allocation might be effective from a safety perspective, but they are not always desirable from human factors perspective due to problems related to performance, trust, and acceptance. We therefore propose an adaptive collision avoidance system to provide support depending on the situation in consideration of the driver’s ability to handle the situation and the time to collision. Four driving modes were compared: no support, a haptic shared steering control system with an adaptive haptic authority based on time to collision, a cooperative steering control system that decoupled the driver’s steering input and controlled the vehicle’s direction temporarily, and an adaptive steering control system in which both haptic and cooperative control systems were integrated to provide different automated interventions at the steering wheel depending on the situation. The simulation participants were subjected to various potentially hazardous lane change maneuvers that provided significant differences in their abilities to perceive the risk and handle the situation. Compared to the no-support condition, all systems were more effective in accident prevention. The adaptive system was more effective than the haptic system and less effective than the cooperative system in terms of safety. However, driver performance, trust, and acceptability were significantly improved with the adaptive system compared to the nonadaptive systems.