Design and Evaluation of Adaptive Collision Avoidance Systems

Taking a human factors approach, the present study aims at improving driver interaction with automation by improving driver trust in and understanding of the system and enhancing system design. First, a driving experiment was conducted to investigate how driver understanding of the system capabilities effects driver performance and trust. The experiment compared two driver assistance systems for avoiding collisions during critical lane change: one was a haptic steering control that manipulates the steering wheel friction torque, and the other was an automatic steering control that decouples the driver during critical conditions. The results indicate that, especially in critical situations when driver expectation of the system and system capabilities were not aligned, the driversystem interaction was significantly affected by the way control is allocated between agents. To improve system design in terms of functional allocation and capabilities, the study proposes an enhanced adaptive collision avoidance system in which control is allocated dynamically depending on the situation. This system was assessed in a second driving experiment. While the diver-system interactions significantly improved compared to the haptic and automatic steering control systems, in terms of safety, it did not perform as well as expected. A third experiment, using long term simulator training, was conducted to enhance drivers’ understanding of and trust in the system. The training interaction revealed that drivers adapted more easily to the system, improving driver performance, system effectiveness, and safety. The findings highlight how user training can improve human-automation interaction