Verification of Coaching effect by Instructor-like Assistance System Based on Model Predictive Constraint Satisfaction

Safety and acceptability are the main concerns in the design of driver assistance systems. However, these two requirements sometimes conflict with each other depending on the situation and the driver. This conflict is particularly emphasized in the case of elderly drivers. To solve this problem, this paper proposes a driver-vehicle cooperation scheme, an "instructor-like assisting control" consisting of model predictive constraint satisfaction and a multi-modal human-machine inter-face. The proposed assisting scheme is expected to improve the drivers’ inherent driving characteristics, which is recognized as a "coaching effect" in cognitive science. This effect was verified by long-term experiments over one month using a driving simulator.