操舵制御に関するドライバの意図認識システム

抄録

This paper proposes recognition systems of driver's intention on choosing steering-control modes, lateral control tasks, and on his/her driving mood, which would be essential part of an intelligent steering-assistance system. The steering-control mode recognition system determines which steering-control mode the driver selects, manual or automatic, by applying a sequential probability ratio test to the time-series of prediction errors in two AR Model of preprocessed steering torque signal. The control task recognition system consists of two recognition subsystems, one utilizes Hidden Markov Model Filter, and the other employs ARMAX Model and statistical testing. The driving mood recognition system calculates an index of relative change of aggressiveness of driving, defined as a weighted sum of moving averages of signals such as vehicle yaw rate, longitudinal acceleration, and driver's remnant. The weights on each factor of the index of relative change of aggressiveness are obtained by using the learning algorithm of Support Vector Machine. The results of simulation study using the data drawn from actual vehicles show that the systems would achieve high recognition capability. Furthermore, the proposed systems show a robustness to the vehicle speed change and variation of control characteristics among drivers.