MPC-based switched driving model for human vehicle co-piloting considering human factors

In vehicle longitudinal control, improved comfort and reduced operation workload for human drivers are achieved with the ACC (Adaptive Cruise Control), which still requires the driver to maintain full attention on monitoring, meanwhile, the risk of distraction and fatigue rises resulting from the long-term supervising task, which has a strong impact on the successful take-over. The key point to make full use of ACC advantages and make up human's weakness on supervising task is the reasonable arrangement of control time of ACC and human driver. In this paper, an MPC (Model Predictive Control) based optimized switching strategy of longitudinal driving authority transition is proposed, which aims to provide proper advice for human drivers to handover and takeover control authority, through minimizing the overall index consisted of the operation workload, fuel consumption, takeover risk and tracking errors. In addition, a new driver longitudinal model considering actual reaction time delay and insensitive distance perception of human drivers is proposed as well, which combines with ACC to constitute a typical switched control system called the switched driving model as the predictive model for MPC. The stable condition of the new driver longitudinal model is derived by using describing function method and a sufficient condition to ensure steady switched system is given by using the Lyapunov method and LMI approach. The results of simulator experiments show that the new driver model describes the car-following behavior of real human driver better. What's more, the simulation results demonstrate that the performance index of switched driving is smaller than the human driving and ACC driving only, and the optimization time is short enough to meet the requirement of engineering practice.