Stochastic predictive control based motion planning for lane change decision using a Vehicle Traffic Simulator

This paper describes the design and evaluation of a model predictive control algorithm for automated driving on a motorway using a Vehicle Traffic Simulator. For the development of a highly automated driving control algorithm, motion planning is necessary to satisfy driving condition in various road traffic situations. There are two key issues in motion planning of automated driving vehicles. One of the key issues is how to handle potentially dangerous situations that could occur in order to guarantee the safety of vehicles. The second key issue is how to guarantee the robustness of the controller under model uncertainties and external disturbances. To improve safety with respect to the future behaviors of subject vehicles, not the current states but rather the predicted behaviors of surrounding vehicles should be considered. The desired driving mode and a safe driving envelope are determined based on the probabilistic prediction of surrounding vehicles behaviors over a finite prediction horizon. To obtain the desired steering angle and longitudinal acceleration for maintaining the subject vehicle in the safe driving envelope during a finite prediction horizon, a motion planning controller is designed based on an MPC approach. The developed control algorithm has been successfully implemented on a vehicle ECU. The proposed control algorithm has been evaluated on a real-time vehicle traffic simulator. The throttle, brake, and steering control inputs and the controlled vehicle behavior have been compared to those of manual driving.