Robust Model Predictive Control for Path Tracking of Autonomous Vehicle

Path tracking is one of the critical technologies in the autonomous vehicle. Its performance may be seriously affected by disturbance resulting from unpredictable environment like changes in road friction coefficient and parameter uncertainty such as cornering stiffness and mass caused by errors of measurement. Besides, since the vehicle system consisting of many systems is an extremely complex nonlinear system, it is almost impossible for us to establish a precise model of a vehicle especially when it is moving. These inevitable factors influence the control accuracy and even threaten the stability and safety of the vehicle system. This paper proposed a promising solution to this problem, robust MPC (Model Predictive Control) combined with the optimal preview controller for path tracking problems of an autonomous vehicle. The state space model in tracking error variables of a passenger vehicle used for path tracking application is established. In order not to regard road curvature as additive disturbance, the error dynamics of the vehicle for a given road curvature is transformed to the error model deviating from the steady state trajectory. A reachable set of the error state is computed on-line based on bounded disturbance. The constraints of nominal state and input are obtained, which ensure state and input constraints are satisfied in presence of disturbances and uncertainties. Then the preview controller is proposed to improve the tracking performance. Simulation using a passenger vehicle is conducted on different road friction coefficient, mass uncertainty and road bank angles, respectively, which are all treated as additive disturbance. Simulation results show the effectiveness of the proposed framework under the test of double lane change.