Path Tracking Control Based on an Adaptive MPC to Changing Vehicle Dynamics

Abstract

In this paper, an adaptive Model Predictive Controller (MPC) is proposed as a solution for path tracking control problem for autonomous vehicles. The effect of feeding the MPC with a continuously changing vehicle’s mathematical model is studied, so that the controller becomes more adaptable to changing parameter values accompanied with instantaneous states. The proposed MPC is compared with both Stanley controller and a similar MPC that uses a fixed vehicle model. The performance is measured by the ability to minimize both lateral position and heading angle errors. A dynamic bicycle model for the vehicle is deployed in the MPC and the controllers are simulated in CarSim-MATLAB/Simulink co-simulation environment using three common maneuvers: S-Road, double lane change and curved road. Results show that the proposed controller gives better tracking performance than the two others with minimal instantaneous and root mean square RMS errors.