Vehicle trajectory tracking control based on variable horizon of model predictive control

Traditional Model Predictive Control (MPC) with fixed horizon cannot balance computational burden and tracking performance under complex conditions. To improve autonomous vehicle path-tracking performance, this study introduces a dual-horizon adaptive MPC strategy, optimizing both control and prediction horizons. This helps to reduce trajectory tracking errors while optimizing computing resources dynamically. First, we establish a vehicle dynamics model and a path-tracking model. Subsequently, the variations in the control and prediction horizons are determined via fuzzy control. The used integration scheme comprehensively considers three key parameters: vehicle speed, lateral deviation, and yaw rate, where the yaw rate measurement is used Kalman filter-based observation. The MPC controller first adjusts the control and prediction horizons. Next, it calculates the desired front-wheel steering angles and direct yaw moment. High-speed condition simulations are performed using MATLAB/Simulink and CarSim co-simulation platforms. Results demonstrate enhanced trajectory tracking accuracy with the variable-horizon MPC approach. Simultaneously, the method improves the computational efficiency, saving about 15.5% of the computation time on average.