Adaptive model predictive control for robust lateral motion tracking of semi-autonomous vehicles with dynamic parameter variation

This study presents a novel adaptive model predictive control (AMPC) framework for robust lateral motion tracking in semi-autonomous vehicles operating under dynamically uncertain conditions. Traditional controllers often struggle to maintain stability and accuracy in the presence of nonlinear vehicle dynamics, time-varying parameters, and external disturbances. The proposed AMPC system integrates real-time parameter estimation via recursive least squares with a predictive optimization structure, allowing continuous adaptation to variations in vehicle mass, speed, and tire-road friction. A comprehensive simulation environment developed in MATLAB/Simulink was used to evaluate the AMPC across multiple scenarios, including aggressive lane changes, crosswind disturbances, and low-friction conditions (friction coefficient μ = 0.4). Quantitative comparisons with conventional model predictive control (MPC) and linear quadratic regulator (LQR) controllers demonstrate that AMPC achieves a 43% reduction in lateral tracking error and a 37% improvement in yaw angle root mean square error (RMSE), maintaining peak yaw errors below 0.275 radians even under severe disturbances. Steering control signals remain smooth and within actuator limits, with maximum steering rates under 0.48 rad/s. Additionally, a human-in-the-loop simulation confirmed the controller's ability to handle delayed driver intervention (1-3 s) without compromising vehicle stability or trajectory tracking. These results validate AMPC's superior robustness, adaptability, and real-time performance in managing complex lateral control tasks. The framework provides a scalable solution for enhancing safety and reliability in shared-control driving environments, addressing both technical and human-centric aspects of advanced driver assistance systems (ADAS).