Tube-based Robust Model Predictive Control for Autonomous Vehicle With Complex Road Scenarios

This paper presents a Tube-based Robust Model Predictive Control (Tube-RMPC) strategy for autonomous vehicle control, designed to address model parameter uncertainties and variations in road-tire adhesion coefficients under complex driving conditions. The proposed approach enhances the representation of vehicle dynamics by introducing a unified vehicle-tire modeling framework, capturing nonlinear characteristics more effectively. To facilitate real-time implementation, the model is systematically linearized and discretized, ensuring computational efficiency.

A Tube-RMPC strategy is then formulated to improve the vehicle’s adaptability to uncertainties in road surface adhesion. By constructing a robust invariant tube around the nominal trajectory, the proposed controller effectively mitigates disturbances, ensuring stability and robustness under varying road conditions. Additionally, a co-simulation platform integrating CarSim and Simulink is employed to validate the proposed method. Simulation results demonstrate that Tube-RMPC improves path-tracking performance, reducing maximum tracking error by up to 9.17% in S-curve and 2.25% in double lane change, while significantly lowering RMSE and enhancing yaw stability compared to MPC and PID.