Tube-Based Model Predictive Control for Robust and Stable Lateral Path Tracking of Autonomous Vehicles

Abstract This article presents a tube-based model predictive control (tube-MPC) framework for lateral path tracking of autonomous vehicles under model uncertainty and bounded disturbances. Based on a nominal MPC model, the controller incorporates a prestabilizing feedback gain and a disturbance-invariant set to improve disturbance rejection and constraint handling. Compared with the conventional MPC that assumes perfect model accuracy, the tube-MPC introduces a disturbance-invariant tube around the nominal trajectory, within which all perturbed system states are forced to evolve. Under the adopted modeling assumptions, this formulation is intended to improve recursive feasibility and constrain satisfaction in the presence of bounded disturbances and parameter variations. A matlab–carsim–integrated co-simulation platform is established. Compared with conventional PID and nominal MPC controllers, the proposed tube-MPC achieves lower tracking errors in lateral velocity, yaw rate, yaw angle, and lateral position under the considered simulation scenarios. In the single-lane-change test, it reduces the lateral position mean absolute error (MAE) from 0.152 m (PID) and 0.081 m (MPC) to 0.049 m—an improvement of 81.3% and 57.1%, respectively—with integral of mean square error (IMSE) = 0.0015 and PATEST = 97.2%. In the double-lane-change case, MAE = 0.052 m and PATEST = 96.7%, corresponding to reductions of 80.3% and 58.2%. Overall, the results indicate that tube-MPC improves lateral tracking performance and disturbance tolerance under the considered simulation conditions, keeps the lateral deviation within approximately ±0.05 m in the tested cases, and demonstrates feasibility in simulation, although broader validation is still needed.