Path tracking controller design for autonomous vehicle based on robust tube MPC

Abstract

A robust tube MPC controller based on tube-division with a linear time-varying (LTV) model is proposed for autonomous vehicle path tracking. To reduce the conservativeness, a novel offline method is designed to calculate the tubes by dividing the original N-steps invariant sets into sequences of tighter candidate tubes. The propagation limits of the vehicle time-varying parameters within a prediction horizon are used in the division to ensure each candidate tube contains any state trajectory starting at its origin. A corresponding tube will be selected instead of being calculated online at each sampling instant in terms of vehicle states, which makes a more efficient online computation. The results of the simulation show the improved performance of the proposed robust tube MPC controller.