Autonomous Vehicle Path Tracking: Stochastic Tube Model Predictive Control With Covariance Steering and Discounted Chance Constraints

The path tracking performance of autonomous vehicles degrades due to model uncertainties and external disturbances. Tube-based Model Predictive Control (MPC) has been regarded as an effective approach to address this issue. However, the over-conservative nature of tube-based MPC has limited its broader application. this paper proposes a stochastic tube MPC method subjected to state and control chance constraints. The probability distribution of system uncertainty is assumed as a prior knowledge. Leveraging covariance steering theory, the dynamic feedback gain mechanism of the perturbed system is designed accounting for the variability in the probability distribution of disturbances. Instead of robust positive invariant sets, the concept of probabilistic reachable sets is used to determine the constraint-tightening and convert chance constraints into deterministic ones. Furthermore, the discounted state chance constraints are incorporated in the cost function of nominal system, prioritizing near-term tracking performance over long-term behavior. A joint CarSim-Simulink simulation demonstrated that the proposed stochastic tube MPC method reduces conservatism and enhances the tracking performance. Further verification through hardware-in-loop (HIL) experiment confirmed the effectiveness of the proposed approach.