A model‐free adaptive control algorithm based on reinforcement learning and its application in tracking control of unmanned ground vehicles

Abstract This paper proposes a novel reinforcement learning‐based model‐free adaptive control (RL‐MFAC) algorithm for the path tracking of autonomous vehicles. By using dynamic linearization techniques, lateral and longitudinal dynamic vehicle processes are dynamically linearized, and then an MFAC controller is designed. To optimize the MFAC controller parameters for better performance, RL, with its powerful self‐learning capabilities, is introduced. The proposed control strategy offers rigorous stability. Through strict semi‐physical simulation validation on the CarSim‐MATLAB/Simulink software integration platform, this method demonstrates good performance in vehicle longitudinal and lateral control simulations, validating the effectiveness of the algorithm.