Lateral control for autonomous land vehicles via dual heuristic programming

Abstract

In this paper, to improve the lateral control performance of autonomous land vehicles (ALVs) under different longitudinal velocities, a novel learning-based lateral control method based on the dual heuristic programming (DHP) algorithm is presented. A new way to calculate the lateral control errors based on the geometric relationship between the vehicle and the path is utilized. To minimize the lateral control errors of ALV, the lateral control problem is modelled as a Markov decision problem (MDP). To approximate the optimal control policy of the MDP, a learning controller based on the DHP algorithm is designed, where the critic is used to approximate the derivative of the value function and the actor is utilized to improve the control policy based on the output of the critic. The inputs of the critic and actor are comprised of the lateral control error and the vehicle's longitudinal velocity, which makes the proposed method be effectively adaptive to different longitudinal velocities. Simulation results demonstrate the proposed lateral control scheme has advantages over widely used lateral control methods, such as the pure pursuit, PD control and Stanley methods.