An approximate dynamic programming approach for path following control of an autonomous vehicle

Intelligent control of autonomous vehicles has been an important research topic due to the model uncertainties and complexities of vehicle dynamics. In this paper, we proposed an approximate dynamic programming (ADP) approach for path following control of an autonomous vehicle. The idea is to use the Dual Heuristic Programming (DHP) algorithm, which is an efficient class of ADP methods, to directly control the steer angle of the vehicle´s wheels. In order to perform data-driven simulation studies, we established the kinematic model of the vehicle and the model of the steering wheels. Then, the DHP algorithm was designed for path following control by using a Markov decision process model. Simulation results illustrated that the proposed method can achieve good performance for the path following control problem.