Nonlinear H∞ path following control for autonomous ground vehicles via neural network and policy iteration algorithm

Abstract

This paper addresses the crucial issues of autonomous ground vehicles (AGVs), that is, smooth and accurate path following, especially when input constraints and external disturbances exist. A nonlinear H∞ control scheme is proposed based on the neural network (NN) and policy iteration (PI) algorithm. Firstly, a nonlinear system model for path following of AGVs is constructed, which takes vehicular nonlinear cornering behavior into account. Then, a nonlinear H∞ controller is designed based on the Hamilton-Jacobi-Isaacs (HJI) equation. Wherein, the established NN is leveraged to approximate the HJI equation’s performance function, and the PI algorithm is developed to learn the solution to the HJI equation. Finally, the convergence and closed-loop stability of the proposed strategy are demonstrated. Simulation results show that the proposed controller can significantly improve the response speed, driving smoothness and accuracy of path following in high-speed driving condition with strong robustness against external disturbances.