RBF-based supervisor path following control for ASV with time-varying ocean disturbance

Abstract

A robust model-free path following controller is developed for autonomous surface vehicle (ASV) with time-varying ocean disturbance. First, the geometrical relationship between ASV and virtual tracking point on the reference path is investigated. The differentiations of tracking errors are described with the relative motion method, which greatly simplified the direct differential of tracking errors. Furthermore, the control law for the desired angular velocity of the vehicle and virtual tracking point are built based on the Lyapunov theory. Second, the traditional proportional-integral-derivative (PID) controller is developed based on the desired velocities and state feedback. The radial basic function (RBF) neural network taking as inputs the desired surge velocity and yaw angular velocity is developed as the supervisor to PID controller. Besides, RBF controller tunes weights according to the output errors between the PID controller and supervisor controller, based on the gradient descent method. Hence, PID controller and RBF supervisor controller act as feedback and feed forward control of the system, respectively. Finally, comparative path following simulation for straight path and sine path illustrate the performance of the proposed supervisor control system. The PID controller term reports loss of control even in the unknown disturbance.