Neural-based direct adaptive control for a class of general nonlinear systems

Abstract

A direct adaptive controller based on high-order neural networks (HONNs) is presented to solve the tracking control problem for a general class of unknown nonlinear systems. The plant is assumed to be a feedback linearizable and minimum-phase system. Firstly, an ideal implicit feedback linearization control (IFLC) is established using implicit function theory. Then a HONN is applied to construct this IFLC to realize approximate linearization. The proposed controller ensures that the closed-loop system is Lyapunov stable and that the tracking error converges to a small neighbourhood of the origin. The requirements of an off-line training phase and the persistant excitation condition are eliminated. Simulation results verify the effectiveness of the proposed controller and the theoretical discussion.