Adaptive Neural Network Control for a Class of MIMO Uncertain Pure-Feedback Nonlinear Systems

Abstract

In this paper, robust adaptive neural network control is investigated for a class of multi-input-multi-output (MIMO) pure-feedback nonlinear system with unknown nonlinearities. The unknown nonlinearities could be come from unmodeled dynamics, modeling errors, or nonlinear time-varying uncertainties. Based on the backstepping design technique and the universal approximation property of the neural network (NN), robust adaptive control is synthesized by employing a single NN to approximate the lumped uncertain nonlinearities. The proposed control can eliminate the circularity problem completely, and guarantees semiglobal uniform ultimate boundedness (SGUUB) of all the signals in the closed-loop and convergence of the tracking error to an arbitrarily small residual set.