Adaptive control of nonlinear uncertain systems using neural network

Abstract

The robust neural adaptive control scheme is developed for a class of nonlinear MIMO uncertain systems. Two types of uncertainties are considered: parametric uncertainty and unknown nonlinearities. In the control procedure, neural networks are implemented to estimate the uncertain system owing to parametric uncertainty and robust compensating controllers are designed in H <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</inf> sense for attenuating the unmatched nonlinearities, in which no property of uncertainties is assumed and needed for design, only its possible bound which may not unique, is used for analytical purposes. And also relaxes some of restrictive assumptions that are usually made in neural adaptive control schemes, one such assumption is the requirement of a known bound on the network reconstruction error. It is shown that the proposed control is continuous, guarantees global stability and the H <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</inf> performance index. Extensive simulations on the tracking control of a two-link rigid robotic manipulator verify the effectiveness of the design methodology and controller.