Robust Adaptive Neural Control Fora Class Of Nonlinear Systems

Adaptive control using on-line function approximators for feedback linearizablesystems has proven to be a very e?ective way to design controllers based on ap-proximate knowledge of the system dynamics. Practical implementation of suchcontrollers on the aircraft have demonstrated their stability and performance char-acteristics, as well as superior fault tolerance when there is redundancy in the control.These real world experiments provide great momentum for theoretical research innonlinear adaptive control systems using neural networks to approximate unknownfunctions.In this thesis we present adaptive neural network state feedback control waysfor nonlinear systems with virtual control coe?cients functions, unknown nonlinearfunctions, unknown parameters and stochastic disturbances.In Chapter 1 of this paper, we introduce the background of the adaptive controland neural networks.In Chapter 2 and Chapter 3, adaptive neural control is investigated for a classof nonlinear time-delay systems with unknown virtual control coe?cients, unknownnonlinear functions and unknown disturbances by using appropriate Lyapunov-Krasovskii functionals, neural networks and Nussbaum functions. It is provedthat the proposed design method is able to guarantee semi-global uniform ultimateboundedness of all the signals in the closed-loop system, and the tracking error isproven to converge to a small neighborhood of the origin. In Chapter 2, we introducethe problem formulation and preliminaries and present in detail the state-feedbackcontrol design for the ?rst order systems. Then in chapter 3, we expanded thesecond chapter's method, studied the robust adaptive control design for nth-order systems by adaptive backstepping. Simulation results are provided to illustrate theperformance of the proposed approach.In Chapter 4, adaptive neural control is investigated for a class of nonlinearstochastic systems with stochastic disturbances and unknown parameters. Underthe condition of all system states being available for feedback, by employing thebackstepping method, a suitable stochastic control Lyapunov function is then pro-posed to construct an adaptive neural network state-feedback controller, and un-known parameters are reasonably disposed.It is shown that, the closed-loop systemcan be proved to be global asymptotically stable in probability .The simulation re-sults demonstrate the e?ectiveness of the proposed control scheme.