The well-known capacity of the neural networks structures to approximate functions and nonlinear systems, to any degree of accuracy, on a compact domain, has motivated its use in actual control problems where the presence of uncertainties is inevitable. Neural networks models can be used to parameterize unknown nonlinearities, hence, the linear parameterization assumption typically imposed in most nonlinear adaptive control designs can be avoided. Several works, based on Lyapunovs stability theory have demonstrated the feasibility, stability properties and performance characteristics of such schemes. However, most of such controllers are only applicable to some classes of SISO uncertain nonlinear systems. Hence, in this dissertation we propose neural networks based adaptive algorithms for identification, adaptive observation, state feedback and output feedback control for more general classes of nonlinear systems. Initially, we introduce a nonlinear adaptive observer for MIMO unknown general nonlinear systems. The proposed observer uses linearly parameterized neural networks whose weights are adaptively adjusted in a suitable form to guarantee stability for the state and weight estimation errors. No strictly positive real (SPR) assumption on the output error equation is required for the construction of the proposed observer. Next, we consider the adaptive tracking control of unknown general nonlinear systems with full state measurement. Bounded linearly parameterized approximators for adaptive compensation of unknown nonlinearities and usual Lyapunov arguments are used to design two controllers that guarantee semi-globally uniform ultimate boundedness (UUB) for all signals in closed loop. In the first scheme, where only adaptive control action is used, it is shown that all signals in closed loop are UUB. The second controller preserves the stability properties of the first, and in addition, by incorporating a variable proportional control action, it is shown that the tracking error converges to a residual set whose size can be arbitrary pre-specified by the designer without small approximation errors and without high gains in the feedback loop. The proposed controllers have advantageous features, such as robustness and the possibility of regulating the control signal effort and rate. Finally, we develop an output feedback tracking controller for a class of MIMO unknown nonlinear systems. By using an adaptive observer, which is similar to the previously introduced, and usual Lyapunov arguments, a controller is designed. It is proved that the output converges to a neighborhood of the desired trajectory. Moreover, the proposed controller enables control effort manipulation in order to satisfy physical limitations of the actuators.
Identifer | oai:union.ndltd.org:IBICT/oai:agregador.ibict.br.BDTD_ITA:oai:ita.br:2553 |
Date | 00 December 2002 |
Creators | José Alfredo Ruiz Vargas |
Contributors | Elder Moreira Hemerly |
Publisher | Instituto Tecnológico de Aeronáutica |
Source Sets | IBICT Brazilian ETDs |
Language | English |
Detected Language | English |
Type | info:eu-repo/semantics/publishedVersion, info:eu-repo/semantics/doctoralThesis |
Format | application/pdf |
Source | reponame:Biblioteca Digital de Teses e Dissertações do ITA, instname:Instituto Tecnológico de Aeronáutica, instacron:ITA |
Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0016 seconds