| The design of stabilizing controllers for nonlinear plants with unknown nonlinearities is a challenging problem. The inability to identify exactly the nonlinearities on-line or off-line motivates the design of stabilizing controllers based on approximations or on approximate estimates of the plant nonlinearities. The price paid in such case, could be lack of theoretical guarantees for global stability, and non-zero tracking or regulation error at steady state.; In this dissertation several nonlinear robust adaptive control schemes are designed and analyzed for a class of single-input single-output (SISO) nonlinear systems, a class of multi-input multi-output (MIMO) nonlinear systems, and a class of large-scale MIMO nonlinear systems with unknown nonlinearities. We use single layer neural networks as approximation models of the unknown nonlinearities. The proposed control schemes provide a general approach to bypass the stability problems where the estimated plants become uncontrollable without any restrictive assumptions. The controllers are continuous and guarantee closed loop semi-global stability and convergence of the tracking error to a small residual set. The semi-global stability is characterized by a region of attraction whose size depends on the compact sets used to approximate the unknown nonlinearities. The size of the residual tracking error can be specified a priori and is guaranteed by choosing certain design parameters. Several procedures for choosing these parameters are presented. The properties of the proposed control schemes are demonstrated using several examples that include control of a hypersonic aircraft. |
