Abstract:
Adaptive neural networks control can be defined as a method of robust control, in which neural networks systems are used to approximate unknown functions, some of the system parameters are adapted online to ensure the stability of the closed loop system. The main objective In this thesis, An adaptive neural networks control approach is developed and proposed for a class of single-input single-output (SISO), and multi-input multi-output (MIMO) nonaffine nonlinear dynamic. Based on the implicit function theory, the existence of an ideal controller, that can achieve control objectives, is firstly shown. Then, Taylor series expansion is employed to transfer the normal nonlinear system into the standard affine form in the neighborhood of the ideal. Based on this elegant result , we have proposed adaptive neural controller direct and indirect for a class of single-input single-output (SISO) nonlinear systems, and multi-input multi-output (MIMO) nonlinear systems, The common feature between all developed adaptive controllers is the use of neural networks systems, that are updated on-line, with the purpose of producing approximations of the system‟s dynamics, in some cases (indirect methods), or of some unknown stabilizing controllers, in others (direct methods). In addition, stability and robustness analysis of the proposed control schemes are performed by using the Lyapunov synthesis method, and for each scheme, simulation results are given to highlight its performance.
