Variable neural adaptive robust control and observation of uncertain systems

One of the essential elements in many controller design processes is a mathematical model of the dynamic system to be controlled. However, it is often difficult to obtain a good mathematical model due to unknown system parameters or dynamics. The objective of this research is to develop practical high performance adaptive robust controllers and observers for uncertain systems whose mathematical models are not exactly known. The essential ingredient of the proposed controllers and observers is a practical function approximator. In the first stage, a novel variable-structure radial basis function (RBF) network is proposed as a self-organizing approximator for a class of continuous-time uncertain systems. In the second stage, adaptive robust state and output feedback control architectures are proposed for the output tracking control of both single-input singleoutput and multi-input multi-output uncertain systems. In the third stage, a variable neural adaptive robust observer is proposed for the state estimation of a class of multiinput multi-output uncertain systems. The controllers and observers proposed in the second and the third stages incorporate the aforementioned variable-structure RBF network for self-organizing approximation of unknown system dynamics. The structure variation of the RBF network is taken into account in the stability analysis for the proposed variable neural adaptive robust controllers and observers. A number of numerical simulations are performed to demonstrate their efficacy