Hierarchical fuzzy neural networks for nonlinear dynamic systems

Abstract

In this thesis, the combination of hierarchical fuzzy logic system with neural network methodology for the adaptive control of nonlinear dynamic systems is addressed. The philosophy behind this study is that a hierarchical fuzzy neural network control system is designed to combine the advantages of the fuzzy inference and neural network methodology. Original contributions of this dissertation include that a theoretical framework of hierarchical fuzzy neural network is proposed based on the fuzzy hierarchy error approach algorithm and the effectiveness of nonlinear control using the hierarchical fuzzy neural network is investigated. Firstly, the theory of hierarchy in neural network and fuzzy logic system is introduced and the functional equivalence relation between the radial basis function (RBF) network and the fuzzy inference system is discussed, and a simplified description model of the proposed fuzzy neural network is presented. Secondly, according to the basic optimization principle of artificial neural networks, a novel neural network model for solving the quadratic programming problem is proposed. Thirdly, a hierarchical fuzzy neural control scheme is discussed. Then, a structure of the hierarchical fuzzy neural network, which is composed of an antecedent network and a consequent network, is explained. In the network learning and training phase, a concise and effective algorithm based on the fuzzy hierarchy error approach (FHEA) is formulated to update the parameters of the network. A model reference adaptive control structure incorporating the proposed fuzzy neural network is studied. Finally, stable fuzzy neural tracking control of a class of unknown nonlinear systems based on the hierarchy approach is illustrated. The adaptive fuzzy neural controller is constructed from the hierarchical fuzzy neural network with a set of fuzzy rules. The corresponding network parameters are adjusted on-line for the purpose of controlling the plant to track a given trajectory. The stability analysis of the unknown nonlinear system is discussed based on Lyapunov's principle.