Self-Tuning Adaptive Control using Fourier Series Neural Networks

Abstract

A neural network architecture, called the Fourier Series Neural Network (FSNN), has been developed [1] for modeling unstructured dynamic systems using system input and output frequency spectrums. This paper addresses the issues concerning on-line implementation of self-tuning adaptive control using the FSNN as an estimator. An underlying controller design method based on the estimation of the system frequency response is proposed in the principle of the laglead compensation. The performance of this Neuro-Self-Tuning Regulator (NSTR) is evaluated using the performance parameters from the frequency domain such as the system bandwidth, phase margin and the gain margin. Simulations for the evaluation of the NSTR were conducted and the results are discussed.