Model reference adaptive neuro-controller with on-line parameter estimation

Abstract

In this paper, a Model Reference Adaptive Neuro- Controller is developed, in which the error between the outputs of the plant and the reference model is used to adapt the controller parameters. The Model Reference Adaptive System (MRAS) was originally proposed to control a time varying systems where the performance specifications are given in terms of a reference model. A neural network model, called Hybrid Multi Layered Perceptron (HMLP) network will be used for this Adaptive Neuro-Controller (ANC). The Recursive Least Square (RLS) algorithm will adjust the ANC parameters to minimize the error between the plant output and the model reference output. The performance of the HMLP network is compared with Multi Layered Perceptron (MLP) networks. These networks have been tested using a linear and nonlinear plant with some variations in operating conditions. The results for both plants sets indicated that HMLP network gave significant improvement over standard MLP network.