Adaptive neuro-controller design based on MLP network

Abstract

Back Propagation (BP) algorithm is the most commonly used algorithm for training artificial neural networks. But, it suffers from extensive computations, relatively slow convergence speed and possible divergence for certain conditions. The main objective of this paper was to compare the performance of BP algorithm and Recursive Least Square (RLS) algorithm for Adaptive Neuro-Controller (ANC). These algorithms are used to update the parameter of the ANC. A neural network model, called Multi Layered Perceptron (MLP) network is used for this ANC. The Model Reference Adaptive Control (MRAC) is used to generate the desired output path and to ensure the output of the controlled system follows the output of reference model. In this paper, the comparison between two algorithms is based on the convergence speed and robustness of the controller. These controllers have been tested using a linear and a nonlinear plant with several varying operating conditions. The simulation results show that RLS algorithm have better performance compared to BP algorithm.