Neural Controller Utilizing Genetic Algorithm Technique For Dynamic Systems
Loading...
Date
2009-10
Authors
A Ali, Marwan
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
This research presents a method of learning multilayer Neural Network (NN) using Genetic Algorithms (GAs) techniques. The evolutionary techniques based on GAs are studied and employed for the Model Reference Adaptive Control (MRAC) scheme of different plants. The GAs are used for selecting an optimal number of hidden nodes for the neural controller, as well as training the network to minimize the error between the output of the plant and the output of the model reference. The simulation examples demonstrate how the hidden nodes are adapted through the generation until they reach their optimal integer number, which depends on the complexity of the controlled plant. The real-coding operators of GAs have been used in this work because of the limitations of traditional binary coding. Moreover, a hybrid selection method plus elitism strategies are used for reproduction process of the GAs. A comparison between the proposed neural controllers with a classical genetically tuned, Proportional Derivative Integral (PID) controller is made, in which the final outcome is used as a feedback controller based on model reference. A comparative analysis is also made based on the speed of convergences, the tracking ability of the desired model reference and robustness of these controllers to output disturbances. Based on simulation results, it is concluded that the proposed neural controller is better than PID controller in both: robustness and tracking ability.
Description
Keywords
Neural Controller Utilizing Genetic Algorithm Technique , Dynamic Systems