Self-tuning multivariable fuzzy and neural control using genetic algorithms

In this paper genetic algorithms (GAs) are applied to improve the performance of fuzzy and neural multivariable controllers. In the case of fuzzy multivariable controllers the GA task is to determine an appropriate compensation element which will perform partial decoupling between system inputs and outputs thus permitting the application of a decentralized control scheme. A novel robust fuzzy logic controller is then applied to each one of the SISO subsystems. The neural multivariable controller is based on the Gawthrop-type MIMO controller but now a neural network is used for the prediction of the system outputs. A genetic algorithm decides how much the control signal should contribute in the cost function thus succeeding the optimal compromise between system stability and zero steady-state error.