Optimization of fuzzy controllers by neural networks and hierarchical genetic algorithms

This paper deals with the optimization of fuzzy controllers using neural networks and hierarchical genetic algorithms. The method combines the training advantage of neural networks, and the aptitude to find a global optimum offered by genetic algorithms. The fuzzy controller is implemented as a neural network where each layer represents a part of the fuzzy controller. The training process consists in optimizing the connection weights which code the various parameters of the controller. Once the training is finished, the parameters coded chromosomes take part in the evolution process using selection, crossover and mutation. This hybridization is applied to nonlinear system.