Multilevel qualitative and numerical optimization of fuzzy controller

Often the optimization of fuzzy controllers is done using new global optimization techniques like genetic algorithms, simulated annealing or benefiting from learning capabilities of neural networks. In this paper, we restrict ourselves to the use of classical techniques like gradient algorithms and Rosenbrok´s algorithms associated in a multilayer structure to a qualitative (symbolic) supervision level. Some simulation examples (linear, nonlinear, stable, unstable systems) show the efficiency of the proposed methodology