Automatic tuning of fuzzy controller parameters based on RBF neural network

Parameter selection for fuzzy controller often depends on expert's knowledge to controlled objects. The quality of a fuzzy controller can be drastically affected by the choice of quantization factor and proportion parameter. Thus, some measures for tuning fuzzy logic controller should be taken. RBF (radial basis function) neural network has the approximation ability and have been proved to approximate any nonlinear function. In this paper, a new method to tune parameter of fuzzy controller is proposed on the base of two RBF neural networks. The first RBF neural network is used to approximate the part of fuzzy inference and get the Jacobian information between input and output of fuzzy controller. The second RBF neural network identifies the controlled model and obtains Jacobian information about output of fuzzy controller and that of controlled system, which is needed in the process of optimizing quantization and proportion parameters. The simulation results demonstrate that this method has perfect dynamic performances, quick response speed.