Simulation Experiment Of Model Free Adaptive Fuzzy Controller

The existent controllers are either designed based on the mathematical model of the controlled object, or designed based on the fuzzy experience of experts. But in practice it is difficult to get control experience in some control process. Self-learning and getting the fuzzy rules became one of the important problems in the area of fuzzy control.This paper presents a new controller called Model Free Adaptive Fuzzy controller, which can obtain fuzzy rules without the mathematical model of the controlled object. This kind of controller uses a multilayer perception (MLP) as the identifier to learn the outside property of the controlled object online, and a Fuzzy MLP (FMLP) which automatically modifies its network parameters with the information of the controlled object supplied by the identifier. The network can converge much faster with the FMLP as the controller, which can use the linguistic information about the experience of the control process. This new type of controller structure modifies its parameters of the identifier and the controller, which is responsible for the advantage of strong adaptability, as well as the strong robustness. It is proved that this new structure of controller has the ability of properly controlling simple nonlinear systems, which can be seen from the experiments carried on some simulated systems.