Self-learning fuzzy controller
Yoshiyuki Yamamoto, H. Fujikawa, Shinichi Yamada, K. Shida · 2003
A novel design for a fuzzy controller with a self-learning function is proposed. It is very difficult to determine the optimal condition of a fuzzy controller, since the control performance of the controller depends on both the membership functions and the control rules. The proposed algorithm, which can automatically determine the suitable condition of the controller, resolves this difficulty. In the algorithm, the control rules of the fuzzy controller are modified according to the time response of plant output signals. The validity of this algorithm is evaluated through numerical experiments. The proposed design method is shown to have the following advantages: an arbitrary initial condition of the control rules is allowable: the design is easier than with a conventional PI (proportional plus integral) controller: and even if the plant parameters are uncertain, it is possible to design a suitable controller.>