Fuzzy control is one of the most successful field for research in the applications of fuzzy set theory. Fuzzy logic controller is designed with the human's intuition and experiences. So, fuzzy logic controller has the function of human's decision. However, in the systems having complex dynamics, it is nearly impossible to determine the fuzzy control rules for exact performance only with the above method. To design more intelligent controller, it is required to make use of the functions of human's learning and adaptation.
Fuzzy learning controller organizes the fuzzy control rules by itself with the abilities of learning and adaptation. In this paper, it is proposed a new method of learning algorithm determining change in control input of fuzzy learning controller in nonlinear unstable systems. The state feedback gains are used from the linearized system of the nonlinear system. Through this method, it is easy to determine the learning rates only with two parameters.
Acrobatic robot system is selected as an example (one-input two-output unstable system). It is guaranteed the good convergence and confirmed the performance of fuzzy learning controller is better than that of linear controller through the simulation. Fuzzy learning controller is implemented through the experiment, and it is compared the responces of several controller(fuzzy learning controller, optimal controller, linear controller with pole-placement). And it is carried out the experiment with disturbance.
