ON-LINE RULE WEIGHTING FOR PID-TYPE FUZZY LOGIC CONTROLLERS USING EXTENDED KALMAN FILTER

In this study, we propose a method for online tuning of fuzzy rule weights of PID-type fuzzy logic controller via Extended Kalman Filter. In the application phase of Extended Kalman Filter to this new online parameter optimization setting, suitable state and observation vectors is needed; in this framework, the rule weights of the rule base are defined as the states and the output of the fuzzy system is defined as the observation vector. We apply the weight adjustment not to the consequent of the rule but instead to the complete rule. The effectiveness of the proposed on-line weight adjustment method is then demonstrated on linear and non-linear systems by simulations. The performance of the proposed tuning method is evaluated according to four performance measures and performance amelioration is observed in all measures. Moreover, the proposed on-line tuned PID-type FLC can handle the noise more successfully than the conventional one.