Abstract
It is necessary to tune inference rules whenever fuzzy reasoning is employed to control an object. Some self-tuning methods which can tune inference rules automatically were proposed. But, these conventional methods don't have sufficient generalization capability. We propose a new selftuning method of fuzzy inference rules using delta rule, which is a learning algorithm of neural network. This method have high generalization capability, because this method optimizes antecedent parts which are arranged each rule. In this paper, formulation of the proposed method is described first, and its applications to some numerical examples and a moving obstacle avoidance are reported to show its advantages.
