The least trained samples method to tune knowledge bases of CMAC based FLC

The least trained samples method to tune fuzzy knowledge bases of CMAC based FLC is proposed in this paper. The fuzzy knowledge bases can be tuned by CMAC through experienced output of the proposed structure. The scheme for least trained samples is also utilized in the training process so that the trained times decreases. We demonstrate that the proposed structure has wide learning capability with few computations because CMAC possesses local generalization ability and needs only a small number of activated units (memory cells) in both training and control processes. Simulations are applied to solve the rear-loading truck problem, which employs the least trained samples to validate fast learning and accurate approximation by the proposed method.