Fuzzy CMAC with Online Learning Ability and Its Application

The binary behavior of activation function in receptive field of conventional cerebellar model articulation controller (CMAC) affects the continuity of the network output. In addition, the original learning scheme of CMAC may corrupt the previous learning data. A control scheme, which parallely combines the fuzzy CMAC (FCMAC) and PID, is proposed in the paper. The weights are updated according to the credits which are assigned to the hypercubers according to their learning histories and fuzzy membership degrees. The FCMAC is powerful in control time-varying processes due to the online learning ability of the FCMAC. Experimental results of temperature control have shown that the FCMAC with online learning ability can accurately follow the control trajectory and reduce the tracking errors.