Abstract: A digital tracking controller based on Cerebellar Model Articulation Controller(CMAC) neural network is presented, aiming at some non-linear and uncertain factors in the high precision servo systems. This controller consists of feed forward controller, Pulse Doppler (PD) controller and CMAC. It absorbs fast-learning and precise approaching advantages of CMAC to enhance traditional control method. Data form experiments demonstrate that the amplitude error of CMAC method is more precise compared to PD method by one order of magnitude. Meanwhile, it has good inhibitory effect on distortion caused by friction. The proposed controller is shown to be effective and valid, and has good referenced value to improve the dynamic tracking performance of high precise servo system.

Key words: CMAC, compound control, Least Mean Square(LMS), servo system, friction compensation