The PID controllers are widely used in many engineering fields. The existence of the control disturbance and measurement error may greatly reduce the controller's stabilization, and even lead the control to failure. In this paper, we proposed a novel PID controller, which use the cerebellar model articulation controller (CMAC) neural network to compensate the control error online and use the Kalman filter to diminish the affection of control disturbance and the measurement error. Simulation results show that the proposed controller work very well.