Neural network adaptive control and repetitive control for high performance precision motion control

This paper presents neural network adaptive control and repetitive control for precision motion control applications. Two neural controllers are applied in the proposed adaptive control scheme. The first one is a classic supervised neural controller using radial basis function network as a performance baseline for motion control. To eliminate the non-zero periodic error coming from the deterministic reference signals, another neural controller is added as a second controller which contains a discrete-time repetitive controller by including the deterministic internal model into the neural controller. The proposed neural network adaptive control scheme was applied to a piezo-actuated system for tracking periodic profiles. The experimental results obtained demonstrate that the motion control performance is improved by the neural-repetitive controllers.