A Neural-Network-Based Equivalent-Input-Disturbance Approach

In this paper, a neural-network-based equivalent-input-disturbance (EID) approach is developed for a servo control system to improve the disturbance-rejection performance, where the radial basis function neural network (RBFNN) is used to approximate an EID of unknown and mismatched disturbances. First, by treating the Luenberger state observer as an ideal dynamics of the real system, the comparative output between the two systems is used to construct the RBFNN-based EID estimator. Then, by exploiting the two-degree-of-freedom property of the EID approach, the system analysis and design of the closed-loop control system is simplified to those of the tracking subsystem and the disturbance-rejection subsystem. Further, a design algorithm together with some guides is given. Finally, a case study of a motor driver system demonstrates the validity of the developed method.