Iterative Learning of Dynamic Inverse Filters for Feedforward Tracking Control

A novel method to construct dynamic inversion compensation is proposed for feedforward tracking control. In contrast to common approaches involving parametric system identification followed by inversion synthesis, in this article we apply iterative learning control to track an impulse signal, where the converged control input is directly used to construct the inverse filter. The method is applicable to multivariable systems without a diagonalization process. The proposed method is implemented on a linear motor and an active magnetic bearing system, respectively. The experimental results are presented to demonstrate the feedforward tracking performance.