Design of feedforward filters for improving tracking performances of existing feedback control systems

In this paper, we propose a design procedure for feedforward filters that can be used to improve the tracking performance of an existing closed-loop system. Our frequency domain approach does not require a parametric model but only the frequency response of the closed-loop system, which can be measured accurately. It allows the tradeoff between model-matching and control efforts through the selection of weighting filters. Time-domain specifications can be met indirectly by various relations with frequency domain specifications. The algorithm is intuitively simple and easy to compute for SISO systems. Extension to MIMO systems is possible by formulating convex optimization problems. The design is implemented on a nanopositioning stage for semiconductor manufacturing, to improve its transient response.