A comparison of ILC architectures for nanopositioners with applications to AFM raster tracking

Abstract

In previous work, we compared the raster tracking performance of two distinct combined feedforward/feedback control architectures while using model-inverse-based feedforward control [1], [2]. In this paper, we extend that work into the application of parallel and serial iterative learning control (ILC) architectures. These ILC architectures naturally relate to the two previously studied combined feedforward/feedback control architectures, feedforward closed-loop injection (FFCLI) and feedforward plant injection (FFPI). Experimental learning results from an atomic force microscope (AFM) raster scanner are provided as well as results comparing the FFPI and FFCLI architectures with those of the learned performance for parallel and series ILC. We show that the value of ILC over model-inverse-based feedforward methods is increased in the presence of model uncertainty or variation.