Dual-adaptive feedforward control for raster tracking with applications to AFMs

Abstract

We evaluate the performance of a dual-adaptive feedforward control architecture applied to the raster scan of a piezo-based positioning system. In previous work [1], we introduced the adaptive-delay algorithm that improved the tracking performance of the feedforward plant-injection architecture. The key benefit of the adaptive-delay algorithm is the adaptation calculation that does not require knowledge of plant parameters. In [1], the algorithm uses model-inverse-based feedforward control to increase raster-tracking bandwidth. It is well known that model-inverse-based feedforward control designs can perform poorly in the presence of large model variation or uncertainty. System identification methods reveal that the frequency response of our piezoscanner includes a large amount of variation as the user requests various operating points within the stage's range. As a result, tracking performance degrades as we vary from the conditions with which the model was identified. To correct for this, we combined the adaptive-delay algorithm with partial-parameter adaptation that updates critically variant parameters. This partnership of adaptive feedforward controllers improved experimental tracking results and robustness to model uncertainties.