AAS 95-377

Use of Non-Causal Digital Signal Processing in Learning and Repetitive Control

Y. Wang and R. W. Longman, Columbia University, New York, NY

Abstract

This paper develops new easily implemented learning and repetitive control algorithms based on concepts from digital signal processing. Most learning control approaches have been based on feedback control concepts that require causality. However, learning control and to some extent repetitive control have the luxury that control action updates based on previous repetitions can be based on data future to the time step of interest in that repetition, i.e. we are not constrained by causality. This opens up the large body of digital signal processing methods. These methods are applied here to develop learning control laws that are adjusted for good transient behavior based on frequency response concepts.