Abstract
Repetitive controllers based on trial and correction method have been successfully utilized to achieve desired output trajectories of various mechatronic systems. However, arbitrary use of these controllers often results in intolerable increase of the output error in transient trial stages, which restricts potential applicability of these controllers. No systematic design procedures to overcome this drawback have been proposed so far.
In this paper, a design of a class of repetitive controllers for discrete-time linear systems is considered. The conditions that some natural norms of the output error sequence decrease monotonically as the number of the trials increases are examined. A simple design method is proposed, which determines the coefficients of the error correction algorithm from finite sequece of the impulse response of the plant. Excessive increase of the output error in transient trial stages can be prevented by appropriate use of this design method. The relationship between the proposed design method and approximate inverse models of the plant is clarified. A numerical example is given to illustrate the effectiveness of the proposed method.
