Iterative Learning Controller Design Using the Inverse Model of a Nominal Feedback Control System

Integrating iterative learning control (ILC) with feedback control systems progressively enhances tracking performance by learning from data, such as control inputs and tracked errors accumulated over multiple trials. Despite the integration of ILC systems with existing feedback control systems, learning controllers have often been designed without effectively leveraging the information used in feedback controller design. Furthermore, the improper utilization of this information can degrade the performance of ILC systems. In this study, the ILC system comprises two learning filters: a learning filter and a robustness filter. The learning filter is directly derived from the inverse of the nominal feedback control system, while the robustness filter is a low-pass filter that ensures robust convergence under uncertainty. To design learning controllers, uncertainty is isolated using linear fractional transformation (LFT) within a transfer function based on established convergence conditions. A robust convergence condition in the -norm sense is formulated, and it is represented by the robustness filter, uncertainty weighting function, feedback controller, and nominal plant. Based on the derived convergence condition, criteria for the straightforward design of learning controllers were presented. The performance weighting function employed in the design of the feedback control system was excluded from the design of the ILC system, thereby ensuring unobstructed enhancement of the learning performance. Finally, simulation studies were conducted to demonstrate the feasibility of the proposed method.