University of Zielona Góra

Iterative Learning Control (ILC) is renowned for its capability to achieve precise tracking control for systems with repetitive actions at a fixed time interval. However, pursuing the dual objective of high-precision tracking and rapid convergence is a persistent challenge in the field of learning control. To address this problem, a novel ILC method is designed for a class of discrete-time linear systems subject to non-repetitive disturbances in this paper. Particularly, the updating term in ILC is constructed inspired by the principle of sliding mode control (SMC), which results in the learning process being divided into two distinct stages: a rapid reaching stage and a slow sliding stage. As a result, a balance between convergence speed and tracking performance can be ensured via the proposed ILC method. In addition, to attenuate the effects of non-repetitive disturbances, the disturbance compensation mechanism is integrated into the proposed ILC method. Moreover, the optimal value of the learning gain can be determined using the predicted root mean square (RMS) errors of subsequent iterations, eliminating the need for additional tuning actions. Finally, simulation examples are provided to validate the effectiveness and superiority of the proposed new ILC method. Note to Practitioners—For many mechanical components in mechatronic systems and robotics, the motions are repeatable. Iterative learning control (ILC) is a well-established technique ideally suited for enhancing the performance of such repetitive tasks without excessive requirements on sensor-feedback quality or control-loop bandwidth. However, most existing ILC approaches in the literature primarily focus on improving convergence accuracy, while little attention is paid to convergence speed in the iteration domain, especially in the presence of disturbances. This paper addresses the limitations of classical ILC schemes, and draws inspiration from the sliding mode control (SMC) technique. To be specific, a novel SMC-based ILC algorithm is proposed that allows to achieve a good balance between the fast convergence and precise tracking performance, especially in case of iteration variant disturbances. Also, it will be shown how the optimal learning gains can be determined. Base on the examples of multi-axis gantry robot and injection molding process, simulations support the theoretical results, and meanwhile show the effectiveness and advantage of the proposed ILC strategy.