Bettering adaptive iterative learning for rotation maneuvers of rigid bodies via segment-wise rectification

Abstract

This paper proposes a segment-wise rectification method to enhance the learning efficacy of iterative learning control (ILC), which is suitable for repetitive tracking tasks of rigid bodies. Despite the conventional paradigm of adaptive ILC, two challenges persist in adaptive ILC for rotation maneuvers: the unboundedness problem, characterized by the unbounded nature of estimates and input amplitudes, and the temporal coupling problem, involving the unnecessary impact of the past tracking performance on the current tracking performance along the time axis. Either of the two issues may cause undesirable overestimation. To address the unboundedness problem, we propose a learning mechanism based on segment division, facilitating the continuous evolution of estimates within each segment throughout iterations. This mechanism naturally ensures the boundedness of all system signals, mitigating overestimation risks. Moreover, to resolve the temporal coupling problem, we develop the segment-wise rectification method regardless of the unity constraint of states. By leveraging these methodologies, we establish an innovative adaptive ILC framework that achieves quasi-perfect (respectively, perfect) attitude tracking when initial state errors are non-zero (respectively, zero). Furthermore, the proposed adaptive ILC is applied to the repetitive attitude tracking tasks of rigid bodies, demonstrating exceptional tracking and learning performance in the presence of uncertainties.