Iterative Learning Control: Practical Implementation and Automation

Iterative learning control (ILC) has been well recognized for its output tracking ability in systems that perform repetitive tasks, such as robot manipulators. In practice, however, the application of ILC remains challenging as it generally requires the repetition of the initial settings and such industrial manipulators do not provide measurements of actual joint positions. This article presents a practical and fully automated implementation of ILC for industrial robot manipulators while assuming complete model uncertainty and the unavailability of joint velocity measurements. We propose the employment of a dedicated controller tasked only with bringing the arm back to its initial position while reducing the joint position reinitialization error after every iteration. We show that the convergence of the position tracking error is bounded by the initial mismatch. Finally, we provide experimental justification of the proposed approach on a 4-DOF commercial robot manipulator while only using the measurements of relative joint positions and show that our fully automated process outperforms two ILC algorithms with thorough
manual
homing.