Human-Robot Interactive Skill Learning and Correction for Polishing Based on Dynamic Time Warping Iterative Learning Control

Abstract

To achieve rapid and flexible deployment of robots in the finishing process of small batch workpieces, this article proposes a framework for human-robot interactive (HRI) skill learning and correction based on improved dynamic time warping iterative learning control (DTW-ILC). First, we incorporate Gaussian mixture model (GMM) with DTW-ILC approach to enable the robot to learn polishing skills from human demonstration and interaction. Second, to ensure accurate force tracking under the condition of varying polishing feed speed, we propose an iterative force tracking method based on DTW-ILC and impedance control. Notably, we propose to iteratively estimate the polishing stiffness and incorporate it into the path updating law, resulting in simplified parameter settings and faster error convergence compared with traditional iterative learning control (ILC) methods with fixed parameters. A polishing experiment is carried out to prove the effectiveness of the proposed framework and method.