Path Learning by Demonstration for Iterative Human-Robot Interaction With Uncertain Time Durations

This article presents a path learning method through physical human-robot interaction (pHRI) based on a stretch-compression iterative learning control (ILC) scheme and contouring impedance control. The robot learns a task path desired by the human user through a kinaesthetic interface and provides physical assistance to the human user in repetitive interactions. Due to the uncertainty of the human user's force and motion, the time duration of each iteration may be different, so a novel ILC scheme based on stretch and compression operation is proposed to update the reference trajectory of the robotic manipulator. By attaching the Frenet-Serret frame to each point on the reference path, the control task is decomposed into impedance control in the tangential direction and position control in the normal or binormal direction constraining the human user on the reference path. Experiments on a 7-DOF Sawyer robot are carried out to show the effectiveness and robustness of the proposed method.