Task geometry aware assistance for kinesthetic teaching of redundant robots

Abstract

Kinesthetic teaching allows the direct skill transfer from the human to the robot through physical human-robot interaction. However, it is heavily affected by the robot’s dynamics and the control scheme utilized for the physical interaction. In this work, we aim at assisting the human-teacher by reducing her/his physical and cognitive load. To this aim, we propose a controller with virtual fixtures and inertia optimization for assisting kinesthetic teaching, exploiting knowledge of the task geometry and the robot redundancy. Experimental results utilizing a KUKA LWR4+ robot for the teaching of a brush painting motion on a curved surface validate the method and demonstrate its performance in comparison with a gravity compensation scheme and the utilization of virtual fixtures alone. The system is proved to be passive under the exertion of a human force.