Adaptive Virtual Guides for Compliance Control Skill Teaching in Partially Known Tasks

Abstract

Compliant manipulation is widely studied in the domain of robotics due to urgent industrial demands of flexible production. In many tasks that require human demonstration of the end-effector’s pose and contact force, cognitive and physical loads on human operators are high. Fortunately, virtual guides (VGs) can effectively assist human users to carry out the corresponding work. This article aims to design an adaptive virtual guides (AVGs) scheme for kinesthetic teaching that can be applied to surface finishing and to reduce human load. For this purpose, the end-effector’s orientation is automatically adjusted according to the workpiece’s 3D model and the end-effector’s position. The force provided by the robot along the end-effector’s z-axis is adjustable, which is used to reduce human’s control effort for contact force demonstration. Our method utilizes the virtual mass-damper system and joint admittance control model for the AVGs design, whose validity is verified on a Sawyer robot in simulation.