Compliant motion planning for a two-arm robot via human demonstration

Abstract

Two-arm robot system can execute complicated tasks and handle parts with complex shapes using cooperation. However, two-arm robot is not widely used because its motion planning is difficult. Human executes complex assembly tasks using two arms with experience-based learning. Therefore, assembly motions of the human are good models for the two-arm robot. This study focuses on planning compliant motion of two-arm robot system based on analysis of the assembly motion of human. A data acquisition system is constructed in order to measure the assembly motion of human. Position profiles of the part and force/moment profiles between two contacting parts are obtained through the data acquisition system. These profiles are classified into several sections representing similar features and used to plan the compliant motion. The proposed method is verified through experiments with peg-in-hole assembly using two-arm robot system.