A MR-Assisted and Scene Perception System for Human-Robot Collaborative Disassembly of Power Batteries

Abstract

To make up for the deficiency of manual disassembly which is low efficiency, high cost and poor safety, robots have great advantages in performing safe and effective disassembly for end-of-life power batteries. Human-robot collaboration (HRC) is mostly pre-programmed, lacks adaptability as well as flexibility to dynamic scenarios. Therefore, it is challenging to ensure the efficiency and safety of human-robot collaborative disassembly(HRCD) for retired power batteries. Targeting that, this research proposes a mixed reality (MR) assisted scene perception system for human-robot collaborative disassembly. The scene semantic recognition module integrates instance segmentation, gesture recognition and action recognition for the perception of the elements in the disassembly scene, including the position of the operation, the robot, the power battery and the tools. The MR technology is utilized to simulate and visualize the HRCD processes, and ensure the safety distance between the human and the robot. The results of the experimental case study indicate that the proposed approach increase the efficiency and enhance the safety of the HRCD system.