A Human-Robot Collaboration System for Object Handover

Yifei Yang,Longzhong Lin,Yifan Zhang,Zhongxiang Zhou,Yue Wang,Rong Xiong
DOI: https://doi.org/10.1109/robio54168.2021.9739333
2021-01-01
Abstract:Humans and robots are the two main labor in industry, and they have respective advantages. Human-Robot Collaboration combines the advantages of humans and robots, which makes it a research hotspot. While there are a lot of research and developments in the field of Human-Robot Collaboration, how to apply and optimize existing technologies to realize a complete Human-Robot Collaboration system is not a very focused problem. So, we realize a Human-Robot Collaboration system for object handover: the human holds a container (denoted as ContainerA), and the robot holds the other container (denoted as ContainerB) with some contents as a collaborator; the robot uses computer vision to detect ContainerA and estimate position and orientation of it; then the robot plans and controls the motion of the robot arm based on the visual information, and hand over the contents in ContainerB to ContainerA. For computer vision, we use Fast- SCNN for semantic segmentation and REDE for pose estimation, which can obtain the accurate pose of the hand-held container in real time. In terms of robot arm motion planning and control, we perform speed control based on artificial potential field and admittance control to achieve a smooth follow-up movement.
What problem does this paper attempt to address?