Improved Safety and 3D Scanning with Human-Robot Collaboration

Madison Wright, Karlym Nam, Jinguang Tong, Sundaram Muthu, Lars Andersson, Chuong Nguyen

Abstract: In many technical working environments, humans need robots to perform, or assist with, repetitive tasks. Since robots remain a tool from our perspective, the question is how to safely use this tool to its full potential. We focus on two use cases: one is safety and the other is knowledge sharing. First, collaborative robots, or cobots, have been used more and more in factory work cells due to their ability to stop and prevent serious injuries when they collide to human operators. It would be even better if the collision is avoided in the first place with robotic vision. Second, robots could carry out tasks but may lack the knowledge or skill to deal with difficult situations, for example: completing a full 3D scanning of an object. In this paper, we present two approaches to improve safety and 3D scanning in a research lab environment. The setup is a work cell where a human operator works with a robot arm which is equipped with an RGBD camera and a gripper to manipulate and scan objects of interest. A second RGBD camera looks at the work cell for human gestures and their body movement. Robotic Operating System (ROS) version 2 is employed for hardware-software integration. One of our approaches allows the robot to track and follow objects of interest as they are moved around on a table, and stop when it detects a human hand in the way. Our other approach allows the robot to recognise human gestures to capture RGBD data from multiple views for 3D scanning and flip the object for scanning the hidden side. The captured data is then combined to have a more complete 3D information of the object. We demonstrate the effectiveness of our robot performing these tasks and the improvement in work safety and the quality of 3D object reconstruction.