Vision based collision detection for a safe collaborative industrial manipulator

摘要

In this work we present an improved method for collision detection that extends the ability of collaborative systems to share the workspace between the operator and the robot. Our work is based on the state-of-the-art GPU-Voxels algorithm, with a dynamic swept volume method for collision detection that minimizes intervention in robot trajectories. We propose a sensor calibration method that allows multiple cameras to be placed around the robot to ensure coverage without shadows. In addition, we show how our previous work on advanced force control algorithms, which in combination with the collision detection algorithm presented, meets all safety requirements of the technical specification ISO/TS 15066. Our ultimate goal is to derive a framework for collaborative industrial manipulators, which are able to work side by side with people.