Machine Perception Platform for Safe Human-Robot Collaboration

Abstract

Speed and separation monitoring, operation defined in safety standards for collaborative robots, is meant for real-time collision avoidance. Laser scanners are safety-certified devices and a traditional sensor choice for this application. Unfortunately, the limited amount of target information they provide restricts their use in realistic collaborative robot scenarios, in which knowledge about the nature of the detected targets is required. We propose a machine perception platform for safe human-robot collaboration based on a broadband W-band radar, a 3D camera, and a laser scanner. Besides computing range and angle-of-arrival information, we use the micro-Doppler signatures of the radar echo signals to distinguish between humans and objects.