Multi-Sensor Obstacle Tracking for Safe Human-Robot Interaction

Abstract:
For human-robot interaction in industrial applications, monitoring the robot surroundings is essential in order to guarantee the safety of humans sharing the workspace with robots. This contribution presents an approach for obstacle detection and tracking based on the fusion of multiple heterogeneous depth sensors which are mounted on board a mobile manipulator. The proposed methods can track arbitrary obstacles all around the robot. Furthermore, a novel method for extrinsic calibration of the sensors is proposed, using the manipulator as a sensor calibration target. This approach enables a robust and accurate calibration without the need for a dedicated calibration object. Experimental results validate the real-time performance of the proposed obstacle tracking method.