Analysis of Different Human Body Recognition Methods and Latency Determination for a Vision-based Human-robot Safety Framework According to ISO/TS 15066

David Bricher, Andreas Müller

Abstract

Today, an efficient and flexible usage of lightweight robots in collaborative working spaces is strongly limited by the biomechanical safety regulations of ISO/TS 15066. In order to maximize the robot performance without contradicting the technical standards and recommendations, a safety framework is introduced, which makes use of state-of-the-art deep learning algorithms for human recognition and human body part identification. Particularly, a generic vision-based method for the determination of the occurring latencies is proposed. To this end, the different latency contributions from the recognition process up to the process of adapting the robot speed to an ISO-conform level are analyzed in detail.

Download


Paper Citation