Collision experiments in the human-robot interaction (HRI) context showed that the effective robot mass is one of the main parameters that influence human injury probability during a collision. Also the current standard ISO/TS 15066 highlights the importance of this parameter and provides a method to determine the maximum safe robot velocity based on the effective mass. To enable both safe and efficient robot applications, it is crucial to derive the robot's instantaneous effective mass sufficiently accurate based on either a), a kinematic and dynamic model or b), a suitable collision experiment. In this paper, we describe and quantitatively compare the well-established reflected mass model by Khatib and the simplified model provided in the ISO/TS 15066 for the KUKA LWR IV + and the Franka Emika Panda robot. Furthermore, we propose a method to practically determine the effective mass using a passive mechanical pendulum setup. Our results show that the simplified ISO/TS model can lead to a significant safety-relevant error. With our preliminary experimental setup, however, we can verify that the reflected mass obtained by the dynamics model only differs 1.1-7.8 P% from the measured value.     «

Collision experiments in the human-robot interaction (HRI) context showed that the effective robot mass is one of the main parameters that influence human injury probability during a collision. Also the current standard ISO/TS 15066 highlights the importance of this parameter and provides a method to determine the maximum safe robot velocity based on the effective mass. To enable both safe and efficient robot applications, it is crucial to derive the robot's instantaneous effective mass sufficie...     »