Minimizing the effective robot mass in human-robot collisions by exploiting redundancy

Abstract

Enabling safe human-robots interactions is an essential goal in nowadays robotics research because modern light-weight robots allow direct physical contact. In recent years, many studies were conducted to evaluate human safety during collisions with a robot. In the experiments, the robot velocity, contact geometry and effective robot mass were found to be the key parameters that affect human injury severity. For improving collision safety, these parameters may be varied reasonably. The velocity may e.g. be lowered to a safe value by control and blunt surfaces may be chosen instead of sharp geometries. In terms of mass, one can modify the mechanical structure of the robot on the one hand. On the other hand, the robot pose has influence on the robot mass perceived by the human as well. In this thesis, the robot pose is optimized such that the effective robot mass is minimized while following a nominal trajectory. This is done by exploiting the robot’s redundant degrees of freedom. Online and offline algorithms are proposed to minimize the effective mass. Using a model of the DLR Light-Weight Robot III, the methods are tested in simulation and compared to each other.