Experimental Evaluation and Modeling of Passive Falls in Humanoid Robots

Abstract

Humanoid robots are being tested in multiple applications in different environments, ranging from household and health care facilities to industrial manufacturing or disaster scenarios. Although the first priority of a humanoid robot in any application is to keep its balance and prevent falling, this possibility can never be entirely ruled out due to an internal failure of the robot or to external perturbations. Furthermore, there is no guarantee that the robot can be actively controlled during the fall, which means that the robot will passively fall in the worst case scenario. In order to ensure the safety of humans sharing the same workspace, of nearby equipment, and of the robot itself, it is required to gain knowledge on the expected impact forces when such passive fall occurs, and to create mechanisms that mitigate the consequences of a passive fall. This paper presents an experimental study of the consequences of passive falling on the robot body, analyzes different alternatives to mitigate the impact, and presents an analytical model of the fall that helps to predict the accelerations produced at the impact. The study is conducted using a mockup based on the DLR humanoid robot TORO.