Human-robot collision detection under modeling uncertainty using frequency boundary of manipulator dynamics

Abstract

This paper presents the development and experimental evaluation of a collision detection method for robotic manipulators sharing a workspace with humans. Fast and robust collision detection is important for guaranteeing safety and preventing false alarms. The main cause of a false alarm is modeling error. We use the characteristic of the maximum frequency boundary of the manipulator’s dynamic model. The tendency of the frequency boundary’s location in the frequency domain is applied to the collision detection algorithm using a band pass filter (band designed disturbance observer, BdDOB) with changing frequency windows. Thanks to the band pass filter, which considers the frequency boundary of the dynamic model, our collision detection algorithm can extract the collision caused by the disturbance from the mixed estimation signal. As a result, the collision was successfully detected under the usage conditions of faulty sensors and uncertain model data. The experimental result of a collision between a 7-DOF serial manipulator and a human body is reported.