Abstract:
The generalized momentum observer (GMO) is a widely used collision detection technique for collaborative robots (cobots). However, a consensus amongst the robotics commun...Show MoreMetadata
Abstract:
The generalized momentum observer (GMO) is a widely used collision detection technique for collaborative robots (cobots). However, a consensus amongst the robotics community is that the GMO has limited bandwidth and is, therefore, insensitive to hard collisions (rapid occurrence). This article proposes a Back-Input compensation (BICom) approach that can effectively detect both soft (slow) and hard collisions. The basic idea is to utilize the velocity-related residual distribution of model uncertainties and compensate for the deviation between computed and measured torques of robot dynamics, whereas deviations caused by collisions might exceed the upper limits of compensation and then be detected. Besides, the window detector module mitigates false positives of collision detection, and overcomes two common defects of the anti-false alarm scheme, namely the detection delay and the false negatives. Compared with the high-pass filter-based detection technique, BICom can not only respond instantaneously to a hard collision but also monitor the quasi-static contact, which improves detection sensitivity. Experiments on a 6-DoF cobot validate the effectiveness of our method.
Published in: IEEE/ASME Transactions on Mechatronics ( Volume: 27, Issue: 6, December 2022)