Adaptive Collision Detection via Impulse-Momentum Theorem for Safe Sensorless Physical Human-Robot Interaction
Hongzhe Shi, Chao Ye, Chenlu Liu, Weiyang Lin
Abstract
Collision detection is critical for safe physical human-robot interaction (pHRI). The generalized momentum observer (GMO) is a prevalent sensorless method, estimating momentum deviations to detect collisions. However, its reliance on iterative solutions and static thresholds severely limits operational efficiency and dynamic adaptability, leading to reduced sensitivity and accuracy in complex collision scenarios. To overcome these limitations, this paper proposes an impulse-based dynamic threshold generalized momentum observer (IDT-GMO) for adaptive sensorless collision detection. Key innovations include: (i) An impulse-based dynamic threshold mechanism, leveraging the impulse-momentum theorem, enabling adaptive collision detection with enhanced sensitivity and accuracy; (ii) A closed-form analytical solution, formulated through Lagrangian mechanics and Lie group theory, eliminating iterative computation and achieving 45% faster processing than conventional GMO. Experimental validation across soft contact, rigid impact, and multi-contact scenarios confirms that IDT-GMO achieves superior detection sensitivity and accuracy compared to existing methods. Thus, IDT-GMO exhibits significant potential for applications ranging from delicate contact tasks to collision-prone environments.
BibTeX
@inproceedings{ral2026_adaptivecollisio,
title = {Adaptive Collision Detection via Impulse-Momentum Theorem for Safe Sensorless Physical Human-Robot Interaction},
author = {Hongzhe Shi and Chao Ye and Chenlu Liu and Weiyang Lin},
booktitle = {RA-L 2026},
year = {2026}
}