Collision Detection of Robotic Manipulators Based on Momentum Residual Analysis

Collision detection is a critical capability for robotic manipulators operating under high-speed and high-precision conditions, as unexpected contacts may lead to performance degradation or mechanical damage. This paper investigates a collision detection method for robotic manipulators based on momentum residual analysis. First, based on the joint-space dynamic model, first-order and second-order momentum residuals are constructed, and their dynamic characteristics are analyzed. On this basis, magnitude thresholding is adopted as the detection criterion for the first-order momentum residual, and a time-interval identification criterion is designed by exploiting the directional zero-crossing characteristics of the secondorder momentum residual at collision onset and release. Finally, simulations are conducted on a robotic manipulator under both rigid and compliant collision scenarios. Simulation results demonstrate that the momentum residual method can effectively detect collision faults, and the detection approach based on the second-order residual achieves higher accuracy compared with the conventional first-order method.