Robotic Flexible Collision Detection Based on Second-Order Sliding-Mode Momentum Observer

With the continuous development of the industrial sector, the demand for online monitoring of robot contact status has attracted widespread academic attention. Current collision detection methods such as first-order momentum observer have inevitable trade-offs in collision sensitivity and resistance to measurement noise. This paper proposes a second-order sliding mode momentum observer method to address the poor collision detection sensitivity and weak noise immunity of traditional observers. To begin, we establish a human-robot interaction force model for simulation, and subsequently derive the formula for the external force estimation using the conventional first-order generalized momentum. Based on this foundation, we design a second-order sliding mode momentum observer for robot soft collision detection, and demonstrate the stability and finite-time convergence of the entire system using Lyapunov analysis. Subsequently, validation is performed through combined Adams and Simulink simulations. The validation results indicate that the designed sliding mode momentum observer can more accurately estimate external collision torques while exhibiting better noise resilience in observations. This research achievement provides strong support for the application of robots in physical human-robot interaction (PHRI) tasks.