Robotic High-precision Collision Detection and Force Estimation Under Unknown Load

Collision detection is a fundamental problem in the field of human-machine interaction. Inaccurate robotic models caused by unknown loads significantly impact the sensitivity of collision detection and can even result in algorithm failures. This study proposes a collision detection method, inspired by the concept of “interference cancellation” in communication, to overcome the limitations of existing methods in eliminating the influence of unknown loads. The method is based on the self-interference cancellation extended state observer (SICESO). The method establishes a relationship between the output of a feedback control system and the state of robotic motion through dynamics. Computational efficiency is enhanced by employing a reduced-order extended state observer (ESO) to construct an external force observer. The internal disturbance influence caused by dynamic model errors is removed by appropriately setting the observation positions of ESOs based on the delay response characteristics between the physical system and the control commands. The proposed method achieves accurate estimation of external forces independent of the load. Experimental results demonstrate that the proposed method effectively eliminates the influence of unknown loads, achieves collision detection within 0.001s, and accurately estimates external forces with an accuracy of 10⁻² Nm. The proposed method has significantly improved sensitivity and robustness.