Disturbance Recognition and Collision Detection of Manipulator Based on Momentum Observer

Abstract

Increasing requirements for the safety of human-robot interaction and the cost-effectiveness of collision detection rapidly promote the development of collision detection technology without torque sensors. To address nonlinear disturbance factors in collision detection that may cause unstable or even incorrect detection, this paper proposed a research strategy that considered the friction as the disturbance term in manipulator motion for the collision detection. The manipulator joint disturbance model was established based on the LuGre dynamic friction model, and the external torque observer was designed based on the generalized momentum. Then, the friction measurement was realized using the external torque observer, and the model parameters were identified through the genetic algorithm. The collision detection can be reduced errors after the friction model by compensating the disturbance and can be applicable to variable working conditions. Finally, the accuracy of the constructed disturbance model and the performance of the proposed collision detection method were validated by the experimental studies.