Sensorless collision detection and identification utilizing motor current and model-based friction on a 7-DoF manipulator

Abstract

Safe physical human-robot interaction is an essential aspect of modern collaborative manipulators. Reliable collision detection and identification are thereby of crucial importance. Joint torque sensors, which are available at high cost, are commonly used for this purpose. In this thesis, a sensorless approach is presented that estimates external joint torques with a momentum observer utilizing motor current and model-based friction. It shows reliable results for joints in motion. Furthermore, the kinematic and dynamic models are extended by a force/torque (F/T) sensor under the base of the robot. An augmented momentum observer is implemented that provides singularity-free estimates of external Cartesian wrenches at a previously known contact point. Moreover, a hybrid form of base F/T sensor and sensorless approach is presented which, by taking into account dynamic effects, provides fast collision detection at previously known contact points. These approaches are tested and evaluated in experiments on a 7-DoF manipulator. Main ref: Maged Iskandar, Oliver Eiberger, Alin Albu-Schäffer, Alessandro De Luca, and Alexander Dietrich, “Collision Detection, Identification, and Localization on the DLR SARA Robot with Sensing Redundancy”. In: 2021 IEEE International Conference on Robotics and Automation (ICRA). 2021, pp. 3111–3117.