External Force Detection for Physical Human-Robot Interaction Using Dynamic Model Identification

Nowadays as more and more tasks require humans to collaborate with robots in modern industry, and the focus of many robotic researchers worldwide has turned towards human-robot collaboration. In human-robot interaction, ensuring the safety issues has the absolute priority for all other research work. In this context, sensorless collision detection and fast response researches in robotics contribute significantly to solve the safety issues. However, existing approaches for collision detection involve in the usage of external sensors, not fit for closed industrial robots or the offline observer based on robot’s the generalized momentum, poor in the real time response. In this study, a different method of external forces detection for sensor-less industrial robots using dynamics model identification is proposed. The main idea of our method is to identify the external torques by the comparison of the actual motor torques with the predicted joint torques based on dynamics model. Without using any extra sensors, a strict dynamics model including the parameterized friction torques has been formulated only by utilizing the measurements of the joint angles and joint torques. In addition, the essential response strategies in the post-contact stage are the main directions for our following research. Finally, the model accuracy and performance of the proposed method were evaluated in a 6-DOF manipulator. The experimental results demonstrated the reliability of our detection method basically.