On the Interaction Force Sensing Accuracy Of Franka Emika Panda Robot

The precise and reliable interaction force knowledge is among the main aspects of human-robot interaction applications. Since many robots are equipped with joint torque sensors, relying on these measurements avoids increasing cost and complexity, brought by additional sensors. In this paper, we investigate the sensing capabilities of the Franka Emika Panda Robot for interaction force estimation, focusing on the available filtered force estimates provided by the robot controller. We perform experiments involving contact interaction between the robot and an external environment, implemented by using an impedance controlled servomotor. In this way, it is possible to reproduce forces typical of the human behaviour, in controlled and repeatable conditions. The evaluation is based on the identification of the motor impedance parameters, comparing the results obtained by the Panda interaction force estimate with a ground truth value, obtained by load cell measurements. Since joint torque measurements are heavily affected by disturbance torques (e.g. friction) a simple compensation procedure is performed, through data collected without contact. The delay introduced on the external force measurements by the filtering action is compensated, allowing an online usage of the proposed method. Results show that the Panda Robot provides a fairly good estimate of the interaction force.