Robot dynamic calibration on current level: modeling, identification and applications

Dynamic model calibration is an important issue and has broad applications in robotics. However, most of the previous works only focus on the robot dynamic calibration on torque level; that is, the identified parameters can predict the joint torques of robot. Unfortunately, little attention has been paid to the robot dynamic calibration on current level; that is, the identified parameters can predict the motor currents of robot. In order to address this problem, the main contribution of this article is to propose a systematic framework for robot dynamic calibration on current level, which includes modeling, identification and its applications. To the best of the authors’ knowledge, it is the first systematic work on the robot dynamic calibration on current level. Specifically, a novel dynamic identification model on current level is firstly derived. Then, an identification method based on iterations is proposed to identify the dynamic parameters on current level. Afterward, two applications based on the identification results on current level are explored. One application is to use the current-level identification results for identifying joint drive gains accurately. The other application is to use the current-level identification results to compute current residuals for robot collision detection. The advantage of the current residuals is to contain less cumulative errors. Finally, the proposed theories are validated by various experiments on the UR10 robot.