Enhanced Robot Calibration by Minimization of TCP Drifts during Reorientation

Abstract

Kinematic calibration is necessary for every newly assembled industrial robot to achieve specified TCP (Tool Center Point) performances such as accurate linear and reorientation movements in Cartesian space. The compulsory joint level calibration is used to determine the joint offsets in a direct way manually or automatically, with the assumption that geometric characteristics are ideal, while the optional arm level calibration seeks to identify the practical geometric errors of robot individuals by certain optimization algorithms. In terms of TCP performance, the latter tends to present high accuracy thanks to corrected kinematic model by a time-consuming and costly process. For the former, TCP behavior will be less accurate since not only geometric errors are left uncompensated but also a potential poor calibration in joint space will deteriorate the situation of the distal behavior in Cartesian space. To mitigate this risk in a closed-loop manner, this paper firstly reviews the major popular calibration methods in market. Then, combining the two levels of calibrations together, a new method is proposed to perform an enhanced joint level calibration by the similar methodology as arm level calibration, which will identify the joint offset errors by minimization of TCP drifts during reorientation movement. Experimental results reveal that this method, as a joint level calibration solution per se, proves to be more effective and efficient by improving the accuracy of relative Cartesian movements to a level comparable to arm level calibration but with less time and efforts, which is of great practical significance for the improvement of quality control and production efficiency of robot producers.