Robot Calibration Using Least- Squares and Polar- Decomposition Filtering

This paper reports the experimental results of a novel method to calibrate geometric errors of multi-axis robotic manipulators. The method proposed by the authors is based on a least-square estimation of the rotation matrix of a rigid body in three-dimensional Cartesian space. The error is filtered by hnposing the orthogonality constraint on the rotation matrix, using the polar-decomposition theorem. The axis of rotation of the rigid body, then, is computed from the linear invariants of the rotation matrix. Finally, the wrist of the Yaskawa Motoman Robot was calibrated. The measurements of the Cartesian coordinates of points were perfonned using a computer vision system and LED markers on a rigid body grasped by the e~d-effector. It is well known that the current accuracy of robots is outside acceptable tolerance for certain manufacturing applications such as high precision assembly. Different authors have tried to give an estimate of the error involved [10],[12],[15] the need for calibrating robotic manipulators and compensating for the errors thus becoming evident. There are two kinds of errors, namely, geometric and non-geometric. The latter are caused by backlash , structure flexibility, etc. The former exist due to imprecise manufacturing of the