A novel method of automatic hand-eye calibration for robotic manipulator

Abstract

Aiming at the problem of manual operations and low accuracy in current hand-eye calibration process for a robotic manipulator, this paper proposes a novel method of Automatic Pose Generation and Calibration (APGC). First, a series of initial robot poses are automatically generated in Cartesian space. Then, RANSAC and K-means clustering algorithms are introduced to perform a two-step pre-screening process to select the optimal initial poses, so as to acquire robot poses that contribute to improving the calibration accuracy. Finally, the existing dual quaternion and convex relaxation global optimization theory are employed to solve the calibration matrix equation. The simulation results show that position and orientation errors of the APGC are more stable than those of two state of the arts to camera focal length noise. The experimental results show that the average position error is 0.516mm, the average orientation error is 0.048°. Compared with two state of the arts, the position and orientation errors are smaller, and the calibration accuracy is higher.