3차원 스캐너를 활용한 로봇 기구학 캘리브레이션

Abstract

This paper proposes a novel feature point estimation algorithm utilizing 3D scanner for robot kinematic calibration. The proposed algorithm identifies feature points from 3D point cloud data obtained by scanning a polyhedral measurement jig attached to the robot’s end effector. By systematically estimating the intersecting points of three planar surfaces on the jig, precise and consistent feature points are detected. The identified points are then employed for kinematic calibration, demonstrating the effectiveness of the algorithm. The calibration experiments were conducted using a 6-DOF robot manipulator, Rainbow Robotics RB5-850. Calibration accuracy was verified using distance error modeling without requiring transformation matrices between the robot and scanner coordinate systems. Experimental results show that the algorithm significantly improves positioning accuracy, reducing distance error by up to 93.16% in training data and 67.89% in test data. The proposed method offers a practical alternative to expensive laser trackers, enhancing the usability of 3D vision systems in industrial environments.