A Sensorless Self-Calibration Method With Pose Constraint for Collaborative Robot

Abstract

Collaborative robots are designed to adapt quickly to new tasks. To enhance their absolute accuracy, self-calibration methods using portable measurement devices offer a cost-effective solution. In this work, a novel pure mechanical self-calibration device based on a three-ball-groove Kinematic Coupling is designed for collaborative robot without any sensors. Taking advantage of the hand-guiding ability of the collaborative robot, the robot’s end-effector can be precisely located to unique poses through the self-calibration device, which forms the pose constraints. A self-calibration model which reflects the relationship between the robot kinematic error and the relative pose error of two measured configurations is established utilizing the local frame representation of the Product-of-Exponential (Local POE) formula. To tackle the issues for self-calibration and base-tool calibration, a two-step calibration algorithm with the Random Sample Consensus (RANSAC) method is proposed. Simulations have been conducted for the evaluation of the effectiveness and the robustness of the proposed method. The experiment on the KUKA LBR iiwa 7 R800 shows that the mean position error and the mean orientation error are significantly reduced from 7.98 mm, 0.0069 rad to 0.41 mm, 0.0015 rad, respectively.