@article{LI2026, 
author = {Ming LI and Rongsheng LU},
title = {Three Step Identification Method for Kinematic Parameters of Industrial Robots},
year = {2026},
journal = {Journal of South China University of Technology (Natural Science Edition)},
volume = {54},
number = {7},
pages = {66-78},
keywords = {kinematics, axis method, attitude error model, position error model, binocular stereo vision},
url = {https://www.sciopen.com/article/10.12141/j.issn.1000-565X.250343},
doi = {10.12141/j.issn.1000-565X.250343},
abstract = {Currently, the absolute positioning accuracy of industrial robots is considerably lower than their repeatability precision. To improve the pose accuracy of the robot’s Tool Center Point (TCP), this paper investigates the theory of robot kinematic parameter errors and binocular stereo vision pose measurement techniques, and proposes a three-step identification method for kinematic parameters. This three-step method first employs a single-axis rotation approach to fit the axes of Joints 1, 2, and 3, from which the first part of the kinematic parameters is calculated. Then, a pose error model of the target sphere is constructed to identify the angular components of the remaining kinematic parameters. Finally, a position error model of the target sphere center was constructed to identify the length components of the remaining kinematic parameters. A corresponding experimental system was built, and binocular stereo vision technology was used to measure and study the changes in target sphere pose during the robot’s single-axis rotational motion. The results show that the proposed three-step method effectively overcomes the issue of low measurement accuracy for the axes of the robot’s last three joints, achieving separate identification of the angular and length parameters in the kinematic model. Experimental results indicate that after calibration using the three-step method, the robot’s positioning error is reduced to 66.01% of its original value, and the uncertainty is reduced to 85.62% of its original value, demonstrating that the method can effectively reduce industrial robot positioning errors. Comparative analysis of compensation outcomes across one-step, two-step axis, two-step, and three-step methods validate the superiority of the proposed three-step method in kinematic parameter identification, achieving better identification performance.}
}