Robot kinematic parameter error calibration based on Levenberg–Marquardt and artificial rabbits optimization algorithm

Positioning accuracy of the robot is generally improved by using the calibration method. The Levenberg-Marquardt algorithm is treated as the typical algorithm for robot calibration, which exists as the rounding error and then restricts further improvement of its calibration accuracy. Therefore, this paper proposes a calibration method based on Levenberg-Marquardt and artificial rabbits optimization algorithm. First, the kinematic error model of the robot is established by the modified Denavit-Hartenberg model. Second, the robot kinematic parameter errors are initially calibrated using the Levenberg-Marquardt algorithm. Third, these errors from the initial calibration are brought into the artificial rabbits optimization algorithm and accurately calibrated by the artificial rabbits optimization algorithm to obtain the optimal values and compensated into the robot kinematic error model. Finally, the simulation, calibration, and validation experiments are conducted to verify the effectiveness of the proposed calibration method and to compare the advantages of other methods. After calibration by the Levenberg-Marquardt and artificial rabbits optimization algorithm, the mean positioning error is reduced from 1.6348 mm (95% confidence interval: 1.4509-1.8187 mm) to 0.0244 mm (95% confidence interval: 0.0220-0.0268 mm) for the simulation experiments and from 0.8303 mm (95% confidence interval: 0.7341-0.9265 mm) to 0.1636 mm (95% confidence interval: 0.1485-0.1787 mm) for the validation experiments. In the application scenario experiment of measuring a standard ring gauge, the measurement error is reduced from 0.1239 mm (95% confidence interval: 0.1067-0.1411 mm) to 0.0623 mm (95% confidence interval: 0.0548-0.0698 mm). This method can more accurately calibrate the robot kinematic parameter error and significantly improve the accuracy level of the robot.