Two level calibration method for industrial robots based on LM and RBF Neural Network

In order to improve the positioning accuracy of a six-degree-of-freedom robot in practical applications, this paper proposes a two-stage error calibration method that can enhance the absolute positioning accuracy of industrial robots. This method integrates model-based calibration and non-model-based calibration techniques. In the first stage, the model-based calibration is performed using an improved Denavit-Hartenberg (MD-H) model, which establishes a comprehensive geometric parameter error model for the industrial robot. Then, the Levenberg-Marquardt (LM) algorithm is employed to identify the geometric parameter errors of the robot. In the second stage, a residual error prediction model based on Radial Basis Function (RBF) neural network is established to predict and compensate for the residual errors after correcting the geometric parameters. Finally, the Six degrees of freedom ABB IRB 120 industrial robot was tested and verified. After two-stage calibration, the average comprehensive position error of the robot end center point is reduced from 3.433 mm to 0.479 mm, and the average comprehensive attitude error is reduced from 0.483 ° to 0.087 °. Therefore, the proposed two-stage calibration method based on LM and RBF neural network effectively improves the absolute positioning accuracy of the robot.