Programmable Dynamic Modeling and Parameter Identification for Serial Manipulators

Manipulators have a wide range of applications in industry, and their dynamic parameter identification can improve model accuracy and contribute to the digital twin. Generally, parameter identification requires prior surrogate modeling. Due to the coupling between the generalized coordinates, the complexity of the model obtained via the overall modeling approach poses a challenge for parameter identification. To solve this problem, recursive modeling should be employed, offering the additional advantage of simplified programming. Moreover, this paper proposes a programmable manipulator parameter identification process. First, the linearized dynamic model is derived from the Newton-Euler method’s force balance relationship between connecting links, utilizing both geometric parameters and experimental data. Then, the minimum parameter set is calculated based on QR decomposition, and the dynamic equations are changed into an easily identifiable form. Finally, the least squares method is used for parameter identification. To assess the accuracy of our process, we calculated the root-mean-square error (RMSE) between reconstructed torque and theoretical torque in simulation experiments of the UR10 robot. This paper’s key contributions are a modular process and programmable methodology, offering the foundation for developing manipulator parameter identification software that streamlines the identification process.