Improved Data-Weighted Iterative Parameter Identification Method for Accurate Dynamic Modeling of Collaborative Manipulators

Collaborative manipulators have become increasingly necessary in industry and human-robot collaboration, and an accurate dynamic model of manipulators serves as an important foundation for applications such as precise position/force control and collision detection. However, data bias, resulting from data sampling errors, and complex non-linear friction significantly reduces the accuracy of identification methods. To address these issues, we propose an improved data-weighted iterative identification method for manipulator dynamic models. The improved data weight function is employed in the inner loop iteration to prevent information loss caused by a fixed weight threshold. Furthermore, a novel friction model considering friction anisotropy and the Stribeck effect is introduced to iterate in the outer loop. Finally, experiments are conducted on three different collaborative manipulator datasets, and we compare the identification accuracy of our proposed method with other state-of-the-art algorithms. The experimental results demonstrate that the accuracy of the proposed method is improved by more than 15% compared to others. The proposed method exhibits excellent torque estimation accuracy and good applicability to different manipulators.