A parameter separation-based method for kinematic identification of industrial robots without prior kinematic information

Abstract

Accurate kinematic parameters are crucial for deploying industrial robots in high-precision manufacturing applications, such as machining workpieces. Traditional kinematic identification methods often assume that nominal parameter values are known and used as initial estimates. However, in industrial applications, obtaining such nominal values is challenging due to the limited access to detailed design data and the specialized knowledge required for kinematic modeling. This lack of prior kinematic information poses significant challenges to the accuracy and efficiency of parameter identification. To address these issues, we introduce a variable projection (VP) method that eliminates linear parameters through orthogonal projection, transforming the kinematic parameter identification problem into a nonlinear least squares problem involving only the nonlinear parameters. First, the separable structure of the kinematic model is explicitly derived. Then, a novel approach is proposed to integrate the separation of redundant parameters with the VP method. By focusing solely on non-redundant nonlinear parameters, the proposed method significantly reduces reliance on the accuracy of the initial estimates. Simulations and experiments demonstrate that the proposed method achieves more stable parameter estimates and faster convergence in the absence of prior kinematic information. Furthermore, the identified parameters are successfully applied for error compensation in a robotic machining case, leading to an 80% improvement in machining accuracy.