Physically-Consistent Full Dynamic Parameter Identification Based on Nonlinear Friction Model

Abstract

Aiming at the shortcomings of the current dynamic parameter identification methods in fitting nonlinear friction characteristics, the satisfaction of physically-consistent constraints, and the efficiency of dynamic calculation of the minimum inertial set, a physical consistent dynamic full parameter identification method based on the nonlinear friction model is proposed. Firstly, according to the characteristics of each component of the robot's dynamic equation, the joint trajectory that can cancel the non-friction moment term during the forward and reverse rotation of the joint is designed, and the friction moment is separated from the measured moment; Secondly, the nonlinear friction model is used to model the friction torque, and the friction parameters are obtained by solving the nonlinear least square problem; Finally, the regularized weighted least squares problem with physically-consistent constraints is solved by convex optimization method, and the full parameter set of robot dynamics including friction parameters and inertia parameters is obtained. The experimental results show that the RMS error of the proposed dynamic parameter identification method is the smallest compared with the CAD numerical calculation method and the penalty optimization identification method.