Dynamic Model Identification of Collaborative Robots Using a Three-Loop Iterative Method

Abstract

Accurate dynamic model is of crucial importance for collaborative robot to achieve a satisfying performance in model-based control or other applications. Nevertheless, it is often difficult to identify dynamic parameters accurately due to the joint elasticity and nonlinear fricitonal impact. Considering viscous friction nonlinearity and a minor Stribeck effect commonly seen in the practical situations, we present a three-loop iterative method for the dynamic model identification of collaborative robots. The inner loop is well-designed to identify the mass-inertial parameters and linear friction parameters of the robot with a least-squares approach. As for the middle loop, a suitable nonlinear viscous friction model is applied to improve the identification results. The impact of the Stribeck effect is thoroughly studied in the outer loop. Experiments on a six-dof collaborative robot have proved the identification errors can be significantly reduced by our method.