Dynamic Parameters Identification of an Industrial Robot: A Constrained Nonlinear WLS Approach

Abstract

This paper brings an identified model for a 6 degrees of freedom (dof) industrial robot, the Denso VP-6242G robot, with an end-effector composed of a spherical handle, fixed on a force sensor. This robot is intended to experiments in the field of Physical Human-Robot Interactions (PHRIs), for co-manipulation purposes. The control algorithms that are necessary to achieve a good PHRI, require a good knowledge of the robot dynamical model, especially the inertia matrix which should be positive definite whatever the configuration of the robot. However, most of industrial robots are supplied without any datasheet containing the inertial parameters nor Computer-Aided-Design (CAD) model. Hence, we propose to apply an identification procedure to experimental data, based on the Inverse Dynamic model Identification Method (IDIM). To ensure the positive definiteness of the inertia matrix, the used optimization step addresses the problem of nonlinear Weighted Least Squares (WLS), derived from the mathematical formulation of the identification problem, under a set of nonlinear constraints in the parameters. A validation step permits to check the efficiency of this approach for the Denso robot.

A nonlinear robot joint friction compensation method including stick and sliding characteristics

“…Due to its simplicity, Coulomb model with linear viscous friction is widely used. Hao et al (2016), Bahloul et al (2018) and Guo et al (2018a, 2018b) integrated Coulomb model with linear viscous friction into robot dynamic model and performed unified linearization and parameter identification. Zhang et al (2020) separated Coulomb model with linear viscous friction from link dynamic model and reckoned friction parameters independently.…”

Joint Dynamics and Adaptive Feedforward Control of Lightweight Industrial Robots

“…Dynamics identification have been conducted for several industrial robots including the 7 DOF DLR light-weight robot [18], COMAU Smart-3 S2 [53], KUKA KR 15 [54], Mitsubishi PA-10 [55], Staubli RX-60 [56], Staübli TX40 [57], ER-16 [58], SCHUNK Powerball LWA 4P [59], KUKA KR 6-2 [60], KUKA LBR iiwa [61], and Denso VP-6242G [62]. However, the works on robot dynamics identification assume simple models for the joint dynamics which introduce bias to the inertial parameter estimates.…”