A novel LLSDPso method for nonlinear dynamic parameter identification

Abstract

Purpose This paper aims to propose a novel method to identify the parameters of robotic manipulators using the torque exerted by the robot joint motors (measured by current sensors). Design/methodology/approach Previous studies used additional sensors like force sensor and inertia measurement unit, or additional payload mounted on the end-effector to perform parameter identification. The settings of these previous works were complicated. They could only identify part of the parameters. This paper uses the torque exerted by each joint while performing Fourier periodic excited trajectories. It divides the parameters into a linear part and a non-linear part, and uses linear least square (LLS) parameter estimation and dual-swarm-based particle swarm optimization (DPso) to compute the linear and non-linear parts, respectively. Findings The settings are simpler and can identify the dynamic parameters, the viscous friction coefficients and the Coulomb friction coefficients of two joints at the same time. A SIASUN 7-Axis Flexible Robot is used to experimentally validate the proposal. Comparison between the predicted torque values and ground-truth values of the joints confirms the effectiveness of the method. Originality/value The proposed method identifies two joints at the same time with satisfying precision and high efficiency. The identification errors of joints do not accumulate.