System Identification Method for Robotic Manipulator based on Dynamic Momentum Regressor

Abstract

The paper presents a momentum regressor using both numerical integration and Hamiltonian dynamic description for robotic manipulator. It has an advantage in that the proposed momentum regressor does not require the acceleration estimation or sensor for the identification of dynamic parameters. In addition, the identification algorithm is proposed by using Lagrange multiplier and bordered gramian matrix for the minimization of difference between the previous and the current estimates of dynamic parameters. The developed algorithm is intuitive and easy to implement in real-time fashion. Finally, the effectiveness of the proposed momentum regressor and identification method is shown through the numerical simulations about a planar two-link robotic manipulator system.