|A Decentralized Model Identification Scheme by Random-work RLS Process for Robot Manipulators: Experimental Studies
Sang-Deok Lee, Yeong-Geol Bae, and Seul Jung*
International Journal of Control, Automation, and Systems, vol. 17, no. 7, pp.1856-1865, 2019
Abstract : In this paper, a parameter identification method by randomly excited trajectories for decentralized joints
of robot manipulators is presented. Each joint of a robot manipulator is decoupled and identified as a second
order linear equation by a recursive least square method. Although robot manipulators are a nonlinear and coupled
system, decentralized models are required for either the independent joint control such as model-based linear control
methods, a time-delayed control (TDC) method or the input torque estimation. The random walk-based parameter
identification scheme of using a recursive least square (RLS) method is applied to a mobile manipulator, KOBOKER
as a test-bed. Then the identified models are used for designing a state observer to estimate the states of KOBOKER
more accurately when the robot follows the sinusoidal trajectory. The accuracies of the identified model and the
estimated state are verified experimentally by comparing with the torque of a linearized motion equation.
Decentralized control, model identification, random walk, robot manipulators, state observer.