Decoupling Identification and Physical Parameter Estimation for Serial Two-Link Two-Inertia System

Abstract

This paper proposes a multivariable identification method for a SCARA-type planar two-link robot arm with elastic joints caused by reduction gears, which is treated as a serial two-link two-inertia system. The arm for the experimental verification is equipped with accelerometers that have the two roles described below. The proposed method consists of three steps. The first step is the rigid model parameter estimation by the least-squares method. The second step is the elastic model identification using a multi-input multi-output state space model technique, which enables the nonlinear interaction torques between two links to be decoupled. The torques, calculated using the accelerometer signals and the rigid model parameters, are employed as inputs for decoupling in the multi-input identification. The angular velocities of the links, calculated using the accelerometer signals and the motor encoder signals, are employed as outputs for improving accuracy in the multi-output identification. The decoupling method divides the serial two-link two-inertia system into two one-link two-inertia systems in this step. The third step is the physical parameter estimation of the one-link two-inertia systems. The physical parameters consist of motor inertias, link inertias, joint-friction coefficients and joint-spring coefficients. Furthermore fine tuning of the estimated physical parameters is carried out using closed-loop simulations with the nonlinear least-squares optimization. Experimental results using the two-link arm have shown the accuracy of the proposed identification method.