Online estimation of impedance parameters for a variable impedance controlled robotic manipulator

Abstract

The aim of this work is to estimate the impedance parameters, namely the damping and stiffness, of a variable impedance dynamic system. The estimation is performed using a Constrained Extended Kalman Filter (CEKF). Comparing the various non-linear estimation techniques, Extended Kalman Filter shows a superiority with respect to speed of execution. This is a major requirement in case the estimation is used for a task involving online-tuning of the parameters of a variable impedance controlled robotic manipulator in contact with a variable impedance dynamic environment, for example, during human-robot physical interaction. In order to have a ground truth, the algorithm was experimentally tested on a system with known variable impedance, namely, a variable impedance controlled KUKA LWR. For the estimation procedure, the position of the end-effector was used as the measurement and the external force applied on it as a known input. Without giving explicit information on the dynamics of the variable impedance parameters of the controlled manipulator, the CEKF appreciably tracked the real parameters. The performance of the estimator declines in case the impedance variation is highly non-linear.