Reliable home error identification of a 2-DOF parallel robot based on regularization methods

Abstract

This paper deals with the home error (the initial position error of active link) identification of a 2-DOF parallel robot. The identification model containing the home errors of the robot and the assembly errors of a new measuring instrument is developed using distance measurement. After that, an adaptive ridge regression method and a regularized Kalman filter method are proposed to improve the identification reliability and accuracy. Particularly, a modified L-curve method is proposed to provide suitable regularization parameters for the regularized Kalman filter. Based on the selected optimal measurement positions, experiments are carried out, in which the two regularized identification methods are compared with the ordinary ridge regression and the Kalman filter methods. Results show that the reliability and accuracy of the two methods are much better than the ordinary ridge regression method, and the divergence problem of the Kalman filter can be well resolved by the regularized Kalman filter.