An analytical method to eliminate the redundant parameters in robot calibration

Abstract

Model based error compensation of a robotic manipulator, also known as robot calibration, requires the identification of its generalized errors. These errors are found from measured data and used to predict, and compensate for, the end-point errors as a function of configuration. However, the generalized error formulation introduces redundant parameters, often non-intuitive, that may compromise the robustness of the calibration. The existing numerical methods to eliminate such errors are formulated on a case-by-case basis. In this paper, the general analytical expressions and physical interpretation of the redundant parameters are developed for any serial link manipulator, expressed through its Denavit-Hartenberg parameters. These expressions are used to eliminate the redundant parameters from the error model of any manipulator prior to the identification process, allowing for systematic robot calibration with improved accuracy. Simulations are conducted to verify the theory presented in the paper.