Accurate kinematic calibration of a six-DoF serial robot by using hybrid models with reduced dimension and minimized linearization errors

Authors:
Publisher website
Google Scholar
Abstract
Purpose: In typical model-based calibration, linearization errors are derived inevitably, and non-negligible negative impact will be induced on the identification results if the rotational kinematic errors are not small enough or the lengths of links are too long, which is common in the industrial cases. Thus, an accurate two-step kinematic calibration method minimizing the linearization errors is presented for a six-DoF serial robot to improve the calibration accuracy. Design/methodology/approach: The negative impact of linearization on identification accuracy is minimized by removing the responsible linearized kinematic errors from the complete kinematic error model. Accordingly, the identification results of the dimension-reduced new model are accurate but not complete, so the complete kinematic error model, which achieves high identification accuracy of the rest of the error parameters, is combined with this new model to create a two-step calibration procedure capable of highly accurate identification of all the kinematic errors. Findings: The proportions of linearization errors in measured pose errors are quantified and found to be non-negligible with the increase of rotational kinematic errors. Thus, negative impacts of linearization errors are analyzed quantitatively in different cases, providing the basis for allowed kinematic errors in the new model. Much more accurate results were obtained by using the new two-step calibration method, according to a comparison with the typical methods. Originality/value: This new method achieves high accuracy with no compromise on completeness, is easy to operate and is consistent with the typical method because the second step with the new model is conveniently combined without changing the sensors or measurement instrument setup.

This publication has 25 references indexed in Scilit: