Mechnical Design Systems Laboratory, Hiroshima University

Prediction accuracy of the kinematic model plays a pivotal role in offline numerical compensation for three-dimensional (3D) positioning error of a six-axis robotic manipulator. In addition to the Denavit-Hartenberg (D-H) error that was typically discussed in the analogous research, this paper proposes a novel scheme to identify the bidirectional angular positioning deviations of all the rotary axes without using the commercial laser tracker. This proposed scheme employs the R-Test, which is significantly more cost-efficient than a laser tracker employed in many past works. Compared to the conventional D-H model, significantly improved prediction performance of the proposed model was experimentally investigated on three rectangular paths covering the entire workspace. Experimental results indicate that the proposed scheme reduces the residual error from 505.2 to 215.2 mm, compared to 118.8 mm with a laser tracker, while maintaining a significantly lower cost that is 1/100 of a commercial laser tracker. Uncertainty analysis also clarify that when only the D-H error parameters are identified, as in many conventional works, the angular positioning deviations can be a major uncertainty contributor to the identification of the D-H errors. The proposed scheme does not require the R-Test locations to be calibrated, and its influence on the uncertainty in the identified errors is assessed.

Keywords: R-Test, Angular positioning deviations, Calibration, Uncertainty analysis, Industrial robot