Modeling and calibration of high-order joint-dependent kinematic errors for industrial robots

Abstract

Abstract Robot positioning accuracy is critically important in many manufacturing applications. While geometric errors such as imprecise link length and assembly misalignment dominate positioning errors in industrial robots, significant errors also arise from non-uniformities in bearing systems and strain wave gearings. These errors are characteristically more complicated than the fixed geometric errors in link lengths and assembly. Typical robot calibration methods only consider constant kinematic errors, thus, neglecting complex kinematic errors and limiting the accuracy to which robots can be calibrated. In contrast to typical calibration methods, this paper considers models containing both constant and joint-dependent kinematic errors. Constituent robot kinematic error sources are identified and kinematic error models are classified for each error source. The constituent models are generalized into a single robot kinematic error model with both constant and high-order joint-dependent error terms. Maximum likelihood estimation is utilized to identify error model parameters using measurements obtained over the measurable joint space by a laser tracker. Experiments comparing the proposed and traditional calibration methods implemented on a FANUC LR Mate 200i robot are presented and analyzed. While the traditional constant kinematic error model describes 79.4% of the measured error, the proposed modeling framework, constructed from measurements of 250 poses, describes 97.0% of the measured error. The results demonstrate that nearly 20% of the kinematic error in this study can be attributed to complex, joint-dependent error sources.