Multi-directional Position Accuracy Variation Calibration for Collaborative Robots

Abstract

Position accuracy is crucial for collaborative robots (cobots) to perform high-precision tasks. In traditional calibrations, the reduction in accuracy caused by multi-directional position accuracy variation (MPAV) remains challenging to address due to the lack of knowledge about the main influence factors. This article proposes a MPAV calibration method by mechanism modeling and error prediction to enhance the position accuracy of cobots. First, the factors leading to MPAV are analyzed using the kinematic model, and error properties (extrema, symmetry, and directionality) are investigated through measurement experiments. Based on the derived error properties, a position error model is built using Preisach method and kinematic Jacobian mapping, yielding accurate predictions for compensating MPAV error. Then, an interpolation-based prediction for MPAV error is constructed to overcome the dependence on dual encoders at joints in traditional identification. Combining the error model with the prediction method, MPAV error can be compensated, thereby enhancing position accuracy. Finally, calibration experiments under four different joint torque conditions are conducted on a six degrees of freedom (DoF) cobot. The results show that position accuracy can be evidently improved compared with existing methods, verifying the effectiveness of the proposed method.