Mechnical Design Systems Laboratory, Hiroshima University

Industrial robots are increasingly considered as a potential alternative especially for machining tasks while the poor positioning accuracy prevented it from broader application. Model-based compensation, which improves the accuracy of a nominal model by incorporating additional error parameters, is acknowledged as an effective and efficient solution. In such a compensation, the precision of the kinematic model’s predictions is crucial for achieving high accuracy robotic machining. Many researchers employed the conventional Denavit-Hartenberg (D-H) error model, which describes the kinematic errors by positional and orientational deviations of axis average lines. However, it has been noticed that the identification results of D-H errors could vary with the posture of downstream axes. Such phenomena cannot be properly captured by a set of static D-H error parameters. This research introduces a novel modelling scheme that contains posture-dependent D-H errors. Validation results demonstrate that the proposed posture-dependent model improves the positioning accuracy of the investigated robot with satisfactory performance, which reveals the posture-dependent properties of D-H errors.

Keywords: industrial robot, error prediction, posture-dependent errors, accuracy