A new calibration method for joint-dependent geometric errors of industrial robot based on multiple identification spaces

This paper proposes a new calibration method for joint-dependent geometric errors of six-DoF industrial robots. Chebyshev polynomials are adopted to characterize the high-order joint-dependent geometric error model, revealing the impact of strain wave gearing errors and other sources more accurately. This effort also brings higher observability index on condition of an appropriate order. Furthermore, the geometric errors are lumped into different groups according to different sensitivities and the corresponding identification models are also established. In this way, each identification subspace contains much fewer error parameters with similar sensitivity. The simulations prove that better measurement configurations can be acquired using the proposed method according to the evaluation of observability indices. For implementation, sensor systems are designed to be fixed on joint 3 and joint 6 respectively to establish the multiple identification spaces. Alternative strategy for the robot without mechanical interface on joint 3 is also provided. Based on this, a set of real calibrations are performed and the results with joint-dependent models and multiple identification spaces indicate better identification accuracies.

Automated summary

This study proposes a new method for calibrating joint-dependent geometric errors of six-DoF industrial robots, characterized by Chebyshev polynomials and sensitivity grouping for more accurate modeling and reduced parameter dimension. Sensor systems are designed for calibration, leading to improved measurement configurations and identification accuracy based on simulation evaluations.