Scorecard Model-Based Motion Stability Evaluation of Manipulators for Robot Training

Motion stability is vital for the industrial application of robotic manipulators; therefore, its evaluation becomes an important task for robot training. This paper proposes an evaluation system based on the scorecard model to evaluate the stability performance of manipulators. Firstly, the wavelet modulus maximum filtering method is employed to divide the vibration signal into structural vibration components and motion-generated vibration components. Secondly, the evaluation index (EI) is designed based on the processed data to judge whether the vibration signal of the moving process is stable. Finally, the scorecard model is proposed to evaluate the stability of vibration samples, including the process of attribute extraction, binning, WOE coding, feature selection, logistic regression, and score transformation. Additionally, an endpoint localization method is employed to segment the vibration signal of a continuously moving manipulator into single-step motion segments, making it possible to score the motion subprocesses. Experiments are conducted on physical robotic arms to verify the effectiveness of the proposed scheme, and the results show that the model can reasonably score the motion stability of robotic arms for robot training.