Deep Learning Based Singular Spectrum Analysis for Realization of Wideband Force Sensing

This paper proposes a new approach for realization of wideband sensor-less force sensing based on deep learning based singular spectrum analysis. The force sensation function is performed by a disturbance observer. The wideband force sensing is realized at high value of observer pole. To extract force information from the noisy estimation of the disturbance observer, the deep learning based singular spectrum analysis is designed with consideration of variable noise characteristics during operation. The deep learning algorithm is employed to design the online estimation of the embedded dimension, which is related to the noise extraction performance of the singular spectrum analysis. The effectiveness of the proposed method is verified by numerical simulations and simulations based on experimental results.