Optical Fibre Bragg Based Sliding-tactile Sensing and Classification Training Method for Material Recognition
Pan Ruizhi,Feng Yan,Liu Hexiang,Wang Haoxiang,Zhang Hongpu,Zhang Yinxiang,Zhang Hua
DOI: https://doi.org/10.3788/gzxb20245302.0206006
IF: 0.6
2024-01-01
ACTA PHOTONICA SINICA
Abstract:With the development of smart robots,intelligent tactile sensing is increasingly applied in industrial production,which can greatly improve efficiency and accuracy.Compared with traditional electrical sensors,optical fiber Bragg Grating(FBG)sensors have significant advantages,such as flexibility,electromagnetic immunity,and small size.They also demonstrate high sensitivity and rapid response in perceiving strain and pressure.Current researches on FBG-based tactile sensing mainly focus on strain,temperature,sliding positioning and contact force deduced from the Bragg wavelength shift of FBG.However,there are relatively few researches on combining feature extraction,machine learning,and other cutting-edge technologies to achieve more sophisticated intelligent perception,such as material recognition. In this work,we presented a FBG based sliding-tactile sensing and classification training method for online material recognition by the differential properties of contact surface materials,such as roughness and stick-slip phenomenon.We developed a horizontal two-layer silicone rubber covered FBG sensing unit and its sliding-tactile perception system.When sliding on the certain material,a continuous strain exerts to FBG through the silicone rubber sensing unit and FBG's response changes. To classify efficiently,this paper extracted the mean maximum difference λB(max)--λB,extreme difference λB(max)-λB(min),and standard deviation Δλ(std)of the FBG's wavelengths as the three-dimensional feature for mapping the material properties.And the classification training of the Support Vector Machine(SVM)algorithm and its classification model was developed.The results show that the classification accuracy is 96.6%for rough cloth,PLA and 800-grit sandpaper under the mixed dataset of 5 cm/s,10 cm/s and 15 cm/s sliding speeds.Compared with the direct wavelength and traditional mean/median feature classification methods,this three-dimensional feature-based method exhibits superior classification capability and adaptability. In order to achieve further intelligent applications,this paper also designs an interactive computer control system,including wavelength acquisition,speed control and material recognition result display.It can control the sliding speed and online material recognition as well.Utilizing the prediction function trainedModel.predictFcn(t_test),the corresponding predicted results were presented after extracting three-dimensional features.In 36 tests,5～15 cm/s random sliding speed(3 types of materials×3 samples×4 times slip)were carried out,and the correct predictions were 34 tests,which verifies that this method is effective and accurate. This work indicates that the FBG sensor has great potential in the field of material recognition by slip-tactile sensing.The research results can provide a novel online material recognition method for intelligent sensing robots.