Continuous Distributed Micro/Nano Fiber Tactile Skin Enabled by Optical Frequency Domain Reflectometry and Deep Learning

Existing fiber‐based tactile skins struggle to achieve direct spatial localization due to quasi‐distributed or pointed architecture, which compromises spatial resolution and scalability, especially for large‐area applications requiring dense force mapping. Our work addresses this by integrating continuous micro/nano fiber (MNF) arrays with optical frequency domain reflectometry (OFDR)‐based distributed sensing. OFDR provides direct axial localization along the fiber through Rayleigh backscattering frequency‐position mapping, while transverse localization is achieved through the sensitive wavelength shift responses of multiple MNFs and deep learning demodulation. A bio‐inspired silicone/PDMS bilayer mimics human dermal mechanics, where the PDMS‐encapsulated MNFs achieve ultrahigh force sensitivity (4683 pm/N) through enhanced evanescent field and small size. The cross‐correlation spectral image replaces conventional curve to provide multi‐dimensional strain features through rich spatial‐frequency joint correlation information, resembling biological decision‐making that relies on multi‐dimensional information. Powered by a physics‐augmented ResNet‐18 model trained on expanded spectral data (71.43% synthetic samples), the system decouples force‐distance dependencies and surpasses the OFDR's 1.48 mm spatial resolution. Simultaneous 2D contact localization (48.5 m/55.3 m MAE) and 9.6 mN force resolution are finally achieved across a 50 40 mm 2 area. This work bridges micron‐level resolution with centimeter‐scale coverage, advancing distributed tactile intelligence for safe human‐robot collaboration.