Intrinsic Environmental Referencing in Coaxial Nanofiber Yarns Enables Humidity‐Robust Triboelectric Tactile Recognition

ABSTRACT

Triboelectric tactile sensors coupled with machine learning achieve high material recognition accuracy under controlled conditions, but ambient humidity introduces a domain shift that degrades performance in real‐world settings. Here we develop a coaxial porous nanofiber sensing yarn that embeds environmental referencing directly within its sensing structure. The yarn combines triboelectric material recognition with a capacitive baseline ( C ₀ ) whose monotonic humidity dependence serves as an intrinsic environmental indicator, co‐localized and temporally synchronized with the sensing signal. Using C ₀ to partition data into humidity‐defined bins, a humidity‐adaptive learning framework recovers cross‐humidity classification accuracy from 70.2% to 95.7% without external sensors. The capacitive mode simultaneously provides pressure sensitivity enhanced approximately to 2.74 times over dense‐coated yarns, stability over 6000 cycles, ultralight‐object detection, and stiffness discrimination. Integrated into a soft robotic gripper, a gated multimodal fusion network achieves 99.9% classification accuracy across five object categories and 30‐70% relative humidity. This work shows that environmental sensitivity in sensing systems, rather than being an obstacle to suppress, can serve as an intrinsic information source for adaptive perception.