Deep Learning-Enhanced All-Nanofiber Sensors for Pressure–Temperature Sensing and Stimulus Discrimination

Abstract Flexible sensors for elderly-assistive robots typically chase high sensitivity, while comfort and multisensory capabilities are largely ignored. Here, we break this pattern with a deep-learning decoupled sensor based on tunable all-nanofiber membranes. Unlike conventional sensors that rely on multiple complex signals, our device uses a single sensing mode and decouples the pressure–temperature signal. The all-nanofiber sensors not only offer excellent pressure–temperature sensing performances but also demonstrate outstanding elderly-friendliness, including biocompatibility, breathability, and degradability. A deep learning algorithm decouples the mixed signals, enabling accurate recognition of liquid temperature and volume in a cup, as well as classification of 15 daily objects during robotic grasping with 97.78% accuracy. This work provides a simple, ecofriendly, and high-performance sensing solution for elderly-assistive robotics and smart home systems.