An Ultrarobust Dual-Modal Neural Fiber With In-Sensor Computing

Touch behavior constitutes the most direct physical manifestation of human intentions, meaning that intention decoding requires human-machine interfaces to simultaneously sense touch position and pressure. Conventional systems, however, rely on physically stacked discrete sensors and von Neumann architectures, causing data redundancy and impaired wearability. Here, we present an in-sensor-computing dual-modal neural fiber (ID neural fiber) that overcomes this fundamental limitation. The ID neural fiber achieves an unprecedented position resolution of < 5 mm and a pressure detection limit of 0.6 N, combined with exceptional robustness-sustaining 300 times load-bearing capacity and exhibiting < 8% performance fluctuation under bending. Coupled with flexibility, waterproofing, and reconfigurability, the ID neural fiber can be seamlessly integrated into textiles for reliable operation in harsh environments, including underwater robotic control. Crucially, the ID neural fiber enables in-sensor computing that dramatically increases the information density and allows synchronized feature fusion through programmable logic gates, enabling efficient real-time intention recognition. To capture touch habits, an intelligent system with the ID neural fiber achieves a remarkable identification accuracy of 98.21%. This work establishes a new material-architecture paradigm that unifies ultrarobustness, dual-modality, and in-sensor computing, setting the stage for next-generation wearable human-machine interfaces with both comfort and intelligence.