BigFiberNet: LLMs and Fabric Computing Empowered Large-Scale Non-Disturbance Mobile Sensing Networks

Benefiting from 5G advancements, the integrated sensing and communication capability nowadays is further enhanced by the emerging intelligence of large language models (LLMs), which are reshaping various mobile applications. In parallel, fabric computing technology demonstrates unique potential for measuring physical parameters, complementing 5 G and LLM capabilities in communication and data analytics. Its inherent user-friendliness and seamless integration into daily life enable non-disturbance sensing. However, the limited computing capacity and network bandwidth of edge fabric devices restrict the functionality of LLMs. Ultra-dense multimodal sensory data will quickly reach the system storage limit and block data transmission. When fabric computing meets with 5G and LLMs, this article proposes “BigFiberNet”, a novel architecture for large-scale mobile networks that enables non-disturbance sensing and cognition. Given the flexibility and diversity of LLM tasks, user demands for multimodal data vary depending on the contextual scenario. BigFiberNet dynamically extracts semantics from raw data, tailoring it to meet diverse user requirements based on semantic relevance. This allows the LLMs to derive insights from human activities and physiological health metrics efficiently. We implement the BigFiberNet platform, demonstrating its scalability and adaptability through experiments and case studies.