Space-air-ground Integrated Networks With Task-driven Connected Intelligence

Hui Gao, Ruohan Cao, Wenjun Xu, Caixia Yuan, Hsiao-Hwa Chen

Abstract: Space-air-ground integrated network (SAGIN) aims to offer seamless data connectivity across various vertical layers, including satellites and aeri-ai/ground/marine vehicles/ships at different altitudes. In this article, we foresee that with the rapid advancement of platforms' capabilities in sensing, computing, communication, and intelligence, traditional communication nodes will evolve into intelligent agents to facilitate task-driven functional networking across different SAGIN segments. In particular, we propose a novel task-driven multi-agent intelligent networking (MAIN) architecture for connected intelligence in SAGIN. Specifically, the intelligent agents in MAIN enjoy a high level of semantic intelligence fueled by advanced large language models with rich common and domain knowledge, which enhance the sensing, decision-making, control, and communication functions to meet the requirements of various complicated tasks and improve task-accomplishment efficiency. Furthermore, a novel intent-driven cross-domain management and control mechanism is conceived for MAIN, wherein the physical entity, communication, and mission intent domains are all managed and controlled by an intent-driven plan to comprehensively orchestrate the cross-domain functions and resources within SAGIN for specific tasks. Finally, we identify the challenges in the technical transition toward task-driven connected intelligence with MAIN. This work sheds light on task-driven multi-functional MAIN within SAGIN, highlighting the significance of semantic intelligence in the designs of task-driven networks.

External IDs:dblp:journals/wc/GaoCXYC25