AI 中文总结
本研究针对SAGINs的耦合能源需求，提出原生适用于HAP的智能体AI框架，经灾难案例验证，可提升能源效率等性能，为SAGINs可持续发展提供关键方向。
AI 中文摘要
空天地一体化网络（SAGINs）可拓展连接范围，但其通信、计算及平台运行会产生紧密耦合的能源需求。太阳能驱动的高空平台（HAPs）结合了持久的区域覆盖、可再生能源采集及机载计算能力，有望成为理想的中间层。然而，要实现这一潜力，仅优化单个链路或处理器是不够的，因为无线电传输、任务执行、回传链路使用及电池维护共享同一能源预算。为此，我们提出一种原生适用于HAP的智能体人工智能框架，该框架持续感知通信、计算、能源、移动性及任务状态，调用定量工具进行预测与验证，并通过闭环控制回路协调可执行动作。此外，该框架采用多时间尺度设计，将快速无线电控制与任务编排及长期能源规划分离。我们通过一个灾难恢复案例研究说明，该框架如何应对回传链路拥塞、流量激增及太阳能发电下降的情况，与其他基线方法相比，其在能源效率、任务完成率及延迟方面均有所提升。最后，我们指出可信控制、多HAP协同编排及数字孪生辅助的终身自适应是迈向可部署、可持续且弹性SAGIN智能的关键步骤。
英文摘要
Space-Air-Ground Integrated Networks (SAGINs) can extend connectivity, but their communication, computing, and platform operations create tightly coupled energy demands. Solar-powered High-Altitude Platforms (HAPs) offer a promising middle layer by combining persistent regional coverage, renewable-energy harvesting, and onboard computing. However, realizing this potential requires more than optimizing individual links or processors, as radio transmission, task execution, backhaul use, and battery preservation share a common energy budget. Therefore, we introduce a HAP-native Agentic AI framework. It continuously perceives communication, computing, energy, mobility, and mission states; invokes quantitative tools for prediction and verification; and coordinates executable actions through a closed control loop. Then, a multi-timescale design separates fast radio control from task orchestration and long-term energy planning. Furthermore, a disaster-recovery case study illustrates how the framework responds to backhaul congestion, traffic surges, and declining solar generation, improving energy efficiency, task completion, and latency over other baselines. We finally identify trustworthy control, collaborative multi-HAP orchestration, and digital-twin-assisted lifelong adaptation as key steps toward deployable, sustainable, and resilient SAGIN intelligence.
