Related Work

Since 2017, 3GPP has been continuously releasing technical specifications and reports regarding the integration of satellites and 5G [4], [5], [27]–[30], sparking widespread discussion in the academic community about mobile satellite networks [31][33]. Due to recent breakthroughs in technology and cost control for LEO satellites, as well as their inherent advantages over higher orbit satellites, many efforts have been made to integrate LEO satellites with mobile networks.

自2017年以来，3GPP已持续发布关于5G与卫星融合的技术规范和报告[4], [5], [27]–[30]，这引发了学术界关于移动卫星网络的广泛讨论[31]-[33]。由于近期在LEO卫星技术和成本控制方面的突破，以及其相比高轨卫星的固有优势，已有许多工作致力于将LEO卫星与移动网络相集成。

One mainstream approach is to modify the network structure to accommodate the movement of LEO satellites [34]–[38]. This includes discussing the deployment locations of network functions and introducing new network functions. However, most of these works do not focus on user plane issues or fail to provide sufficient improvements in reducing latency. A recent work [34] aims to shorten control plane latency by deploying part of the core network functions on satellites. However, it overlooks user plane issues, resulting in users still experiencing long end-to-end latency.

一种主流方法是修改网络架构以适应LEO卫星的移动性[34]–[38]。这包括讨论网络功能的部署位置以及引入新的网络功能。然而，这些工作中的大部分并未关注用户面问题，或未能在降低时延方面提供足够的改进。近期的一项工作[34]旨在通过在卫星上部署部分核心网功能来缩短控制面时延，但它忽略了用户面问题，导致用户仍经历较长的端到端时延。

Another approach attempts to overcome satellite mobility from a higher perspective by introducing new anchor management mechanisms without modifying the mobile networks themselves [26], [39], [40]. These efforts often focus only on the latency variations caused by satellite mobility, rather than considering the entire end-to-end path. Work [26] proposes a global mobility management mechanism, which provides low-latency global internet service to users through an anchor manager and distributed satellite anchor points. However, this mechanism focuses on latency changes within the satellite network and can only allocate anchor points that meet latency requirements rather than the optimal anchor point.

另一种方法试图从一个更高的视角，通过引入新的锚点管理机制来克服卫星移动性，而无需修改移动网络自身[26], [39], [40]。这些工作通常仅仅关注由卫星移动性引起的时延变化，而未考虑整个端到端路径。文献[26]提出了一种全局移动性管理机制，通过一个锚点管理器和分布式的卫星锚点为用户提供低时延的全球互联网服务。然而，该机制关注的是卫星网络内部的时延变化，且只能分配满足时延要求的锚点，而非最优锚点。

Our proposed architecture involves deploying the network function (i.e., S-UPF) on satellites. However, by redesigning the PDU session establishment process, we do not introduce additional overhead on the control plane. On the other hand, by expanding the available anchor points and comprehensively considering multiple end-to-end paths, we achieve a significant reduction in end-to-end latency, surpassing existing schemes. We consider both the user plane and the control plane and conduct comprehensive system-level experiments on a real data-driven platform, ensuring that the experimental results closely reflect real-world scenarios.

我们提出的架构涉及到在卫星上部署网络功能（即S-UPF）。然而，通过重新设计PDU会话建立流程，我们并未在控制面上引入额外的开销。另一方面，通过扩展可用锚点并综合考虑多条端到端路径，我们实现了端到端时延的显著降低，超越了现有方案。我们同时考虑了用户面和控制面，并在一个真实数据驱动的平台上进行了全面的系统级实验，确保了实验结果能紧密反映真实世界的场景。