Enabling 6G and Beyond Network Functions From Space: Challenges and Opportunities

这篇文章算是一篇 NTN + 5G 的综述, 讲得很仔细, 非常适合作为入门材料. 了解一下当前的行情与技术背景

1. 背景
   • 空间蜂窝网络的潜力: 卫星因其广阔的地面覆盖范围, 在其上部署RAN的话可消除地面网络的死区
2. 相关工作
   • GEO + bent pipes: 过于古早. 仅在中继 UE 和地面站之间的无线电信号, 不进行星载处理
   • 现有stateful架构: 存在根本性问题!
     • 依赖有状态的逐跳会话来绑定 UE_ID、QoS、计费和安全等状态信息, 并假设基础设施是固定的、始终在线的和可信的
     • 然而, LEO 具有极端移动性与可扩展性
3. 我们的做法: "逐跳+基于session的stateful" -> "无状态+按需蜂窝服务"
   • UE辅助: 将蜂窝功能状态卸载到 UE
     • UE 在初始注册到归属网络 (home) 后, 会复制并存储其state, 形成一个分布式状态存储库
     • 这些state在 LEO 卫星移动时保持本地化, 避免了不必要的会话迁移
   • 实际效果:
     • 路由: 根据目标IP, 每个LEO卫星无状态据dst转发
     • 发起:
       1. UE掌握state, 可以根据自身需要灵活选择 access sat
       2. 在进行GSL链路建立时, 将state metadata上传至sat: 认证成功就星上直接安装network func; 反之回滚到基础情况, 确保运营商始终保持局面掌控

---

Cellular networks from space, powered by recent technological advances in satellite mega-constellations, promise to expand operators’ service areas to anywhere on Earth for commercial revenues and social goods. While the existing 5G has attempted to support satellites via its non-terrestrial network enhancements, its hop-by-hop stateful session-based network architecture impedes this progress due to various functionality, scalability, reliability, and security concerns under the mega-constellation’s extreme mobility in harsh outer space. This article investigates these issues and explores how to remove this architectural barrier for 6G and beyond from space. We propose a user-centric design that shifts from the hop-by-hop session-based to stateless on-demand cellular service. We discuss how this paradigm shift can stabilize the cellular function inter-networking under extreme satellite mobility, refactor in-orbit cellular functions to be stateless with the assistance of local user equipment, and foster the efficient use of satellites for low capital costs and large-scale deployments.

在巨型卫星星座近期技术进步的推动下，天基蜂窝网络有望将运营商的服务区域扩展至全球任意角落，以实现商业营收和社会公益。尽管现有的5G已尝试通过其非地面网络（NTN）增强来支持卫星，但其“逐跳、状态化、基于会话”的网络架构，在巨型星座于严酷外太空环境下的极端移动性面前，引发了功能性、可扩展性、可靠性和安全性等多方面的担忧，从而阻碍了这一进程。

本文旨在探究这些问题，并探索如何为天基6G及未来网络消除这一架构障碍。

我们提出了一种“以用户为中心”的设计，该设计从“逐跳、基于会话”的服务转变为“无状态、按需分配”的蜂窝服务。

我们将讨论这种范式转变（paradigm shift）如何在极端的卫星移动性下稳定蜂窝功能的互联互通，如何借助本地用户设备（UE）的辅助将“在轨”蜂窝功能重构为无状态，以及如何促进卫星的高效利用，以实现低资本成本和大规模部署。

Space is the next frontier for cellular networks. Thanks to their broad terrestrial coverage, satellites can potentially eliminate terrestrial networks’ dead zones for the 2.7 billion “unconnected” global users via 5G, 6G, and beyond. Recent technological advances in direct-to-cell low-Earth orbit (LEO) satellite megaconstellations, such as SpaceX’s Starlink, AST SpaceMobile, Iridium, and Globalstar, further accelerate this paradigm’s real deployment by placing satellites closer to phones/Internet of Things (IoT) devices for better radio quality, faster data speed, cheaper hardware, and lower power consumption. They can significantly save operators’ capital costs for terrestrial infrastructure in underserved areas and expand their service anywhere on Earth to enable new subscribers for commercial revenue and social good.

Mobile satellites have been operational for decades since the 2G era. Early mobile satellites are single-hop, physical-layer bent pipes in the geostationary orbit (GEO) at an altitude of 35,786 km. As we show, while feasible for old-generation cellular networks in geostationary satellites, this simple bent pipe model has suffered from low network service coverage, missed 4G/5G radio processing deadlines, and unaffordable bandwidth demands in recent LEO megaconstellations due to its heavy reliance on remote ground stations. To this end, modern LEO satellites, like Starlink and AST SpaceMobile, have started to adopt onboard upper-layer 4G/5G network functions for scalable and performant services. The ongoing 3rd Generation Partnership Project (3GPP) 5G nonterrestrial network (NTN) standardization is also exploring the feasibility of offloading radio access and core network functions to satellites.1

However, enabling cellular functions in LEO satellites is hard. Unlike terrestrial infrastructure, LEO satellites in a megaconstellation move extremely fast at approximately 7.6 km/s in harsh outer space and only have intermittent connectivities to ground stations for remote control. This new scenario violates terrestrial cellular networks’ fundamental requirement for fixed, always-on, and trusted infrastructure to establish and maintain hop-by-hop stateful sessions for traffic delivery, mobility management, quality of service (QoS), billing, and other carrier-grade services. Our analysis shows that, if placed in LEO satellites, today’s stateful session-based radio access and core network functions will exhaust satellites with signaling storms, stop functioning repetitively under intermittent connectivity to remote ground stations and random failures, and become prone to attacks in the extreme case.

This article explores an alternative architecture to natively support satellite mega-constellations in 6G and beyond. We observe that rather than hop-by-hop stateful sessions, a stateless session-free architecture is more feasible for satellites under extreme mobility in harsh outer space. The challenge, however, is how to retain carrier-grade services after eliminating hop-byhop stateful cellular sessions.

To this end, we explore a usercentric design to leverage local user equipment (UE) as a naturally stable, scalable, resilient, and secure distributed state repository for fast-moving satellites. By offloading cellular function states to UEs, this paradigm can potentially stabilize the cellular function internetworking under extreme LEO mobility, refactor satellites’ onboard cellular functions to be stateless for scalability, localize their state management with UE assistance for reliability and security, and improve the efficient use of satellites for low capital costs and large-scale deployments. In addition to being beneficial for satellites, this usercentric design can also accelerate the terrestrial cellular network’s signaling procedures for low-latency services, and it facilitates seamless integration between satellite and terrestrial cellular networks. We review recent progress in this direction and discuss various research opportunities.

太空是蜂窝网络的下一个前沿。凭借其广泛的地面覆盖能力，卫星有潜力通过5G、6G及未来网络，为全球27亿“尚未连接”的用户消除地面网络的信号盲区。近期，“直连蜂窝”（direct-to-cell）的近地轨道（LEO）巨型星座（例如SpaceX的Starlink、AST SpaceMobile、Iridium和Globalstar）在技术上取得了进步，它们将卫星部署得更接近手机或物联网（IoT）设备，以获得更好的无线电质量、更快的数据速率、更廉价的硬件和更低的功耗，从而进一步加速了这种范式（paradigm）的实际部署。它们可以显著节省运营商在服务欠缺地区（underserved areas）的地面基础设施资本成本，并将其服务扩展到地球上任何地方，以吸纳新用户，从而实现商业营收和社会公益。

自2G时代以来，移动卫星已经运行了数十年。早期的移动卫星是位于35786公里高度的地球静止轨道（GEO）上的单跳、物理层“弯管”（bent pipes）。正如我们将展示的，虽然这种简单的“弯管”模型对于GEO卫星上的老一代蜂窝网络是可行的，但在近期的LEO巨型星座中，由于其高度依赖远程地面站，该模型已饱受网络服务覆盖率低、错过4G/5G无线电处理时限（deadline）以及带宽需求高昂等问题。为此，现代LEO卫星（如Starlink和AST SpaceMobile）已开始采用“星上”（onboard）的高层4G/5G网络功能，以实现可扩展和高性能的服务。正在进行的第三代合作伙伴计划（3GPP）5G非地面网络（NTN）标准化工作，也正在探索将无线接入网和核心网功能卸载（offloading）到卫星上的可行性。

然而， 在LEO卫星上实现蜂窝网络功能非常困难。 与地面基础设施不同，LEO巨型星座中的卫星在严酷的外太空中以大约7.6公里/秒的速度极端快速地移动，并且它们与地面站的连接是间歇性的（intermittent），仅用于远程控制。 这种新场景违反了地面蜂窝网络对“固定的、永远在线的、可信的”基础设施的基本要求——即需要（这些设施）来建立和维护“逐跳、状态化”（hop-by-hop stateful）的会话， 以实现流量传输、移动性管理、服务质量（QoS）、计费和其他电信级（carrier-grade）服务。

我们的分析表明， 如果将当今“基于状态化会话”的无线接入网和核心网功能部署在LEO卫星上，将会因信令风暴（signaling storms）而耗尽卫星资源，在与远程地面站的间歇性连接和随机故障下反复停止工作， 并在极端情况下易受攻击。

本文探索了一种替代架构，以在6G及未来网络中原生（natively）支持卫星巨型星座。我们观察到，与“逐跳、状态化会话”相比， “无状态、无会话”（stateless session-free）的架构 对于在严酷外太空极端移动性下的卫星更为可行。然而，其挑战在于，在取消了“逐跳、状态化”的蜂窝会话之后，如何保留电信级的服务。

为此，我们探索了一种“以用户为中心”（usercentric）的设计，利用本地用户设备（UE）作为一个“天然稳定、可扩展、有韧性且安全”的分布式状态存储库（repository），来服务于快速移动的卫星。

1. 通过 将蜂窝功能的状态卸载到UE ，这种范式有潜力在极端的LEO移动性下稳定蜂窝功能的互联互通， 将卫星的星上蜂窝功能重构（refactor）为无状态 以实现可扩展性，在UE的辅助下将其状态管理本地化以保证可靠性和安全性，并提高卫星的使用效率以实现低资本成本和大规模部署。
2. 除了有益于卫星，这种“以用户为中心”的设计还可以加速地面蜂窝网络的信令流程以支持低延迟服务，并促进天基（satellite）蜂窝网络和地面（terrestrial）蜂窝网络之间的无缝集成。

我们回顾了这一方向的最新进展，并讨论了各种研究机遇。

WHY CELLULAR NETWORKS FROM SPACE?

Terrestrial cellular networks, such as 4G and 5G, have successfully served billions of users. To offer ubiquitous access, terrestrial cellular networks should deploy radio access networks (RANs) with numerous base stations to cover broad geographic areas and bridge them to the Internet via core networks (Figure 1). While profitable in urban areas with sufficient subscribers, such capital-intensive infrastructure loses mobile operators’ revenues when covering rural areas with few subscribers and is even undeployable over oceans, on airplanes, and in foreign countries, thus leaving 2.7 billion global users and numerous IoT devices unconnected. Furthermore, unexpected natural disasters and wars can easily disrupt these infrastructures.

To this end, mobile operators seek to complement their terrestrial cellular networks with low-cost, resilient infrastructures with broad coverage, such as satellites. As shown in Figure 1, satellites operate at high altitudes to offer broad network coverage at lower costs. They can offer direct cellular access to phones/IoT devices and save mobile operators’ capital costs in underserved areas. Moreover, satellites’ global coverage lets a local mobile operator expand to international roaming services without relying on competitive operators’ expensive, slow, and sometimes untrusted foreign infrastructure. This enables new subscribers for mobile operators for commercial revenue and social good. Since 2022, Apple has partnered with Globalstar to offer direct-to-satellite communications for iPhone 14/15 and above. Huawei’s Mate 60 Pro þ phones can also directly communicate with Tiantong and BeiDou satellites.

Traditional direct-to-cell satellites operate in the GEO at an altitude of 35,786 km. While excellent for broad coverage, GEO satellites are unfriendly to commodity phones/IoT devices since their distant transmission is power hungry, slow, and noisy. Dedicated satphones with high-gain antennas can alleviate this issue, but they are not widely available or affordable to most consumers. Instead, modern satellites like Starlink, Iridium, Globalstar, and AST SpaceMobile operate in LEOs at an altitude of 340–2,000 km to be closer and friendly to phones/IoT devices. Due to each LEO satellite’s smaller coverage, a constellation with tens to thousands of satellites is typically adopted for global coverage.

以4G和5G为代表的地面蜂窝网络已成功服务了数十亿用户。为提供“无处不在”的接入服务，地面蜂窝网络必须部署包含海量基站的无线接入网（RANs）以覆盖广阔的地理区域，并通过核心网（Core Networks）将其桥接至互联网（如图1所示）。

尽管此类资本密集型基础设施在用户充足的城市地区能够盈利，但在覆盖用户稀少的农村地区时，则会造成移动运营商的收益损失，甚至根本无法部署在海洋、飞机上以及（其他）国家。这导致全球27亿用户和海量的物联网（IoT）设备仍处于未连接状态。此外，突发的自然灾害和战争极易摧毁这些基础设施。

为此，移动运营商寻求使用低成本、高韧性（resilient）且覆盖范围广的基础设施（例如卫星）来补充其地面蜂窝网络。

如图1所示， 卫星运行在极高的高度，能以较低成本提供广泛的网络覆盖。它们能为手机/物联网设备提供“直连”蜂窝接入， 并节省移动运营商在服务欠缺地区（underserved areas）的资本开销。

此外，卫星的全球覆盖能力使本地移动运营商得以扩展国际漫游服务，而无需依赖（其他）竞争运营商昂贵、缓慢且时而不可信的海外基础设施。这为移动运营商带来了新用户，从而实现商业营收和社会公益。自2022年起，苹果公司已与Globalstar合作，为iPhone 14/15及以上机型提供卫星直连通信。华为的Mate 60 Pro+手机也能直接与天通（Tiantong）和北斗（BeiDou）卫星通信。

传统的“直连蜂窝”卫星运行在35,786公里的地球静止轨道（GEO）上。尽管GEO卫星在提供广泛覆盖方面表现出色，但由于其长距离传输功耗高、速度慢且噪声大，因此对商用（commodity）手机/物联网设备并不友好。配备高增益天线的专用卫星电话（satphones）可以缓解此问题，但它们未被广泛应用，且大多数消费者也无法负担其价格。

取而代之的是，现代卫星（如Starlink、Iridium、Globalstar和AST SpaceMobile）运行在340至2,000公里高度的近地轨道（LEOs），以便更接近手机/物联网设备并对其提供更友好的支持。由于单颗LEO卫星的覆盖范围较小，通常需要采用包含数十到数千颗卫星的星座（constellation）来实现全球覆盖。

WHAT’S NEW IN SATELLITES?

Different from fixed terrestrial network infrastructure or GEO satellites, commodity hardware-based LEO satellites move fast in harsh, crowded, and unreliable outer space on a global scale:

• Extreme mobility: From Kepler’s laws, satellites at lower altitudes move faster. To date, operational LEO satellites, like Starlink, move at approximately 7.6 km/s. Moreover, except for GEO satellites, LEO satellites’ orbital motions are asynchronous to Earth’s rotation. Both result in complex relative motions between every LEO satellite and the rotating Earth.

• Harsh operation environment: LEOs are highly crowded, with about 8,300 satellites and 27,000 pieces of space junk (e.g., abandoned rocket bodies and pieces from satellite breakups). The recent megaconstellation further congests these orbits and raises the risks of collisions. 3 Furthermore, LEO satellites using the commodity hardware are prone to failures in outer space. For instance, every one out of 40 Starlink satellites may have failed since they use low-cost commodity CPUs without hardening against radiation.

• Limited satellite capability: In contrast to terrestrial infrastructures, satellites are solar powered but need to conduct various power-hungry tasks, such as radio communication to terrestrial nodes, laser-based intersatellite data forwarding, and orbital maneuvers for collision avoidance. With power at a premium, it is difficult for satellites to perform heavy computations. Moreover, once launched into orbit, satellite hardware cannot be easily upgraded afterward. Given their five- to 10-year lifetime, satellites’ in-orbit processing capabilities can easily fall behind those of terrestrial infrastructure.

与固定的地面网络基础设施或GEO卫星不同，基于商用硬件（commodity hardware-based）的LEO卫星在全球范围内的严酷、拥挤且不可靠的外太空中高速移动:

• 极端的移动性： 根据开普勒定律（Kepler’s laws），较低轨道的卫星移动速度更快

  • 目前，在轨运行的LEO卫星（如Starlink）以大约7.6公里/秒的速度移动
  • 此外，除GEO卫星外，LEO卫星的轨道运动与地球自转是异步的（asynchronous）
  • 这共同导致了每颗LEO卫星与旋转的地球之间复杂的相对运动

• 严酷的运行环境： LEO轨道（LEOs）高度拥挤，充斥着约8,300颗卫星和27,000块太空垃圾（例如，废弃的火箭箭体和卫星破碎后产生的碎片）

  • 近期的巨型星座（megaconstellation）进一步加剧了这些轨道的拥堵，并增加了碰撞的风险
  • 此外，使用商用硬件的LEO卫星在外太空中容易发生故障。例如，Starlink卫星使用了未经辐射加固（hardening against radiation）的低成本商用CPU，导致每40颗卫星中就可能有1颗已经失效

• 受限的卫星能力： 与地面基础设施相反，卫星依靠太阳能供电，但需要执行各种高功耗任务

  • 例如与地面节点的无线电通信、基于激光的星间（intersatellite）数据转发以及用于避免碰撞的轨道机动（orbital maneuvers）
  • 在电力极其宝贵（at a premium）的情况下，卫星难以执行繁重的计算任务
  • 此外，卫星硬件一旦发射入轨，后续难以进行升级。鉴于其5到10年的使用寿命，卫星的在轨处理能力很容易落后于地面基础设施

THREATS FOR ORBITAL CELLULAR NETWORKS

• 核心冲突: 当前架构与LEO环境不兼容
  • 当今的 地面蜂窝网络采用“逐跳、状态化会话”（stateful hop-by-hop session）架构，它依赖于“固定的、永远在线的、可信的”基础设施来管理服务质量 (QoS), 计费和安全状态
  • LEO（近地轨道）卫星的特性——极端移动性、严酷的太空环境、不可靠的连接——完全违背了当前蜂窝架构的这些基本假设，从而引发了功能性、可扩展性、可靠性和安全性四大挑战
• 功能性挑战: Bent Pipe模式的失效
  • 传统的“弯管”模式（卫星仅作为物理层信号中继）在 LEO 4G/5G 网络中已不可行
  • 1. 服务覆盖不全： “弯管”模式要求用户和地面站必须同时位于卫星的覆盖范围内，导致远离地面站的广大区域无法获得服务
  • 2. 错过无线处理时限： 4G/5G的无线电功能（如IQ采样处理）要求在250毫秒（ms）内完成。如果通过LEO卫星将IQ数据中继回远程地面站，其往返时延（RTT）将比该时限高出一到两个数量级，导致无线功能崩溃
  • 3. 极端的带宽浪费： 向地面站中继原始的IQ采样数据（例如，一个5MHz信道就需要7.86 Gb/s）会迅速耗尽卫星本已有限的星间链路（ISL）和星地链路（GSL）带宽
• 可扩展性挑战：“星上处理”（Onboard Processing）模式的困境
  • 为解决“弯管”问题而将RAN（无线接入网）功能部署到卫星上（即“星上处理”），同样会因为“状态化会话”而面临严峻的可扩展性问题
  • 1. 信令风暴（Signaling Storms）： LEO卫星覆盖时间极短（例如Starlink约165.8秒）。卫星的快速移动迫使其覆盖下的成千上万用户频繁进行会话迁移（切换），从而产生压垮卫星的“信令风暴”
  • 2. 动态的多对多信任关系： 在卫星租赁模式下（移动运营商租赁卫星运营商的服务），卫星的移动性导致运营商与卫星之间的服务和信任关系需要频繁更新，这使得敏感会话状态的动态管理变得极其复杂
• 可靠性与安全性挑战：恶劣环境的威胁
  • 1. 卫星故障： LEO卫星易受辐射、太空碎片等影响而发生故障，链路也可能间歇性中断。“状态化”的会话流程非常脆弱，任何信令丢失或错误都可能导致整个服务流程中断
  • 2. 卫星攻击： 卫星暴露在（他国）安全域之外，易受劫持、窃听或干扰。在这些易受攻击的卫星上部署状态化功能，意味着大量用户的敏感状态（如鉴权密钥）面临着泄露给对手的风险

TN: stateful hop-by-hop session

目前 地面蜂窝移动网络的 stateful + hop-by-hop 特点:

这一条服务链上的每个人（每一跳）都必须协同合作，并且都持有关于你（会话）的一部分“状态”（内存/记录）

直观体会:

(1) 在蜂窝网络中（如4G/5G），网络功能被分为两个“平面”:

• 控制平面 (Control Plane): 负责“管理”你。比如你的身份认证、移动性管理（换基站）、会话建立、QoS策略协商等。这就像餐厅的“管家团队”
• 用户平面 (User Plane): 负责“传输”你的实际数据。比如你的微信消息、视频流。这就像餐厅的“传菜通道”

(2) flow:

从UE到RAN到AMF/SMF/UPF，每一个节点都必须为你的这一个会话保留一份“状态”记录

OPPORTUNITY: USERCENTRIC ORBITAL CELLULAR NETWORK FUNCTIONS

As we have seen, the current hop-by-hop stateful session-based terrestrial cellular architecture in 4G/5G impedes its native support for satellites due to various functionality, scalability, reliability, and security issues. To remove this architectural barrier, we think a stateless, session-free cellular architecture is worth consideration in the forthcoming 6G and beyond. The challenge, however, is how to retain carrier-grade services after eliminating hop-by-hop stateful sessions. This further raises three fundamental questions:

• What states must exist for carrier-grade services?
• Where are these states placed (if not in satellites)?
• How can orbital cellular functions work with them?

We show that a user-centric cellular architecture is promising to solve these questions. Since its origin, the cellular network has followed the infrastructurecentric design to place heavy signaling and states inside its RAN and core network, which suffers from extreme satellite mobility in harsh outer space (Figure 5). Instead, a usercentric architecture can better support satellites and other NTNs for three reasons:

• Stability: Compared to satellites, UEs’ geographic locations are more stable and invariant of extreme satellite mobility and, thus, more suitable to drive and guide the orbital cellular functions’ internetworking for carrier-grade services.
• Availability: As end users, UEs naturally form a scalable, resilient, secure state repository for fast-moving satellites. They have replicated critical carrier-grade service states after the initial registration. These states are almost always on unless UEs deactivate services or defunction.
• Locality: Unlike satellites exposed to foreign outer space, UEs mostly remain stable in local areas on Earth and, thus, are physically better protected against failures and attacks.

To this end, we follow the end-to-end design principle to simplify orbital cellular network functions with UE-side assistance, as shown in Figure 6. Based on UEs’ stable geographic locations, we can first build a stable satellite backhaul for cellular function internetworking. On top of this stable satellite network substrate, we refactor cellular functions in satellites by offloading their critical states to local UEs. This yields stateless cellular functions from space that are stable, scalable, and resilient to failures/attacks. At runtime, the UE itself drives the in-orbit cellular function orchestration on demand using its local states. This paradigm lets UEs and mobile operators leverage any available and potentially competitive satellites on demand and facilitates the efficient use of satellites for mobile and satellite operators. We next introduce recent progress and future research opportunities in each part.

诚如前文所见，当前4G/5G中“逐跳、状态化、基于会话”（hop-by-hop stateful session-based）的地面蜂窝架构，由于在功能性、可扩展性、可靠性和安全性方面存在诸多问题，已成为其原生支持卫星的障碍。

为移除这一架构障碍，我们认为在即将到来的6G及未来网络中，一种“无状态、无会话”（stateless, session-free）的蜂窝架构值得考虑。然而，其挑战在于，在取消了“逐跳、状态化会话”之后，如何保留“电信级服务”（carrier-grade services）。

这进一步引出了三个基本问题：

• 为实现电信级服务，哪些状态（states）必须存在？
• 这些状态应被放置在何处（如果不在卫星中）？
• 在轨（orbital）的蜂窝功能应如何与它们协同工作？

我们表明，“以用户为中心”（user-centric）的蜂窝架构有望解决这些问题。

自蜂窝网络诞生以来，它一直遵循“以基础设施为中心”（infrastructure-centric）的设计，将其重度的（heavy）信令和状态放置在RAN（无线接入网）和核心网内部， 这种设计在外太空严酷环境下的极端卫星移动性面前备受困扰（如图5所示）:

相反，一个“以用户为中心”的架构能更好地支持卫星和其他非地面网络（NTN），原因有三:

• 稳定性（Stability）： 与卫星相比，UE（用户设备）的地理位置更加稳定，并且不受（invariant of）极端卫星移动性的影响。因此，（UE）更适合用于驱动（drive）和引导（guide）在轨蜂窝功能的互联互通，以实现电信级服务
• 可用性（Availability）： 作为终端用户，UE为快速移动的卫星天然地构成了一个“可扩展、有韧性（resilient）且安全的状态存储库”（state repository）
  • 在初始注册后，它们已经复制（replicated）了关键的电信级服务状态
  • 除非UE主动停用服务或发生故障（defunction），否则这些状态“几乎永远在线”（almost always on）
• 本地性（Locality）： 与暴露在“域外太空”（foreign outer space）的卫星不同，UE大多稳定地保持在地球上的本地（local）区域，因此在物理上能更好地免受故障和攻击的威胁

为此，我们遵循“端到端设计原则”（end-to-end design principle），借助“UE侧辅助”（UE-side assistance）来简化在轨蜂窝网络功能（如图6所示）:

首先，基于UE的稳定地理位置，我们可以为蜂窝功能的互联互通构建一个 “稳定的卫星回程”（stable satellite backhaul）

在这个“稳定的卫星网络基底”之上，我们通过将卫星上蜂窝功能的关键状态“卸载”到本地UE ，来“重构”（refactor）这些功能。这便产生了“无状态的天基蜂窝功能”，这些功能是稳定的、可扩展的，并且对故障/攻击具有韧性。

在运行时，UE自身使用其本地状态，“按需”驱动“在轨蜂窝功能编排”。 这种范式（paradigm）使得UE和移动运营商能够“按需”利用任何可用的、且有潜在竞争力的卫星，并促进了移动运营商和卫星运营商对卫星的高效利用。

接下来，我们将介绍（这一方向的）最新进展以及未来在每个部分的研究机遇。

Stabilizing Orbital Function Internetworking

A prerequisite for functional cellular networks from space is to enable a stable backhaul to bridge the RAN and core network across satellites, ground stations, and UEs. This is hard because of the frequent topology changes under satellites’ orbital motions and maneuvers. While various routing solutions are available to adapt to topology changes in moderate-sized networks under mild mobility, they are not ready to tackle such unprecedented rapid topology changes on a global satellite megaconstellation scale. For example, proactive routing, such as distance-vector/link-state routing,10 label switching, 11and software-defined networking,12 suffers from repetitive routing recomputation or even reconvergence under frequent satellite link churns.2,7 Reactive routing, like Ad hoc On-Demand Distance Vector Routing and Distance Vector Routing, floods route requests on demand across the entire network and may exhaust resource-constrained satellites under frequent topology changes. Recent proposals 12 seek to exploit satellites’ predictable orbital motions for routing optimization but fall short in unpredictable satellite failures and chaotic maneuvers.3

Our usercentric backhaul routing driven by UEs’ geographic locations can potentially avoid these deficiencies. The key is that, as satellites’ primary service areas, Earth’s geographic locations are invariant of extreme satellite mobility. As exemplified in Figure 7, routing to UEs can be guided by their stable geographic locations rather than fast-moving serving satellites. Any satellite covering the target terrestrial destination can forward its traffic. This method is also nearly stateless: satellites do not need to maintain or frequently upgrade routing tables in response to topology changes; they drive their local routing decisions based on the destination UE’s location embedded into its address in packets. We have preliminarily explored this idea and demonstrated its potential to simplify satellite routing, 2 stabilize cellular service areas, 8 accelerate extreme mobility management, 13 and eliminate bandwidth and latency bottlenecks from remote ground stations. 7 We are further exploring this direction to support multiple orbital shells in real LEO satellites, enable hierarchical routing in hybrid GEO–LEO–terrestrial networks, simplify routing processing in resource-constrained satellites, enable QoS-aware routing for premium services, tolerate routing failures, and strengthen security and location privacy.

实现天基蜂窝网络功能的一个先决条件是，必须建立一个稳定的回程（backhaul），以桥接跨越卫星、地面站和UE的RAN（无线接入网）与核心网。由于卫星的轨道运动和机动（maneuvers）会导致频繁的拓扑变化，这一点很难实现。

尽管现有的各种路由解决方案，能够适应中等规模网络在轻度移动性下的拓扑变化，但它们尚无法应对全球卫星巨型星座规模上（如此）前所未有的快速拓扑变化。

例如：

• 主动路由（Proactive routing），如距离矢量/链路状态路由、标签交换和软件定义网络（SDN），在频繁的卫星链路抖动（link churns）下会遭受重复的路由重新计算甚至重新收敛的困扰
• 被动路由（Reactive routing），如AODV（Ad hoc On-Demand Distance Vector Routing），需要在全网按需泛洪（floods）路由请求，这在频繁的拓扑变化下可能会耗尽资源受限的卫星
• 近期的提议试图利用卫星可预测的轨道运动进行路由优化，但在不可预测的卫星故障和混乱的机动面前则无能为力

我们提出的 由UE地理位置驱动的“以用户为中心”的回程路由 ,有潜力避免这些缺陷。其关键在于，作为卫星的主要服务区域，地球的地理位置是恒定的（invariant），不受卫星极端移动性的影响。

如图7所示，发往UE的路由可以由其稳定的地理位置引导，而不是由快速移动的服务卫星引导。任何覆盖目标地面终端的卫星都可以转发其流量:

这种方法也近乎无状态（nearly stateless）： 卫星无需为应对拓扑变化而维护或频繁更新路由表；它们仅根据数据包地址中嵌入的目的UE位置来驱动本地的路由决策。

我们已初步探索了这一想法，并证明了其在简化卫星路由、稳定蜂窝服务区域、加速极端移动性管理以及消除远程地面站带来的带宽和时延瓶颈方面的潜力。我们正进一步探索此方向，以支持真实LEO卫星中的多轨道层（multiple orbital shells）、实现GEO-LEO-地面混合网络中的分层路由、简化资源受限卫星的路由处理、实现面向优质服务的QoS感知路由、容忍路由故障并加强安全性和位置隐私。

Refactoring Orbital Functions as Stateless

With our stable backhaul from space, the next step is appropriately splitting cellular functions across satellites, ground stations, and UEs. As shown in Figure 4, splitting RAN functions among satellites and ground stations will violate the radio processing deadlines and exhaust satellite link bandwidth. We, thus, focus on full-fledged RAN functions in each satellite. Furthermore, it is also of interest to enable core networks in satellites to facilitate orbital edge computing. Note that both RAN and core network functions today are stateful and involved in the hop-by-hop sessions. For scalability, reliability, and security under extreme satellite mobility in harsh outer space, these cellular functions should be refactored to be stateless and session free.

To this end, we propose a UE-assisted approach for stateless, session-free orbital cellular functions, as overviewed in Figure 8. As stated earlier, UEs naturally form a distributed state repository for cellular functions in satellites (which can be viewed as a user-side version of distributed 5G unstructured data storage functions14 ). After registering to the home, a UE has replicated and stored its session states by itself. These states are locally accessible to new satellites to enforce carrier-grade services for this UE, thus avoiding exhaustive state migrations between satellites and ground stations. They remain local despite LEO satellite mobility, thus avoiding unnecessary session migrations in Figure 2. Without being exposed to foreign locations, they are resilient to failures/attacks.

As our first attempt in this direction, SpaceCore8 has demonstrated the feasibility of offloading core network function states from satellites to UEs. Compared to the legacy 4G/5G NTN, this approach, SpaceCore, reduces signaling costs by 17.5–122.2Â, avoids sensitive state leakages, and is more resilient to service disruptions under failures. We are currently extending SpaceCore to enable stateless RAN functions; facilitate the efficient use of satellites by mobile operators; enforce more carrier-grade services, like QoS, billing, and orbital edge computing; and strengthen this approach’s security against malicious UEs and internal/external attackers.

In addition to satellites, this UE-centric architecture can also benefit terrestrial cellular networks and facilitate seamless space–terrestrial network integration. For example, data-plane parallelism via control-plane management 15 leverages this paradigm to accelerate 4G/5G’s control plane for low-latency data access. It formulates the control procedures as a distributed state management problem, localizes them with minimal involvement of remote core networks, and parallelizes them with UE-side state replicas. Without adding more bandwidth, it can achieve up to two orders of magnitude latency reduction in 5G.

Of course, stateless orbital functions are not without costs. By offloading states to devices, the stateless orbital functions may lower the operator’s controllability of critical functions, such as dynamic QoS or billing policies for some UEs. Moreover, some devices can be compromised, selfish, or even malicious. They may manipulate the offloaded states and raise security risks. Our current design resolves these limitations with home-controlled state updates. 6,8,15 The home network is the only participant that can issue, update, and synchronize the UE’s session states. It generates and digitally signs the session states to the UE during the initial registration. Afterward, the serving satellite validates the UE-side state signature. If this fails (e.g., the UE manipulates the states), the satellite rolls back to the legacy procedure by contacting the home network.

在我们稳定的天基回程基础上，下一步是在卫星、地面站和UE之间恰当地切分（splitting）蜂窝功能。如图4所示，在卫星和地面站之间切分RAN功能将违反无线电处理时限并耗尽卫星链路带宽。因此，我们专注于在每颗卫星上部署全功能（full-fledged）的RAN。此外，在卫星上启用核心网功能以促进“在轨边缘计算”（orbital edge computing）也同样值得关注。

值得注意的是，当今的RAN和核心网功能都是状态化的，并且都参与了“逐跳会话”。为了在严酷外太空的极端卫星移动性下实现可扩展性、可靠性和安全性，这些蜂窝功能应被重构（refactored）为无状态（stateless）和无会话（session-free）的。

为此，我们提出了一种 UE辅助（UE-assisted） 的方法，以实现在轨蜂窝功能的无状态和无会话（如图8所示）。

如前所述，UE天然地为卫星上的蜂窝功能构成了一个“分布式状态存储库”（distributed state repository）（这可被视为分布式5G非结构化数据存储功能的用户侧版本）。

1. UE在注册到归属网络（home network）后，已经由其自身复制并存储了其会话状态
2. 这些状态可供（后续的）新卫星在本地访问（locally accessible）
3. 以便为该UE执行电信级服务，从而避免了卫星和地面站之间耗尽式（exhaustive）的状态迁移

尽管LEO卫星在移动，这些状态仍保持在本地，从而避免了图2所示的不必要的会话迁移。由于无需暴露于（他国）外部位置，它们对故障/攻击具有很强的韧性（resilient）。

作为我们在此方向的首次尝试，SpaceCore 已经证明了将核心网功能状态从卫星卸载（offloading）到UE的可行性。与传统的4G/5G NTN（非地面网络）相比，SpaceCore这种方法将信令开销降低了17.5至122.2倍，避免了敏感状态的泄露，并且在故障下对服务中断更具韧性。我们目前正在扩展SpaceCore，以支持无状态的RAN功能；促进移动运营商对卫星的高效利用；执行更多的电信级服务（如QoS、计费和在轨边缘计算）；并加强该方法抵御恶意UE和内/外部攻击者的安全性。

除了卫星，这种“以UE为中心”的架构同样有益于地面蜂窝网络，并能促进“天地一体化网络”（space-terrestrial network integration）的无缝融合。例如，“通过控制平面管理实现数据平面并行化”（data-plane parallelism via control-plane management）就利用了这种范式来加速4G/5G的控制平面，以实现低延迟数据访问。它将控制流程表述为一个分布式状态管理问题，以最小化远程核心网参与的方式将其本地化，并利用UE侧的状态副本来实现并行化。在不增加额外带宽的情况下，它可以在5G中实现高达两个数量级的时延降低。

当然，在轨的无状态功能并非没有代价。通过将状态卸载到设备，无状态的在轨功能可能会降低运营商对关键功能（例如针对某些UE的动态QoS或计费策略）的可控性。此外，某些设备可能已被攻破（compromised）、是自私的、甚至是恶意的。它们可能会篡改（manipulate）被卸载的状态，从而引发安全风险。

我们当前的设计通过 “归属网络控制的状态更新”（home-controlled state updates） 来解决这些局限性

1. 归属网络是唯一可以发布、更新和同步UE会话状态的参与者
2. 它在初始注册期间生成会话状态，并对其进行数字签名后发送给UE
3. 随后，服务卫星会验证（validate）UE侧的状态签名
4. 如果验证失败（例如，UE篡改了状态），卫星将通过联系归属网络来回滚（rolls back）到传统流程

Orchestrating Orbital Cellular Functions

After decoupling stateful sessions from in-orbit cellular functions, the last step is coordinating satellites, UEs, and ground stations to retain carrier-grade services without hop-by-hop stateful sessions. We propose letting each UE initiate and drive the cellular function orchestration using its local states. This approach avoids state migrations between satellites and ground stations (thus reducing signaling storms), localizes the state retrieval without exposing sensitive information to outer space, and improves the signaling procedure reliability since the UE-side state replica is almost always on. Moreover, with its own states, each UE can select any available satellite with cellular functions to serve it. This unleashes more freedom for UEs and mobile operators to enjoy diverse and potentially competitive satellites from multiple satellite operators, which may improve the network coverage and save costs.

Figure 9 exemplifies this UE-driven cellular function orchestration in SpaceCore. 8 When the UE needs satellite services, it piggybacks its state replica to its serving satellite during the radio connection setup. If authorized to access these local states by the home network, the serving satellite can successfully decrypt and install them into its local RAN and core network functions for immediate data services. If the serving satellite fails to decrypt these states, it implies that the (malicious) UE may have manipulated this local state (e.g., for higher QoS or free of billing). In this case, the satellite rolls back to the legacy procedure in Figure 2 by contacting the home network through the remote ground stations. In this way, the home mobile operator retains full control of its carrier-grade services for UEs.

We are extending this basic procedure to support advanced carrier-grade services (e.g., local billing, QoS, and deep packet inspection) and enhance its verifiability without involving remote ground stations.

We would like to point out that the current design focuses on stateless backhaul and core networks. As shown in Figure 6, the full-fledged RAN is currently placed into each satellite. Unlike backhaul or core network states, RAN states are mostly local for the last-hop radio connectivity and, thus, free of signaling storms and leaks. Achieving a fully stateless RAN is hard and unnecessary due to its complex processing. That said, a lightweight RAN with fewer states is still desirable for resource-constrained satellites. It is possible to simplify RAN with UE-side state replicas, as exemplified in Figure 9 for the radio connectivity setup. We leave the lightweight RAN design for satellites as future work.

将在轨蜂窝功能与状态化会话解耦（decoupling）后，最后一步是协调卫星、UE和地面站，以便在没有“逐跳状态化会话”的情况下保留电信级服务。

我们建议让每个UE使用其本地状态来发起和驱动（initiate and drive）蜂窝功能的编排（orchestration）。

这种方法避免了卫星和地面站之间的状态迁移（从而减少信令风暴），将状态检索本地化（无需将敏感信息暴露于外太空），并提高了信令流程的可靠性（因为UE侧的状态副本几乎永远在线）。

此外，凭借其自身的状态，每个UE可以选择任何可用的、具备蜂窝功能的卫星来为其服务。 这为UE和移动运营商释放了更多自由，使他们能够享受到来自多个卫星运营商的、多样化的且有潜在竞争力的卫星服务，从而可能改善网络覆盖并节省成本。

图9举例说明了SpaceCore中这种UE驱动的蜂窝功能编排：

1. 当UE需要卫星服务时，它 在无线连接建立过程中，将其状态副本捎带（piggybacks）给其服务卫星
2. 如果服务卫星获得了归属网络的授权，可以访问这些本地状态，它就能成功解密并将其安装到本地的RAN和核心网功能中，以立即提供数据服务
3. 如果服务卫星解密这些状态失败，则意味（恶意的）UE可能篡改了此本地状态（例如，为了获得更高的QoS或免费计费）
4. 在这种情况下，卫星将通过远程地面站联系归属网络，回滚到图2所示的传统流程

通过这种方式，归属移动运营商保留（retains）了对其UE电信级服务的完全控制权。

我们正在扩展此基本流程，以支持更高级的电信级服务（例如本地计费、QoS和深度包检测(DPI)），并在不涉及远程地面站的情况下增强其可验证性（verifiability）。

我们需要指出，当前的设计侧重于无状态的回程和核心网。如图6所示， 全功能的RAN目前仍被放置在每颗卫星中。 与回程或核心网状态不同，RAN状态（如无线连接）大多是本地的，因此不会产生信令风暴和泄露。由于其处理的复杂性，实现完全无状态（fully stateless）的RAN是困难且不必要的（hard and unnecessary）。

话虽如此，对于资源受限的卫星而言，一个状态更少的轻量级RAN（lightweight RAN）仍然是可取的。利用UE侧的状态副本（如图9中用于无线连接建立的示例）来简化RAN是可能的。我们将卫星的轻量级RAN设计留作未来的工作。

Incremental Deployment

By its design, our usercentric cellular architecture aims for minimal 3GPP network standard changes and vendor/operator-side modifications to foster backward compatibility and incremental deployment. To realize our user-centric geographic backhaul routing in Figure 7, we can embed the user’s geographic location into the legacy IPv6 address and renovate its routing table computation logic. 2,7 To incrementally deploy the orbital cellular function refactoring and orchestration in Figures 8 and 9, the cellular operator and device vendor can add an external proxy for standard cellular functions in their satellites and UEs, as detailed in previous works.6,8,15 Besides incremental deployments in existing 5G, we note that the ongoing 6G standardization offers a precious opportunity to enable native support for these usercentric designs in future cellular networks.

从设计之初，我们的“以用户为中心”的蜂窝架构就旨在实现最小化的3GPP网络标准变更和厂商/运营商侧的修改，以促进后向兼容性（backward compatibility）和增量部署（incremental deployment）

• 为了实现图7中“以用户为中心”的地理回程路由，我们可以将用户的地理位置嵌入到传统的IPv6地址中，并革新其路由表计算逻辑
• 为了增量部署图8和图9中的在轨蜂窝功能重构和编排，蜂窝运营商和设备供应商可以在其卫星和UE的标准蜂窝功能之外，添加一个外部代理（external proxy），如先前的工作中所详述的

除了在现有5G中进行增量部署，我们注意到，正在进行的6G标准化工作，为在未来蜂窝网络中原生支持（native support）这些“以用户为中心”的设计，提供了一个宝贵的机遇。

Conclusion

This article studies the challenges of and opportunities for enabling network functions in 6G and beyond in satellite megaconstellations. We show that existing stateful session-based cellular architecture suffers from functionality, scalability, reliability, and security issues under extreme satellite infrastructure mobility in harsh outer space. To this end, we propose a paradigm shift from stateful session-based to stateless on-demand satellite cellular services in 6G and beyond. Our usercentric approach follows the end-to-end principle to simplify radio access and core network functions in satellites for the users and by the users. We hope our lessons can stimulate more research toward usercentric 6G and beyond in the space era.

本文探究了在卫星巨型星座中部署6G及未来网络功能所面临的挑战与机遇。

我们指出，现有的“基于状态化会话”（stateful session-based）的蜂窝架构，在严酷外太空（harsh outer space）的极端卫星基础设施移动性下，暴露出了功能性、可扩展性、可靠性及安全性方面的诸多问题

为此，我们提出了一种范式转变（paradigm shift）:

在6G及未来网络中，从“基于状态化会话”的服务转向“无状态、按需分配”（stateless on-demand）的天基蜂窝服务。

我们所提倡的“以用户为中心”（usercentric）的路径遵循“端到端原则”，旨在“为用户”（for the users）且“依靠用户”（by the users）来简化卫星上的无线接入网（RAN）和核心网功能

我们期望本文的经验能激发更多面向“太空时代”的、“以用户为中心”的6G及未来网络的研究