The Internet from Space, Reimagined: Leveraging Altitude for Efficient Global Coverage

The Internet from Space has recently attracted renewed attention following technological developments that enable massive constellations of small satellites in low Earth orbit (LEO). While LEO satellite networks (LSNs) promise low-latency global connectivity, they face several fundamental challenges including the need for constant satellite replacement due to orbital decay, affecting environmental sustainability, and increasing congestion in orbital space from emerging players, heightening collision risks.

In this work, we propose to expand the LSN constellation design space by including use of altitude as a flexible design parameter to help solve the aforementioned challenges—i.e., by constructing constellations with orbits throughout the range implied by the classic LEO, medium Earth orbit (MEO), and geosynchronous orbit (GEO) designators. Although altitudes above LEO induce higher propagation latency, they also increase how much of the Earth’s surface is visible to each satellite, thereby significantly reducing the total number of satellites required for global coverage. Building on this intuition, we provide an initial theoretical analysis of the tradeoffs enabled by orbital altitudes from LEO all the way up to GEO and conduct packet-level simulations demonstrating that MEO constellations can achieve present-day Internet latencies while using ∼19× fewer satellites and ∼14× fewer handovers than LEO.

得益于支持在低地球轨道（Low Earth Orbit, LEO）部署大规模小卫星星座的技术发展，“太空互联网”近期再次吸引了广泛关注。尽管LEO卫星网络（LEO Satellite Networks, LSNs）有望提供低延迟的全球连接服务，但它们也面临着若干根本性挑战，包括：因轨道衰减而需要持续更换卫星，这影响了环境的可持续性；以及新兴市场参与者导致轨道空间日益拥堵，加剧了碰撞风险。

在本研究中，我们提议将轨道高度作为一个灵活的设计参数，以扩展LSN的星座设计空间，从而应对上述挑战。具体而言，我们构建的星座其轨道范围涵盖了传统的低地球轨道（LEO）、中地球轨道（Medium Earth Orbit, MEO）和地球同步轨道（Geosynchronous Orbit, GEO）。虽然高于LEO的轨道会引入更高的传播延迟，但它也增加了单颗卫星对地球表面的覆盖范围，从而显著减少了实现全球覆盖所需的卫星总数。基于这一理念，我们对从LEO到GEO的轨道高度所带来的各项权衡进行了初步理论分析，并通过分组级仿真证明：MEO星座能够达到当今互联网的延迟水平，而其所需的卫星数量和切换次数比LEO星座分别减少了约19倍和14倍。

Introduction

In recent years, several efforts in industry and academia alike have reignited global interest in the concept of providing Internet connectivity from space [20, 21, 33, 36, 38]. This resurgence is largely driven by the rapid development and deployment of low Earth orbit (LEO) satellite constellations (LSNs), which offer the promise of high-speed, low-latency Internet access to even the most remote and under-served regions of the world. Companies like SpaceX [16], OneWeb [9], SSST [35], Telesat [19], and Amazon [11] have made significant investments in building and launching such networks, with the goal of closing the digital divide and enabling universal connectivity [39, 41, 52, 56].

However, the low orbital altitudes (e.g., typically between 500 and 2,000 km) that enable LSN’s promise of low-latency, high-bandwidth Internet access [33] also present a key design trade-off. In particular, the coverage area of each satellite is significantly smaller than that of higher-altitude systems and atmospheric drag limits each satellite’s orbital lifetime. As a result, global coverage and local bandwidth demands necessitate the deployment of “megaconstellations” of relatively cheap, disposable satellites. Given reduced per-satellite launch and manufacture costs as well as performance advantages of lower altitudes (e.g., satisfying the <∼10 ms latency requirements for present-day direct-to-cell 5G service which, in turn, requires <∼1000 km altitude [44, 46]), such low-altitude megaconstellations make strong economic sense. However, a growing body of research raises questions about their scalability, sustainability, and long-term costs [5, 12, 30, 49, 53].

Although several recent efforts do tackle the issues of reducing the number of satellites required for high-performance satellite networking [24, 25, 40, 57], they miss the fundamental trade-offs enabled by orbital altitude. On the one hand, efforts that seek optimal points in the LSN design space [24, 40] are limited by the inherently large number of cheap, low-capacity satellites required to meet coverage and bandwidth demands (e.g., ∼1k satellites to achieve continuous global coverage with minimal bandwidth). On the other hand, efforts that integrate higher orbital altitudes [25, 57] miss the nuanced interaction between altitude, coverage area, and propagation latency, focusing instead on routing and integration with the edge-cloud continuum.

To expand the design space, we envision a future where altitude is treated not as a fixed parameter, but as a critical and flexible component in constellation design [25, 26, 29]. Rather than narrow focus on LEO, medium Earth orbit (MEO), or GEO regimes, we envision satellite networks that fluidly blend a wide range of orbital altitudes while carefully balancing latency and bandwidth requirements with the effects of orbital mechanics, coverage radius, propagation latency, and environmental factors such as near-Earth radiation exposure [31, 42, 47]. As an initial step towards this vision, we present (to the best of our knowledge) the first analysis of the networking opportunities and challenges of deploying constellations at intermediate altitudes—i.e., between traditional LEO, MEO, and GEO. We focus in particular on achieving global coverage because (i) the complex time-varying ground-tracks of LEO and MEO satellites make coverage of fixed geographic regions nearly impossible and (ii) given a constellation approach that satisfies global coverage, additional bandwidth demand can be addressed (up to a limit) by adding additional orbital planes with overlapping coverage or (in the case of longer-lived, higher-altitude orbits) by increasing bandwidth capacity of each satellite (e.g., by adding additional spot beams).

Our key insight is that despite their distinct challenges, altitudes just above traditional LEO open up the possibility to significantly reduce the total number of satellites required for global coverage while preserving reasonable network latency. Realizing this possibility demands inter-disciplinary theoretical and empirical exploration of the trade-offs involved. This work, in particular explores the networking implications of exploiting altitude as a powerful design variable rather than as a fixed constraint through the following three contributions. First, we present a theoretical model that quantifies the minimum number of satellites required to achieve global coverage as a function of altitude. Second, we approximate bestcase networking performance metrics such as latency and coverage redundancy across a range of altitudes. Third, we empirically evaluate the impact of altitude on network behavior through realistic, simulation-based scenarios, providing insights that are grounded in operational parameters and practical constraints.

All scripts and NS3-based [8, 13] simulations are available at https://github.com/chris-misa/leveraging-altitude.

近年来，工业界与学术界的诸多努力重新点燃了全球对于“太空互联网”概念的兴趣 [20, 21, 33, 36, 38]。这一复兴主要得益于低地球轨道（LEO）卫星星座（LSNs）的快速发展与部署，它们有望为全球最偏远和服务欠缺的地区提供高速、低延迟的互联网接入。诸如 SpaceX [16]、OneWeb [9]、SSST [35]、Telesat [19] 及亚马逊 [11] 等公司已投入巨资建设并发射此类网络，旨在弥合数字鸿沟，实现全球普惠连接 [39, 41, 52, 56]。

然而，LEO卫星网络赖以实现其低延迟、高带宽承诺的低轨道高度（通常在500至2,000公里之间）[33]，也带来了一个关键的设计权衡。具体而言，单颗卫星的覆盖面积远小于高轨道系统，且大气阻力限制了其轨道寿命。因此，要满足全球覆盖和局部带宽需求，就必须部署由相对廉价、可抛弃式卫星组成的“巨型星座”。考虑到单颗卫星发射和制造成本的降低，以及低轨道带来的性能优势（例如，满足当前直连蜂窝5G服务所需的<∼10毫秒延迟，这要求轨道高度<∼1000公里 [44, 46]），这类低轨道巨型星座具有很强的经济合理性。但是，越来越多的研究对其可扩展性、可持续性及长期成本提出了质疑 [5, 12, 30, 49, 53]。

尽管近期的一些研究致力于解决如何减少高性能卫星网络所需卫星数量的问题 [24, 25, 40, 57]，但它们忽略了轨道高度所带来的根本性权衡。一方面，那些试图在LEO卫星网络设计空间中寻找最优点的工作 [24, 40]，受限于为满足覆盖和带宽需求而固有需要的大量廉价、低容量卫星（例如，实现全球连续覆盖和最低带宽就需要约1000颗卫星）。另一方面，那些集成了更高轨道高度的研究 [25, 57]，则忽略了轨道高度、覆盖面积和传播延迟之间微妙的相互作用，转而关注路由问题以及与边缘-云连续体的集成。

为了扩展设计空间，我们构想了这样一个未来：轨道高度不再被视为一个固定参数，而是星座设计中一个关键且灵活的组成部分 [25, 26, 29]。

我们设想的卫星网络并非局限于传统的LEO、中地球轨道（MEO）或地球同步轨道（GEO）范畴，而是能够灵活地融合各种轨道高度，同时审慎地平衡延迟、带宽需求与轨道力学、覆盖半径、传播延迟以及近地辐射暴露等环境因素之间的关系 [31, 42, 47]。

作为实现这一愿景的第一步，我们（据我们所知）首次分析了在中间高度——即传统LEO、MEO和GEO之间——部署星座所带来的网络机遇与挑战。我们特别关注实现全球覆盖，原因在于：

(i) LEO和MEO卫星时变的复杂地面轨迹使得覆盖固定地理区域几乎不可能

(ii) 在满足全球覆盖的星座方案基础上，额外的带宽需求可以通过增加具有重叠覆盖的轨道平面（在一定限度内），或（对于寿命更长、轨道更高的卫星）通过提升单星带宽容量（如增加点波束）来解决

我们的核心见解在于，尽管面临独特的挑战，但略高于传统LEO的轨道高度为显著减少全球覆盖所需卫星总数提供了可能，同时仍能保持合理的网络延迟。实现这一可能性需要对其中涉及的权衡关系进行跨学科的理论与实证探索。具体而言，本研究通过以下三项贡献，探讨了将轨道高度作为强大的设计变量而非固定约束所带来的网络影响。

首先，我们提出了一个理论模型，量化了实现全球覆盖所需的最少卫星数量与轨道高度之间的函数关系

其次，我们估算了在一系列轨道高度下，延迟和覆盖冗余度等最佳情况下的网络性能指标

第三，我们通过基于现实场景的仿真，实证评估了轨道高度对网络行为的影响，提供了基于实际操作参数和现实约束的见解

所有脚本及基于NS3 [8, 13] 的仿真程序均可在 这里 获取。

Background & Motivation

2.1 Goals of Satellite Networking

The key promise underlying the recent resurgence of interest in satellite networking is the ability of LEO satellite networks (LSNs) to potentially provide low-latency high-bandwidth connectivity to arbitrary regions of the Earth’s surface. Previous generation satellite Internet efforts leveraged designs with small numbers of large satellites placed in high-altitude geosynchronous Earth orbits (GEO) that suffer from inherently high latency (e.g., ∼100 ms one-way delay due to speed-of-light limit) and are incapable of supporting the low-latency requirements of modern web applications and realtime communication protocols. Present-day satellite Internet efforts embraced by the research community leverage designs with large numbers of small satellites placed in low-altitude Earth orbits (LEO) that satisfy lower latency requirements and raise a wide variety of novel technical challenges rooted in their dynamism.

As a backhaul for cellular-type networks, such connectivity enables expanding a wide range of latency-sensitive applications (e.g., real-time communication, web browsing) to billions of “unconnected” users anywhere on Earth [20, 24]. Moreover, as a backup for terrestrial communication technologies, the global reach of LSNs can potentially improve resilience of critical Internet-based services—including emergency response coordination—in the face of multi-hazard risks (e.g., earthquakes, wildfires, etc.) [52]. LSNs have also been proposed for specialized ultra-low latency applications, such as high-frequency trading, due to their potential for faster-than-fiber long-distance communication [33].

近期卫星网络领域兴趣复苏的背后，其核心承诺在于低地球轨道（LEO）卫星网络（LSNs）有潜力为地球表面的任意区域提供低延迟、高带宽的连接服务。上一代卫星互联网采用的设计方案，是将少量大型卫星部署在高空地球同步轨道（GEO），这种方案存在固有的高延迟（例如，受光速限制，单向延迟约100毫秒），无法满足现代网络应用和实时通信协议的低延迟要求。当前被研究界所青睐的卫星互联网方案，则采用将大量小型卫星部署在低地球轨道（LEO）的设计，这既满足了更低的延迟需求，也因其动态性而带来了一系列全新的技术挑战。

作为蜂窝式网络的骨干回传，这种连接能力可将广泛的延迟敏感型应用（如实时通信、网页浏览）扩展至全球数十亿的“无网络连接”用户 [20, 24]。此外，作为地面通信技术的备份，LSNs的全球覆盖能力能够在面对多重灾害风险（如地震、野火等）时，提升关键互联网服务（包括应急响应协调）的韧性 [52]。LSNs还被提议用于专门的超低延迟应用，例如高频交易，因为其长距离通信速度有潜力超越光纤 [33]。

2.2 Limitations of Current Approaches

LSNs must be large. A fundamental challenge in LEO satellite networks (LSNs) is the limited coverage area of individual satellites. Due to their proximity to Earth, a single LEO satellite can only cover a relatively small footprint on the surface at any given time. This geographic coverage limitation necessitates deploying hundreds or thousands of satellites (i.e., mega-constellations) to ensure continuous global coverage.

Furthermore, the limited bandwidth of ground-to-satellite links (GSLs) introduces an additional constraint. In practice, a single satellite cannot satisfy the total demand of a given region, especially e.g., in densely populated areas. This leads to the design requirement that multiple satellites must simultaneously cover each surface cell to meet bandwidth needs as well as provide redundancy. As a result, modern LSNs often incorporate overlapping satellite footprints, driving up the number of satellites needed.

Recent efforts have proposed non-uniform LEO constellations based on repeat ground-track orbits to optimize regional performance. For example, the approach described by [24] suggests selectively augmenting bandwidth in targeted areas through careful orbital planning. While such strategies may provide localized performance benefits, they do not eliminate the fundamental requirement for a larger number of satellites to ensure global coverage. As such, the actual cost savings and satellite reductions from these approaches remain uncertain in real-world conditions.

LEO卫星网络（LSNs）的一个根本性挑战是单颗卫星的覆盖区域有限。由于贴近地球，单颗LEO卫星在任一时刻仅能覆盖地表一个相对较小的区域（即“覆盖区”）。 这种地理覆盖的局限性，使得必须部署成百上千颗卫星（即“巨型星座”）才能确保全球范围的连续覆盖。

此外，地-星链路（GSLs）的有限带宽也带来了额外的约束。 在实践中，单颗卫星无法满足一个给定区域的总需求，尤其是在人口稠密地区。这导致了新的设计要求，即必须有多颗卫星同时覆盖每个地面单元，以满足带宽需求并提供冗余。 因此，现代LSNs的设计通常包含重叠的卫星覆盖区，这进一步增加了所需卫星的数量。

近期的研究提出了基于重复地面轨迹轨道的非均匀LEO星座，以优化区域性能。例如，[24] 中描述的方法建议通过精细的轨道规划来选择性地增强目标区域的带宽。虽然这类策略可能带来局部性能优势，但它们并未消除为确保全球覆盖而需要大量卫星的基本要求。因此，在真实条件下，这些方法实际能节省的成本和减少的卫星数量仍然不确定。

Satellites require continued replacements. In addition to the aforementioned physical and technical constraints, mega-constellations face significant sustainability challenges. Satellites in LEO are subject to atmospheric drag, which gradually degrades their orbits over time and ultimately causes reentry. Unlike higherorbit satellites, LEO spacecraft have relatively short operational lifespans (e.g., typically on the order of 5 to 7 years) before they must be replaced [23, 45]. This impermanence introduces a steady-state requirement for LSNs: to maintain consistent network performance, operators must regularly launch new satellites to replace those that deorbit.

To quantify the impermanence, we analyze the replacement rate required to sustain a given satellite population using empirical lifetime estimates for real-world orbits. In particular, we use STELA [18] to propagate a sample of real-world StarLink orbits (altitude ∼550 km, inclination 53.0 degrees) obtained from Celestrak [1] to obtain a probability distribution (mean and 95-percent confidence interval based on Monte Carlo simulation) for the lifetime of these orbits. We then extrapolate these lifetime estimates to larger numbers of satellites and show in Figure 1a the annual number of replacement launches needed to maintain steady-state population (e.g., in StarLink) as a function of the population size. Concurrently, Figure 1b estimates the amount of aluminum oxide (AlO) deposited into the mesosphere by reentering satellites, based on models from [30].

At the steady-state population of 1584 satellites, which corresponds to the target in a Federal Communications Commission (FCC) filing for one StarLink shell, the network would require the disposal of a median of 202 satellites per year. This process would release an estimated 6 metric tons of AlO annually. For context, this single network layer alone would contribute approximately 35% of the total AlO deposited from all satellite re-entries worldwide in 2022. Given that StarLink’s full constellation includes multiple such layers with similar orbital parameters and “growing space race” across countries [4, 6, 7, 14], we believe the cumulative environmental impact will be substantial if not worse.

除了上述物理和技术限制，巨型星座还面临严峻的可持续性挑战。LEO卫星受大气阻力影响，其轨道会随时间逐渐衰减，并最终再入大气层。与高轨道卫星不同，LEO航天器的运行寿命相对较短（通常为5到7年），之后必须进行更换 [23, 45]。这种非永久性为LSNs带来了一个稳态需求：为维持稳定的网络性能，运营商必须定期发射新卫星以替换那些脱离轨道的卫星。

为了量化这种非永久性，我们利用真实世界轨道的经验寿命估算，分析了维持给定卫星数量所需的替换率。具体而言，我们使用STELA [18] 对从Celestrak [1] 获取的真实StarLink轨道样本（高度约550公里，倾角53.0度）进行轨道推演，从而获得这些轨道寿命的概率分布（基于蒙特卡洛模拟的均值和95%置信区间）。然后，我们将这些寿命估算外推至更大规模的卫星数量，并在图1a中展示了为维持稳态卫星数量（以StarLink为例），所需的年均替换发射次数与星座规模的函数关系。同时，图1b根据 [30] 的模型，估算了卫星再入大气层时在中层沉积的氧化铝（AlO）量。

当卫星数量达到1584颗的稳态水平时（这对应于StarLink某个轨道壳层在美国联邦通信委员会FCC备案中的目标数量），该网络每年需要处理的卫星数量中位数为2022颗。这一过程预计每年将释放6公吨的氧化铝。作为参考，仅这一个网络层所产生的氧化铝就约占2022年全球所有卫星再入大气层沉积总量的35%。考虑到StarLink的完整星座包含多个具有相似轨道参数的此类轨道层，以及各国间日益激烈的“太空竞赛”[4, 6, 7, 14]，我们认为累积的环境影响将是巨大的，甚至可能更糟。

Orbital space is getting overcrowded. As deployment of megaconstellations has accelerated (e.g., Starlink [16], Qianfan or “Thousand Sails” [35]), the resulting saturation of orbital slots introduces a number of critical operational risks [24, 34, 45, 49]. Primary among them is the increased probability of on-orbit collisions. To mitigate this risk, satellites must perform active collision avoidance maneuvers, which consume limited on-board fuel and thereby reduce their operational lifespan. These maneuvers also add complexity to inter- and intra-constellation management and require continuous coordination with other operators and regulatory bodies.

Beyond operational concerns, the growing presence of LEO satellites poses a serious threat to Earth-based astronomy. The reflective surfaces of satellites can interfere with optical observations, while their radio emissions can disrupt sensitive radio telescopes. Recent analyses highlight that unless mitigated, the scale of current and planned LEO deployments will significantly degrade the scientific value of both professional and amateur astronomical observations [12, 53]. This adds another dimension to the environmental and societal costs of scaling LSNs.

随着巨型星座的加速部署（例如Starlink [16]、中国的“千帆”计划 [35]），轨道槽位的饱和带来了诸多关键的运行风险 [24, 34, 45, 49]。其中最主要的是在轨碰撞概率的增加。为降低此风险，卫星必须执行主动的碰撞规避机动，这会消耗有限的星上燃料，从而缩短其运行寿命。这些机动也增加了星座间和星座内的管理复杂性，并需要与其他运营商及监管机构进行持续协调。

除了运行层面的担忧，日益增多的LEO卫星也对地基天文学构成了严重威胁。卫星的反光表面会干扰光学观测，而其无线电发射则可能干扰灵敏的射电望远镜。近期分析指出，除非采取缓解措施，当前及规划中的LEO部署规模将严重降低专业及业余天文观测的科学价值 [12, 53]。这为扩展LSNs的环境和社会成本增添了另一个维度。

The Potential of Higher-Than-LEO Altitudes

The drawbacks of LEO mega-constellations discussed in the previous section can all be traced back to one critical defining design decision: the choice to focus exclusively on low-altitude Earth orbits. We investigate the decision of constellation altitude by first providing informal background on the near-earth environment above LEO, then considering how altitude impacts the networking potential of a single satellite and how these impacts shape the properties of a global-coverage constellation.

上一节讨论的LEO巨型星座的弊端，都可以追溯到一个关键性的设计决策：完全专注于低地球轨道。我们通过首先对LEO以上近地环境提供非正式的背景介绍，然后分析轨道高度如何影响单颗卫星的网络潜力，以及这些影响如何塑造一个全球覆盖星座的特性，来深入探讨星座轨道高度这一决策。

3.1 Near-Earth Space Radiation

Low-latency satellite networking’s near exclusive focus on LEO is often informally justified by the need to avoid regions of near-Earth space known as the Van Allen radiation belts [31, 47]. Although these radiation belts are often described as fixed altitude ranges that must be avoided, their real-world structure and dynamics are far more complex. To illustrate, Figure 2 shows (electron) radiation intensity observed by the CIRBE spacecraft [42, 43] for two consecutive passes (∼9 hours apart) through the Van Allen belts (on April 21, 2023).

Although two distinct regions of higher intensity radiation are apparent, they fluctuate significantly between observations, hinting at the wide-ranging dynamics apparent in CIRBE data [3]. Rather than clean “bands”, near-Earth radiation is highly dynamic and at times (e.g., during solar storms) crosses into regions typically considered outside of the bands (e.g., the right-hand side of Figure 2 corresponds to GEO altitude; the South Atlantic Anomaly [15] persistently extends will into LEO altitudes below 500 km). For the design of satellite constellations for networking, this implies satellites will potentially need to deal with radiation no matter their orbital altitude and hence, the satellite networking community should not artificially limit the design space to narrow “safe” altitude bands.

低延迟卫星网络之所以几乎完全聚焦于LEO，一个常见的非正式理由是需要避开被称为范艾伦辐射带的近地空间区域 [31, 47]。尽管这些辐射带常被描述为必须避开的固定高度范围，但其真实世界的结构和动态远比这复杂。为了说明这一点，图2展示了CIRBE航天器 [42, 43] 在连续两次（相隔约9小时）穿过范艾伦辐射带（2023年4月21日）时观测到的（电子）辐射强度。

虽然图中可以明显看到两个高强度辐射区域，但它们在两次观测之间发生了显著波动，这揭示了CIRBE数据中广泛存在的动态变化 [3]。近地辐射并非清晰的“带状”，而是高度动态的，有时（例如在太阳风暴期间）会进入通常被认为在辐射带之外的区域（例如，图2右侧对应于GEO轨道高度；而南大西洋异常区 [15] 则持续延伸至500公里以下的LEO高度）。对于网络卫星星座的设计而言，这意味着无论轨道高度如何，卫星都可能需要应对辐射问题。因此，卫星网络研究界不应人为地将设计空间限制在狭窄的“安全”高度带内。

3.2 Single-Satellite Impacts of Altitude

Given a more nuanced view of near-Earth radiation at different altitudes, we now turn to characterizing altitude’s impact on networking, first, from the perspective of a single satellite, and later (§ 3.3) from the perspective of satellite constellation design.

在对不同高度的近地辐射有了更细致的理解后，我们现在转而描述轨道高度对网络性能的影响，首先从单颗卫星的视角，然后（在§ 3.3）从卫星星座设计的视角进行分析。

Impact on network latency. The most widely-acknowledged impact of altitude on satellite networking is on latency induced by propagation delay. A lower-bound on this latency can be estimated by dividing the distance of a potential communication link by the speed of light. We visualize this relationship in Figure 3 by considering a single direction of a single ground-to-satellite radio link and showing the theoretical minimum speed-of-light latency (y-axis) for different altitudes (x-axis). We show this latency for both the shortest path, when the satellite is directly overhead the ground station (i.e., at its zenith, solid blue), as well as a more realistic path, when the satellite is 30°above the ground station’s horizon (dotted red). On the right of the figure, the relatively high altitudes required for GEO (∼36,000 km) directly induce the unacceptably high network latency (>100 ms one-way) GEO has become associated with. On the left of the figure, the relatively low altitudes associated with LEO (e.g., ∼600 km) directly enable low-latency networking (e.g., ∼2 ms one-way).

Between the extremes of LEO and GEO, Figure 3 illustrates a wide range of intermediate altitude and latency combinations. To contextualize, we draw horizontal bars (in green) indicating the approximate one-way latency ranges associated with 4G and 5G terrestrial radio links. We also draw vertical bars (in red) indicating the approximate average locations of the Van Allen radiation belts which increase operational costs by requiring addition shielding of sensitive radio components. Combining these ranges yields the insight that constellations at altitudes above the classic LEO range but below the inner Van Allen belt can still achieve lower one-way latency compared with 5G (e.g., <10 ms) and a limited range of altitudes above the inner Van Allen belt can achieve lower one-way latency compared with 4G (e.g., <40 ms).

对网络延迟的影响

轨道高度对卫星网络最为公认的影响是传播延迟。此延迟的下限可以通过将潜在通信链路的距离除以光速来估算。我们在图3中将这种关系可视化，考虑了单个地-星无线电链路的单向传输，并展示了不同轨道高度（x轴）下理论上的最小光速延迟（y轴）。我们分别展示了两种路径下的延迟：

• 最短路径，即卫星位于地面站正上方（即天顶，蓝色实线）
• 更现实的路径，即卫星位于地面站地平线上方30°（红色虚线）

在图的右侧，GEO所需的高轨道（约36,000公里）直接导致了其广为人知的、不可接受的高网络延迟（单向>100毫秒）。而在图的左侧，与LEO相关的低轨道（约600公里）则直接促成了低延迟网络（单向约2毫秒）。

在LEO和GEO这两个极端之间，图3展示了广泛的中间高度与延迟的组合。为了提供参考，我们用水平绿条标示了4G和5G地面无线链路的大致单向延迟范围。我们还用垂直红条标示了范艾伦辐射带的大致平均位置，这些辐射带因需要对敏感的无线电组件增加额外屏蔽而增加了运营成本。结合这些范围可以得出一个见解：位于传统LEO范围之上、但在内范艾伦带之下的星座，仍能实现比5G更低的单向延迟（例如<10毫秒）；而在内范艾伦带之上的有限高度范围内，也能实现比4G更低的单向延迟（例如<40毫秒）。

Impact on Field-of-View. Beyond propagation latency, the next most important consideration in the design of satellite networks is the field-of-view (FoV), or area of the Earth’s surface, that a single satellite can cover. Larger FoV implies a single satellite can cover more of the Earth’s surface and hence fewer satellites are required for global coverage whereas smaller FoV implies the opposite.

Figure 4 shows FoV measured as the radius between the satellite’s nadir and the circle defined by minimum observation angle (i.e., elevation of the satellite above the ground station’s horizon) for several minimum observation angles. Due to curvature of the Earth, lower-altitude satellites have smaller FoV whereas higher-altitude satellites have larger FoV. In particular (assuming a minimum observation angle of 30 degrees), the FoV of a LEO satellite at an altitude of ∼600 km is ∼800 km whereas the FoV of a GEO satellite is nearly 6000 km. (For comparison, an orthographic vantage point (i.e., at infinite distance from the Earth) would cover half the Earth’s circumference or ∼20,000 km.)

Figure 4 indicates that FoV increases significantly with constellation altitude. In particular, satellites above the inner Van Allen belt (again shown as vertical red regions) have over 2× higher FoV compared to typical LEO satellites below the belt. (Note that because we measure FoV as radius, this implies over 4× increase in the surface area covered.)

对视场（Field-of-View）的影响

除了传播延迟，卫星网络设计中次重要的考虑因素是视场（Field-of-View, FoV），即单颗卫星能够覆盖的地球表面区域。更大的FoV意味着单颗卫星能覆盖更广的地球表面，因此实现全球覆盖所需的卫星更少；反之亦然。

图4展示了在几种最小观测角度（即卫星在地面站地平线上的仰角）下，以卫星星下点到最小观测角定义的圆形区域的半径来衡量的FoV。由于地球的曲率，低轨道卫星的FoV较小，而高轨道卫星的FoV较大。

具体来说（假设最小观测角为30度），一个轨道高度约600公里的LEO卫星，其FoV半径约为800公里；而一颗GEO卫星的FoV半径则接近6000公里。（作为对比，一个正交视角（即距离地球无限远）将覆盖半个地球周长，约20,000公里）

图4表明，FoV随星座高度显著增加。特别地，位于内范艾伦带（同样以垂直红区表示）上方的卫星，其FoV是带下方典型LEO卫星的2倍以上。（注意，由于我们用半径衡量FoV，这意味着其覆盖的表面积增加了4倍以上。）

Impact on orbit lifetime. Finally, because LEO satellites operate in a near-Earth domain, their orbits gradually experience non-trivial decay due to atmospheric drag. The effects of atmospheric drag decrease with altitude so that higher-altitude orbits are inherently longer lived. In particular, orbits below ∼1000 km may only last for 𝑂(10) years where as higher orbits (above the influence of the atmosphere) last for 𝑂(100−1000) years and are typically considered more-or-less permanent [23, 45].

对轨道寿命的影响

最后， 由于LEO卫星在近地领域运行，它们的轨道会因大气阻力而逐渐发生不可忽略的衰减。 大气阻力的影响随高度增加而减小，因此更高轨道的卫星寿命天生更长。具体来说，低于约1000公里的轨道可能只能维持𝑂(10)年，而更高（不受大气影响）的轨道寿命可达𝑂(100−1000)年，通常被认为是或多或少永久的 [23, 45]。

3.3 Constellation-Wide Impacts of Altitude

We now integrate the above impacts of altitude on a single satellite’s networking potential into the corresponding network-wide properties assuming a network that provides uniform global coverage (for latitudes up to its inclination).

现在，我们把上述轨道高度对单颗卫星网络潜力的影响，整合到相应的全网属性中，假设网络提供均匀的全球覆盖（纬度最高至其轨道倾角）。

Number of satellites required for global coverage. We first consider how designing higher altitude constellations can reduce the number of satellites required to provide global coverage. Assuming the common Walker-delta constellation geometry [27, 54] used by most recent LEO efforts and a fixed inclination of 65 degrees, we select orbital planes and satellites per plane in order to ensure complete coverage with non-zero overlap between satellites for handoffs. In particular, ascending nodes of orbital plans are separated by 2𝑟 cos(𝜋/4) and satellites in each plane are separated by 2𝑟 cos(𝜋/4)/sin(𝑖) where 𝑟 is the FoV and 𝑖 is the inclination.

Figure 5 shows the total number of satellites required by this approach (y-axis) as a function of altitude (x-axis) for several different minimum observation angles. We observe nearly exponential decrease in the number of satellites required as altitude increases (note the log axes). For example, a constellation at ∼600 km requires ∼1k satellites whereas a constellation at ∼6000 km requires only ∼60 satellites (assuming minimum observation angle of 30 degrees). This implies LEO altitudes require relatively large numbers of satellites to achieve minimal global coverage whereas constellations at higher altitudes achieve global coverage with far fewer satellites.

全球覆盖所需的卫星数量

我们首先考虑设计更高轨道星座如何能减少提供全球覆盖所需的卫星数量。假设采用近期多数LEO项目所使用的通用沃克-δ（Walker-delta）星座几何构型 [27, 54]，并固定轨道倾角为65度，我们选择轨道平面和每平面卫星数，以确保卫星间有非零重叠以支持切换，从而实现完全覆盖。具体来说，轨道平面的升交点间隔为 2𝑟 cos(π/4)，每个平面内的卫星间隔为 2𝑟 cos(π/4)/sin(𝑖)，其中𝑟是FoV，𝑖是倾角。

图5展示了采用此方法所需的卫星总数（y轴）与轨道高度（x轴）的函数关系，并给出了几种不同最小观测角下的情况。我们观察到， 随着轨道高度的增加，所需卫星数量呈近指数级下降 （注意坐标轴为对数尺度）。例如，一个约600公里的星座需要约1000颗卫星，而一个约6000公里的星座仅需约60颗卫星（假设最小观测角为30度）。这意味着LEO高度需要相对大量的卫星才能实现最低限度的全球覆盖，而更高高度的星座则能用少得多的卫星实现全球覆盖。

Approximate end-to-end latency. Next, we consider the added latency overhead associated with higher altitude constellations. In particular, we assume per-hop overheads to be negligible and that the signal travels an arch along the orbital path directly between satellites. This approach approximates a low-bound on the latency achievable by a real-world constellation by replacing the jagged ISL paths required in topologies like +Grid [21, 36] or xGrid [48] with a smooth arch. After summing the total distance traveled in this way from source ground station, to the orbital altitude, and back to destination ground station, we again divide by the speed of light to estimate a theoretical lower-bound on the latency through a potential satellite network.

Figure 6 shows the approximated end-to-end latency for several different example terrestrial distances along the Earth surface. Endto-end latency remains below 100 ms for up to 10,000 km paths for constellations at altitudes of up to 6000 km indicating the potential for constellations above LEO to achieve usefully-low latency (e.g., less than 100 ms). We also note that latency in such above-LEO constellations is significantly lower for shorter distances along the Earth surface indicating their potential use in applications like providing fiber-link backup (e.g., over hundreds rather than thousands of km) in multi-hazard scenarios [52].

近似端到端延迟

接下来，我们考虑更高轨道星座带来的额外延迟开销。具体来说，我们假设每跳开销可忽略不计，且信号直接沿卫星间的轨道路径弧线传播。这种方法通过用平滑的弧线替代+Grid [21, 36] 或 xGrid [48] 等拓扑中所需的曲折星间链路（ISL）路径，估算了真实星座可实现的延迟下限。在计算了信号从源地面站到轨道高度，再返回到目的地面站的总传播距离后，我们再次除以光速，以估算通过潜在卫星网络的理论延迟下限。

图6展示了在几种不同的地球表面距离下，近似的端到端延迟。对于高达6000公里的星座，即使是10,000公里的路径，端到端延迟仍低于100毫秒，这表明高于LEO的星座有潜力实现有用的低延迟（例如，小于100毫秒）。我们还注意到，在这类高于LEO的星座中，较短的地面距离其延迟显著更低，这表明它们在多灾害场景下，有潜力用作光纤链路备份（例如，覆盖数百而非数千公里）[52]。