跳转至

Orbital Edge Computing: Nanosatellite Constellations as a New Class of Computer System

Advances in nanosatellite technology and a declining cost of access to space have fostered an emergence of large constellations of sensor-equipped satellites in low-Earth orbit. Many of these satellite systems operate under a “bent-pipe” architecture, in which ground stations send commands to orbit and satellites reply with raw data. In this work, we observe that a bent-pipe architecture for Earth-observing satellites breaks down as constellation population increases. Communication is limited by the physical configuration and constraints of the system over time, such as ground station location, nanosatellite antenna size, and energy harvested on orbit. We show quantitatively that nanosatellite constellation capabilities are determined by physical system constraints.

We propose an Orbital Edge Computing (OEC) architecture to address the limitations of a bent-pipe architecture. OEC supports edge computing at each camera-equipped nanosatellite so that sensed data may be processed locally when downlinking is not possible. In order to address edge processing latencies, OEC systems organize satellite constellations into computational pipelines. These pipelines parallelize both data collection and data processing based on geographic location and without the need for cross-link coordination. OEC satellites explicitly model constraints of the physical environment via a runtime service. This service uses orbit parameters, physical models, and ground station positions to trigger data collection, predict energy availability, and prepare for communication. We show that an OEC architecture can reduce ground infrastructure over 24× compared to a bent-pipe architecture, and we show that pipelines can reduce system edge processing latency over 617×.

纳米卫星技术的进步以及进入空间成本的降低，促进了低地球轨道（low-Earth orbit）大规模传感器卫星星座的兴起。其中许多卫星系统在一种“弯管”（bent-pipe）架构下运行，即由地面站向在轨卫星发送指令，卫星则回传原始数据。

在本研究中，我们观察到，对于地球观测卫星而言，随着星座规模的增加，“弯管”架构会逐渐失效。通信能力会受制于系统随时间变化的物理配置和约束，例如地面站位置、纳米卫星天线尺寸以及在轨收集的能量。我们定量地证明了纳米卫星星座的能力是由其物理系统约束所决定的。

为解决“弯管”架构的局限性，我们提出了一种轨道边缘计算（Orbital Edge Computing, OEC）架构。OEC架构支持在每颗配备相机的纳米卫星上进行边缘计算，从而在下行链路不可用时能够对感知数据进行本地处理。为了解决边缘处理的延迟问题，OEC系统将卫星星座组织成计算流水线（computational pipelines）。这些流水线基于地理位置并行化数据收集与数据处理，且无需跨链协调。OEC卫星通过一个运行时服务，对物理环境的约束进行显式建模。该服务利用轨道参数、物理模型和地面站位置来触发数据收集、预测能量可用性，并为通信做好准备。

我们的研究表明，与“弯管”架构相比，OEC架构可将地面基础设施减少超过24倍，而计算流水线可将系统边缘处理延迟降低超过617倍

Introduction

A resurgence in the space industry [13, 23, 67, 98], concurrent with standardization of nanosatellite form factors [69] and a declining cost of access to space [31], has stimulated an exponential growth in nanosatellite launches over the past two decades [85]. The largest commercial satellite constellations today consist of hundreds of Earth-observing, cameraequipped nanosatellites [12, 59], each measuring centimeters, massing a few kilograms, and costing only thousands of USD. These emerging systems are a stark contrast to the extremely expensive, monolithic space vehicles of the past. For example, the $192,000,000, 500 kg Earth Observing-1 (EO-1) is a “oneof-a-kind” [70] satellite operating for 16 years under a NASA ground control team. This single-satellite remote sensing mission depended on NASA’s extensive ground infrastructure to support its communication model: operators manually schedule communication on ground stations shared among all other space missions. The EO-1 mission terminated when its ground support was de-funded, leaving no one to schedule communication and data management operations. As nanosatellites proliferate, the viability of building and operating a manual, bent-pipe system architecture diminishes. The scale of this challenge is increasing; several commercial ventures have announced plans to deploy satellite constellations consisting of thousands of devices over the next ten years [34, 40–42, 76, 77, 86–88].

在航天产业复兴 [13, 23, 67, 98]、纳米卫星规格标准化 [69] 以及进入空间成本下降 [31] 的共同推动下，过去二十年间纳米卫星的发射数量呈指数级增长 [85]。当今最大的商业卫星星座由数百颗搭载相机的地球观测纳米卫星组成 [12, 59]，每颗卫星尺寸仅为数厘米，质量几公斤，成本仅数千美元。这些新兴系统与过去极其昂贵的单体式航天器形成了鲜明对比。例如，耗资1.92亿美元、重达500公斤的“地球观测-1号”（EO-1）卫星是一个“独一无二”的 [70] 特例，由一个NASA地面控制团队操作运行了16年。这一单星遥感任务依赖于NASA庞大的地面基础设施来支持其通信模式：操作员在所有航天任务共享的地面站上手动调度通信。当其地面支持资金被取消后，EO-1任务也随之中止，因为无人来安排通信和数据管理操作。随着纳米卫星数量的激增，构建和运营一个手动的“弯管”（bent-pipe）系统架构的可行性正在降低。这一挑战的规模还在不断扩大；数家商业公司已宣布计划在未来十年内部署由数千个设备组成的卫星星座 [34, 40–42, 76, 77, 86–88]。

A trend toward massive constellations of low Earth orbit (LEO) nanosatellites demands a new architecture for space systems. As with large, expensive space vehicles of the past, nanosatellite constellations today still rely on a communication model that sends remote control commands to orbit and delivers sensed data to Earth [20]; this design is referred to as a “bent pipe” by space system architects [58]. Momentum towards large constellations of nanosatellites requires a reimagining of space systems as distributed, edge-sensing and edge-computing systems. As work on warehouse scale computing [6] did for datacenters-as-computers, we aim to raise awareness of system-level research questions for orbital edge computer systems equipped with high-datarate cameras and sensors. This work characterizes and addresses computer hardware and software design challenges for orbital edge computer systems, many of which stem from physical deployment constraints and limitations inherent to ground infrastructure.

Addressing the challenges of the orbital edge is a timely and important problem due to the recent proliferation of nanosatellite systems. Space system architects are eschewing large, costly (e.g. $650,000,000 [28]), “exquisite” [89] satellites for constellations of small, inexpensive (e.g. $65,000 ea.) “CubeSats” [69]. Commercial efforts [12, 59] use this 10,000× lower per-device cost to deploy large, sensor-equipped nanosatellite constellations to LEO and observe the planet with high temporal resolution. Such systems create new capabilities for precision agriculture, environmental and infrastructure monitoring, humanitarian assistance and disaster relief, security, climate, and other commercial uses.

Challenges faced by existing systems under a bent-pipe architecture stem from fundamental physical constraints. The time-varying relationship between the geographic location of ground stations and the orbital position of nanosatellites imposes limitations on link availability and can lead to high downlink latencies. Intermittently available downlinks incur high latency between data collection and data processing in existing systems that simply downlink raw observations. Downlinks can be unreliable; one nanosatellite mission reports 88% packet loss [72]. Commercial ventures require complex, custom downlink solutions [20]. Shared “last mile” infrastructures [2, 99] aid availability but do not address the terrestrial centralization bottleneck. Limits on downlink bitrate prevent bent pipes from scaling to accommodate the extreme data volumes of large constellations and create a need for a new system architecture less reliant on communication.

大规模低地球轨道（LEO）纳米卫星星座的发展趋势，要求一种全新的空间系统架构。与过去庞大昂贵的航天器一样， 今天的纳米卫星星座仍然依赖一种将遥控指令发送至轨道、并将感知数据传回地球的通信模式 [20]；这种设计被空间系统架构师称为“弯管” [58]。大规模纳米卫星星座的发展势头要求我们将空间系统重新构想为分布式的、具备边缘感知和边缘计算能力的系统。正如仓库级计算（warehouse scale computing）[6] 之于“作为计算机的数据中心”所做的工作一样，我们旨在提升学界对配备高数据率相机和传感器的轨道边缘计算系统（orbital edge computer systems）在系统层面研究问题的关注。本研究旨在刻画并解决轨道边缘计算系统的计算机软硬件设计挑战，其中许多挑战源于物理部署的约束以及地面基础设施固有的局限性。

由于近年来纳米卫星系统的激增，解决轨道边缘的挑战已成为一个既及时又重要的问题。空间系统架构师们正在摒弃昂贵（例如6.5亿美元 [28]）的“精致” [89] 卫星，转而采用由廉价（例如每颗6.5万美元）的“立方体卫星”（CubeSats）[69] 组成的星座。商业公司 [12, 59] 利用这种降低了10,000倍的单位设备成本，在LEO部署大型的、配备传感器的纳米卫星星座，以高时间分辨率观测地球。此类系统为精准农业、环境与基础设施监测、人道主义援助与灾难救援、安防、气候以及其他商业应用创造了新的能力。

现有系统在“弯管”架构下面临的挑战源于基本的物理约束。地面站的地理位置与纳米卫星的轨道位置之间的时变关系限制了链路的可用性，并可能导致高昂的下行链路延迟。间歇性可用的下行链路，在现有仅下传原始观测数据的系统中，造成了从数据收集到数据处理之间的高延迟。下行链路可能并不可靠；某纳米卫星任务报告了88%的丢包率 [72]。商业公司需要复杂、定制化的下行链路解决方案 [20]。共享的“最后一英里”基础设施 [2, 99] 有助于提高可用性，但并未解决地面集中化的瓶颈问题。下行链路比特率的限制阻碍了“弯管”架构的扩展，使其无法适应大型星座产生的海量数据，因此亟需一种对通信依赖性较低的新系统架构。

On Earth, sensor systems increasingly leverage edge computing by performing sensor-local data processing in lieu of communication to a cloud datacenter [83]. While access to the cloud from the edge can accelerate computing [8], any benefits depend on backhaul network availability. Highdatarate sensors deployed across large geographic environments face a network bottleneck from the sensor to the datacenter as datarate exceeds bandwidth [44, 83]. Processing data at the edge avoids high-bitrate infrastructure at each sensor and supports a larger population of deployed devices. Edge processing avoids the privacy and security risks of multi-tenant infrastructures in shared datacenters [26, 56, 60].

Applying these terrestrial edge computing techniques directly to space systems is appealing, but nanosatellite constellations are subject to a unique set of operating constraints that typically do not affect terrestrial edge systems. In space, unlike on Earth, all energy for computation and communication must be harvested from the environment — plugging into a power grid is not an option. The small size of a nanosatellite, which is dictated by the cubesat standard, limits solar panel surface area and thus limits power. Unlike on Earth, the quality of visual data in space is fundamentally limited not only by onboard sensors, but also by chassis size (which limits focal length) and orbit altitude (which limits optical resolution). Communication bitrate, which is affected by orbit parameters, ground station capability, and ground station location, dictates the amount of data satellites buffer between downlinks. Any viable orbital edge computer system must directly address these unique physical constraints.

在地球上，传感器系统越来越多地利用边缘计算，通过在传感器本地进行数据处理来替代与云数据中心的通信 [83]。虽然从边缘访问云端可以加速计算 [8]，但任何收益都取决于回程网络的可用性。部署在广阔地理环境中的高数据率传感器，当其数据率超过带宽时，会面临从传感器到数据中心之间的网络瓶颈 [44, 83]。在边缘处理数据避免了在每个传感器处部署高比特率基础设施，并支持更大规模的设备部署。边缘处理还避免了共享数据中心中多租户基础设施带来的隐私和安全风险 [26, 56, 60]。

将这些地面边缘计算技术直接应用于空间系统虽然很有吸引力，但纳米卫星星座受到一系列独特的运行约束，这些约束通常不影响地面边缘系统。在太空中，与地球不同，所有用于计算和通信的能量都必须从环境中采集——接入电网是不可能的。由立方体卫星标准规定的纳米卫星的小尺寸，限制了太阳能电池板的表面积，从而限制了功率。与地球不同，太空中视觉数据的质量不仅根本上受限于机载传感器，还受限于机身尺寸（限制了焦距）和轨道高度（限制了光学分辨率）。通信比特率受轨道参数、地面站能力和地面站位置的影响，它决定了卫星在两次下行链路之间可以缓冲的数据量。任何可行的轨道边缘计算系统都必须直接解决这些独特的物理约束。

We propose Orbital Edge Computing (OEC) as an alternative to existing nanosatellite constellation bent-pipe architectures. OEC colocates processing hardware with high-datarate spectral sensors in small, low-cost satellites. We characterize the physically-constrained design space of a computational nanosatellite, revealing fundamental limitations on data quality and computation inherent to state-of-the-art designs. Based on this design space study, we introduce computational nanosatellite pipelines (CNPs) as an organizational principle for OEC constellations. A CNP distributes sensing, processing, and communication across a constellation in order to remain within latency and energy envelopes.

We then develop cote 1 , the first orbital edge computing simulator (cote-sim) and runtime service (cote-lib). cote physically models orbital mechanics and Earth rotation to track ground station and satellite positions over time. cote models data collection along each satellite ground track, as well as the energy and latency of sensing, computing, and communication for an entire constellation. cote is useful for mission design and simulation (cote-sim) and as an online runtime service for each nanosatellite and ground station (cote-lib).

We use cote-sim to quantitatively demonstrate the limitations of bent pipes, the advantages of OEC, and the benefits of nanosatellite pipelines. cote-lib runs on each device and provides continuous access to a physics-based model of the constellation and ground infrastructure in order to enable autonomy. By directly modeling the physics of the system, each satellite determines at runtime when to downlink, when to process locally, and how to distribute responsibilities across a pipeline without the need for online coordination or crosslink communication. cote-lib enables OEC by eliminating the reliance on remote control through a bent pipe.

我们提出 轨道边缘计算（Orbital Edge Computing, OEC）作为现有纳米卫星星座“弯管”架构的替代方案 。OEC将处理硬件与高数据率光谱传感器共同部署在小型、低成本的卫星中。我们刻画了计算型纳米卫星在物理约束下的设计空间，揭示了当前最先进设计中固有的数据质量和计算能力的根本限制。基于这一设计空间研究，我们引入 计算型纳米卫星流水线（computational nanosatellite pipelines, CNPs）作为OEC星座的一种组织原则。CNP将感知、处理和通信任务分布在整个星座中 ，以确保系统运行在延迟和能量包络之内。

接着，我们开发了cote¹，这是首个轨道边缘计算模拟器（cote-sim）和运行时服务（cote-lib）。cote通过对轨道力学和地球自转进行物理建模，来追踪地面站和卫星随时间变化的位置。cote模拟了每颗卫星沿其星下点轨迹的数据收集过程，以及整个星座在感知、计算和通信方面的能量消耗与延迟。cote既可用于任务设计与仿真（cote-sim），也可作为每个纳米卫星和地面站的在线运行时服务（cote-lib）。

我们使用cote-sim来定量地展示“弯管”架构的局限性、OEC的优势以及纳米卫星流水线的好处。cote-lib运行在每个设备上，提供对星座和地面基础设施的物理模型的持续访问，从而实现自主性。通过直接对系统物理特性进行建模，每颗卫星都可以在运行时决定何时下行数据、何时进行本地处理，以及如何在流水线中分配任务，而无需在线协调或跨链通信。cote-lib通过消除对“弯管”架构下远程控制的依赖，实现了OEC。

In summary, this paper makes the following contributions:

• We demonstrate the limitations of bent pipes using a novel, physics-based simulator that includes orbital dynamics, communication, energy harvesting, and data collection; an OEC architecture can reduce ground station infrastructure over 24× compared to a bent-pipe architecture.

• We characterize the physical design space of an OEC device and identify key limitations that drive constellation design.

• We propose and evaluate computational nanosatellite pipelines, an organizational principle for OEC constellations that distributes work across devices; CNPs can reduce system edge processing latency over 617×.

• We present a runtime service deployed to each nanosatellite and ground station that models the constellation, ground infrastructure, and energy environment in order to autonomously schedule sensing, communication, and computing without the need for cross-link coordination.

总结而言，本文做出以下贡献：

  • 我们使用一个新颖的、基于物理的模拟器（包含轨道动力学、通信、能量采集和数据收集模型）证明了“弯管”架构的局限性；与“弯管”架构相比，OEC架构可将地面站基础设施减少超过24倍
  • 我们刻画了OEC设备的物理设计空间，并识别了驱动星座设计的关键限制因素
  • 我们提出并评估了计算型纳米卫星流水线（CNPs），这是一种将工作分布于不同设备的OEC星座组织原则；CNPs可将系统边缘处理延迟降低超过617倍
  • 我们提出了一种部署在每个纳米卫星和地面站上的运行时服务，该服务通过对星座、地面基础设施和能源环境进行建模，来自主地调度感知、通信和计算任务，且无需跨链协调