On the Automation Optimization and In-Orbit Validation of Intelligent Satellite Constellation Operations

2025-05-02 0 0 7.46MB 18 页 10玖币

侵权投诉

On the Automation, Optimization, and In-Orbit Validation of Intelligent

Satellite Constellation Operations

Gregory Stock

Saarland University

Saarland Informatics Campus E13, 66123 Saarbrücken, Germany; +49 681 302 5635

stock@depend.uni-saarland.de

Juan A. Fraire

Saarland University / INRIA

Saarland Informatics Campus E13, Saarbrücken, Germany; +54 9351 244 6010

INSA Lyon, 20 Avenue Albert Einstein, 69621 Villeurbanne CEDEX, France

juanfraire@depend.uni-saarland.de

Holger Hermanns

Saarland University / Institute of Intelligent Software

Saarland Informatics Campus E13, 66123 Saarbrücken, Germany; +49 681 302 5631

Institute of Intelligent Software, Jinzhu Plaza, No. 221 West Huanshi Avenue, Nansha, Guangzhou, China

hermanns@depend.uni-saarland.de

Eduardo Cruz, Alastair Isaacs, Zhana Imbrosh

GomSpace

GomSpace A/S, Langagervej 6, 9220 Aalborg Øst, Denmark; +352 621 291 207

ecru@gomspace.com

ABSTRACT

Recent breakthroughs in technology have led to a thriving “new space” culture in low-Earth orbit (LEO) in

which performance and cost considerations dominate over resilience and reliability as mission goals. These

advances create a manifold of opportunities for new research and business models but come with a number

of striking new challenges. In particular, the size and weight limitations of low-Earth orbit small satellites

make their successful operation rest on a ﬁne balance between solar power infeed and the power demands

of the mission payload and supporting platform technologies, buﬀered by on-board battery storage. At the

same time, these satellites are being rolled out as part of ever-larger constellations and mega-constellations.

Altogether, this induces a number of challenging computational problems related to the recurring need to

make decisions about which task each satellite is to eﬀectuate next. Against this background, GOMSPACE

and Saarland University have joined forces to develop highly sophisticated software-based automated solutions

rooted in optimal algorithmic and self-improving learning techniques, all this validated in modern nanosatellite

networked missions operating in orbit.

The paper introduces the GOMSPACE Hands-Oﬀ Operations Platform (HOOP), an automated, ﬂexible, and

scalable end-to-end satellite operation framework for commanding and monitoring subsystems, single-satellites,

or constellation-class missions. To this, the POWVER initiative at Saarland University has contributed state-of-

the-art dynamic programming and learning techniques based on profound battery and electric power budget

models. These models are continually kept accurate by extrapolating data from telemetry received from

satellites. The resulting machine learning approach delivers optimal, eﬃcient, scalable, usable, and robust

ﬂight plans, which are provisioned to the satellites with zero need for human intervention—but which are still

under the full control of the mission operator. We report on insights gained while validating the integrated

POWVER-HOOP approach in orbit on the dual-satellite mission GOMX–4 by GOMSPACE that is currently in

orbit.

This is an author-generated technical report of a paper published in the Small Satellite Conference 2021.1

arXiv:2210.11171v1 [cs.NI] 20 Oct 2022

INTRODUCTION

Near-Earth satellites are being launched by the thou-

sands; an unprecedented pace made possible by recent

breakthroughs in technology accompanying a “new

space” culture where cost/performance considera-

tions dominate over resilience/reliability (i.e., emer-

gence of COTS components and CubeSat platforms).

Although these advances create many opportunities

for new research and business models, a number of

striking new challenges need to be tackled in order to

eﬃciently manage the available resources while also

ensuring maximum payload utilization. In particular,

the size and weight limitations of low-Earth orbit

(LEO) small satellites mean that their successful op-

eration rests on a ﬁne balance between solar power

infeed and the power demands of the mission pay-

load and supporting platform technologies, buﬀered

by on-board battery storage. This renders a non-

evident, recurring, and intricate scheduling problem

to be solved on the ground segment, namely the

continual need to make decisions about which task

the satellite is to eﬀectuate next. This requirement

will arguably become the bottleneck for the growing

trend of scaling the space segment to constellations

and mega-constellations.

To this end, we contribute sophisticated software-

based automated solutions rooted in optimal com-

puter science techniques validated in modern nano-

satellite networked missions operating in orbit. This

paper ﬁrst introduces the GOMSPACE Hands-Oﬀ Op-

erations Platform (HOOP), a ﬂexible and scalable

end-to-end satellite operation framework for com-

manding and monitoring subsystems, single-satellites,

or constellation-class missions. By taking advantage

of the vast expertise of GOMSPACE, new space ac-

tors can leverage ﬂight-proven toolchains throughout

the mission lifecycle while proﬁting from partner

ground station networks without the need to invest

in their own operational infrastructure. Second, we

present how HOOP is enhanced by highly eﬃcient

and accurate automated decision-making capabili-

ties exploiting dynamic programming and learning

techniques based on profound battery and electric

power budget models, developed at Saarland Univer-

sity as part of the POWVER initiative. The models

are continually kept accurate by extrapolating data

from telemetry received from satellites. The resulting

machine learning approach delivers optimal, eﬃcient,

scalable, usable, and robust ﬂight plans, which are

provisioned to the satellites with zero need for hu-

man intervention—but which are still under the full

control of the mission operator. Third, we report on

the application of the POWVER-HOOP approach to

GOMX–4, the dual-satellite mission by GOMSPACE

that is currently in orbit. Over a period of more than

a month, a series of in-orbit experiments have been

carried out with the 6U CubeSats, covering Earth

observation, air traﬃc surveillance, as well as inter-

satellite linking capabilities. In these experiments,

the integrated POWVER-HOOP toolchain has shown

its unique strength, namely to operate a mission

without human intervention while persistently deliv-

ering maximum return from its observation payloads

and ensuring the most eﬃcient and safe utilization

of constrained on-board battery resources. We make

these ﬁndings concrete by reporting details of a 48-

hour period selected from the masses of recorded

experimental results.

This pioneering work evidences that humans can

deﬁne and supervise high-level objectives of the mis-

sion while relying on machine learning approaches to

ﬁnally unblock the future of space operations.

CONTEXT

GOMSPACE and Their Mission

Since the foundation of the company in 2007,

GOMSPACE has become a leading manufacturer and

supplier of CubeSats and small satellite solutions

for customers in academic, government, and com-

mercial markets. The key strengths of the company

include systems integration, CubeSat platforms, ad-

vanced miniaturized radio technology, and satellite

operations. The GOMSPACE headquarters are lo-

cated in Aalborg, Denmark. The company also has

a propulsion technology center in Uppsala, Sweden,

and a satellite operations center in Esch-sur-Alzette,

Luxembourg. The company currently employs more

than 150 people and provides services to customers

in more than 60 nations.

GOMSPACE has a track record of successful missions

in space. This is exempliﬁed by the GOMX series

of satellites, all of which were built and operated by

GOMSPACE. GOMX–1, a 2U satellite launched in

November 2013, successfully demonstrated for the

ﬁrst time the reception of ADS-B signals from air-

craft by an orbiting satellite. The satellite remains

in orbit. GOMX–3, launched in 2015, demonstrated

attitude control, downlinking of data, and SATCOM

spot-beam characterization. The satellite success-

fully completed its nominal mission and re-entered

the atmosphere after one year. That mission was

followed by the GOMX–4 mission, a pair of two 6U

CubeSats. This mission demonstrated the ability

of CubeSats to act in coordination through inter-

satellite communication. Payloads of the satellites

are used for surveillance and monitoring of Arctic

regions.

These projects are delivered based on GOMSPACE’s

strong in-house portfolio of established products and

rich capabilities. Currently, GOMSPACE is develop-

ing the Juventas nanosatellite that will form part of

ESA’s Hera mission to the Didymos binary asteroid

system.

The mission will provide valuable scientiﬁc

data from the asteroid system, including radar and

radio science observations of the binary system.

With more than 13 years of experience in the market

and a track record of multiple successful missions

accomplished, GOMSPACE has developed profound

knowledge and competencies within radio technology,

CubeSat platforms, project management and inno-

vation. Starting early as a pioneer in the market,

GOMSPACE has now become a market leader in the

commercialization of nanosatellites and new space

technology.

The numerous lessons learned by GOMSPACE dur-

ing past satellite projects and operations materialize

into HOOP. Described in detail below, HOOP is a

platform enabling automated satellite operation at

low cost that is individually adapted to the speciﬁc

mission requirements. As evidence of HOOP’s ﬂex-

ibility, this paper presents its seamless integration

with the LEOPOWVER toolchain.

POWVER at Saarland University

Since 2013, Saarland University is performing sci-

entiﬁc research on the operation of nanosatellites,

seeded in the EU FP7 project SENSATION

∗

, where

GOMSPACE acted as an industrial partner. The

unique expertise of the Saarland University re-

searchers lies in the application of formal meth-

ods to perform automatic and resource-optimal task

scheduling of satellites and satellite constellations.

Early work developed the scientiﬁc grounds along the

GOMX–1

and GOMX–3 missions, including in-orbit

demonstrations

4,5

. In these contexts, it had become

apparent that there is massive room for improvement

by properly modeling and analyzing the satellite’s

battery, operational constraints, and orbital environ-

ment. The research intensiﬁed as a consequence of a

multi-million dollar award, the ERC Advanced Grant

POWVER

†

, that was awarded in 2016 to Holger Her-

manns by the European Research Council to foster

the research. Another (albeit smaller) grant, the ERC

∗https://cordis.europa.eu/project/id/318490

†https://cordis.europa.eu/project/id/695614

Proof of Concept Grant LEOPOWVER

‡

, is nowadays

the focus point of all application and commercializa-

tion activities of what has been developed success-

fully over the years: orbit-proof software enabling

the continuous, fully automated, energy-optimal, and

proﬁt-maximizing dynamic operation of LEO satel-

lites and satellite constellations.

At the core of the LEOPOWVER software is a unique

collection of highly realistic battery models that en-

able eﬃcient forecasting of battery health and battery

depletion risk with unprecedented accuracy. This is

paired with ultra-eﬃcient optimization and machine

learning techniques that are tailored to the LEO

context, including dedicated support for telemetry

processing and contact plan design for satellite con-

stellations.

Indeed, communication transmitters and

transponders are among the most power-demanding

subsystems of any modern spacecraft. This phe-

nomenon is exacerbated in networked space constel-

lations supported by one or more Inter-Satellite Link

(ISL) interfaces. As a result, a notorious bottle-

neck is provoked by power-hungry networking tasks

that need to be powered by constrained batteries

and solar power infeed. This asks for a very care-

ful time-evolving and data-driven scheduling of the

communication resources.

The methodologies delivered by LEOPOWVER are

targeted at the core of this problem, which needs

to be perpetually solved during the mission lifetime.

Thus, the role of the operator is reduced to the most

important aspects of the mission: deﬁning the goal

and the conditions to achieve it, leaving computer

science algorithms to ensure the optimal and secure

control of the space system. LEOPOWVER harvests

very advanced algorithmic and learning approaches,

which are already unlocking the optimal battery-

aware operations of future cross-linked satellite con-

stellations.7,8

The spirit of LEOPOWVER is to make this unique

combination of technologies ready for take-up by em-

bracing start-ups, academics, and business customers

with the intention to create a prospering user commu-

nity. Thus, favorable conditions will be enabled via

a base version of the software to be released under

open-source licensing in the near future.

To this end,

the LEOPOWVER orchestration toolchain is designed

to provide highly ﬂexible support for integration into

arbitrarily complex operations workﬂows via well-

deﬁned telemetry and commanding interfaces across

the entire spectrum of orbit applications.

‡https://cordis.europa.eu/project/id/966770

Check

https://leopowver.space/

for the latest updates.

Customer

API

Mission

Planner

Service

Assets

Gateway

Assets Data

Ingress

Gateway

GOMSPACE

GSN and

Space

Adapters

Scheduler

Executor

Mission Operator

HANDS-OFF OPERATIONS PLATFORM (HOOP) EXTERNAL ASSETS

Figure 1: Overview of the GOMSPACE Hands-oﬀ Operations Platform (HOOP).

In this paper, we describe how this ﬂexibility is ex-

ploited for a straightforward and successful integra-

tion of LEOPOWVER with the GOMSPACE HOOP

operations platform to then validate its applicability

in the state-of-the-art GOMX–4 nanosatellite mission

currently in orbit.

HANDS-OFF OPERATIONS PLATFORM

The GOMSPACE Hands-oﬀ Operations Platform

(HOOP) is a satellite operations platform built for

automation, scalability, and ﬂexibility. The HOOP

platform was developed by GOMSPACE speciﬁcally to

handle operations for constellations of CubeSats, up

to constellations with thousands of satellites. Since

2018, the platform has been under development with

support from the Luxembourg Space Agency (LSA)

and European Space Agency (ESA).

Unlike traditional operations centers, which carry

out operations manually with large numbers of satel-

lite operators, operations centers for constellations

must automate much of the nominal operations. The

HOOP system has been designed to support and

manage this degree of operations, and thereby allow

operations to scale to support hundreds or thousands

of satellites.

In this way, HOOP streamlines routine operations,

allowing operators to focus on troubleshooting and

improving the mission. This reduces the manual ef-

fort required to monitor and maintain each satellite,

and also means that routine activities, such as pay-

load data capture and download, can be completely

automated and require no manual input at all.

HOOP provides a set of distinguishing features for the

management and autonomous operation of satellites,

from the ﬁrst satellite to a global constellation. A

schematic overview displaying all components that

HOOP consists of is shown in Figure 1.

Conﬁguration Management.

A diﬃcult chal-

lenge in satellite operations is to track and know

the satellite conﬁguration at all times, even when

the satellite is out of visibility of a ground station.

HOOP addresses this challenge by providing a set

of conﬁguration management tools. The last-known

satellite conﬁguration is stored in a database and

visible to operators through a user interface. All

changes planned through HOOP are recorded, and

all downlinked telemetry is monitored for any discrep-

ancies with the expected conﬁguration. Operators

are alerted if any unexpected conﬁguration changes

are detected. The database used by HOOP allows

the conﬁguration of multiple satellites to be tracked.

When an issue arises with a satellite, an operator can

call up the relevant satellite from the database and

review the last known conﬁguration of all on-board

parameters.

Mission Planning.

The operations software han-

dles both manual and automated mission planning.

Operators can deﬁne routine procedures that are

scheduled according to speciﬁed rules (based on

events or time). They can also plan manual pro-

cedures as required, expressed in a high-level yet

ﬂexible procedure language. A plan resolver checks

all operations plans, generated either manually or

automatically, and veriﬁes that there are no conﬂicts

present. If conﬂicts are found, the resolver attempts

to resolve them. If no solution can be found, it will

reject the operations plan and send an alert to the

operator.

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

OntheAutomation,Optimization,andIn-OrbitValidationofIntelligentSatelliteConstellationOperationsGregoryStockSaarlandUniversitySaarlandInformaticsCampusE13,66123Saarbrücken,Germany;+496813025635stock@depend.uni-saarland.deJuanA.FraireSaarlandUniversity/INRIASaarlandInformaticsCampusE13,Saarbrücken,Ger...

展开>> 收起<<

On the Automation Optimization and In-Orbit Validation of Intelligent Satellite Constellation Operations.pdf

共18页,预览4页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

On the Automation Optimization and In-Orbit Validation of Intelligent Satellite Constellation Operations

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: