The Sky is NOT the Limit Anymore:Future Architecture of the Interplanetary Internet

Ahmad ALHILAL∗, Tristan BRAUD* and Pan HUI∗†

*The Hong Kong University of Science and Technology - Hong Kong†University of Helsinki - Finland

1Email: aalhilal@ust.hk, braudt@ust.hk, panhui@cse.ust.hk

Abstract—The ever-growing popularity of space explorationhas attracted many organization and private companies toattempt increasingly challenging missions. Each of these mis-sions bring us one step closer to the colonization of the solarsystem. Many initiatives have started aiming at providing anarchitecture in the so-called Interplanetary (IPN) Internet toserve these upcoming missions. In this paper, we propose anevolutionary, scalable, and extensible architecture adapted thecurrent IPN communications challenges and the milestones offuture space exploration missions. This architecture providesthe foundations for the communication and navigation servicesto run the missions. We introduce the key elements of ourarchitecture, from the communication spectrum and technologiesto the physical elements to place in deep space, going through thecommunication protocols with support of Delay Tolerant Network(DTN), autonomous operation, and Software Defined Networks(SDN). We then analyze the implications of such architectureon data delivery over the (Jupiter −→ Mars −→ Earth) path. Wefinally perform a preliminary evaluation of the performance ofthe architecture to support our proposal and display the evolutionof the end-to-end latency at each milestone.

Index Terms—Interplanetary Internet, Delay Tolerant Net-work, Software Defined Networking.

I. INTRODUCTION

SPACE exploration is not only feeding human curiosity, butalso allows for scientific advancement in environmental

research, and in finding natural resources [1]. Although mediaexposure reached its peak during the Apollo programs, spaceresearch remains a very active domain, with new explorationand observation missions every year.

Following the Apollo program, public and private orga-nizations launched a wide variety of exploration missionswith increasingly complex communication constrains. In 1977,NASA launched Voyager 1 and 2 [2] to explore Jupiter, Saturn,Uranus and Neptune. In September 2007, Voyager 1 crossedthe termination shock at 84 AU which is more than twice thedistance to Pluto. The Voyager Interstellar Mission (VIM), anextension to the 1977 Voyager mission, will explore the outer-most edge of the Sun’s domain and beyond. Regarding nearbyplanets, NASA launched several exploration rovers in 2004and 2012, followed in 2019 by the InSight lander. Currently,6 active satellites orbit around Mars, with primary purposeof studying the atmosphere, relay data for other missions,or test key technologies for interplanetary exploration [2].Private companies are also starting to take part in spaceexploration. SpaceX and Blue Origin were founded to reduce

the cost and increase the safety of space flight, and aim fornear-future Mars exploration. Finally, in 2018, Luxembourgbecame the first country to legislate for asteroid explorationand mining, opening the way for a whole new space industry.However, each mission operates independently, has it owndedicated architecture, uses point-to-point communication, andis dependent on operator-specific resources. In this paper, wepropose an inter-operable infrastructure in a similar fashionto the Internet at stellar scale to simplify communication forupcoming space missions.

In recent years, space exploration managed to attract a lotof media attention, resulting in a clear regain of interest ofthe public for space exploration. As a consequence, spaceagencies started to plan several ambitious missions for the22nd century, both manned and unmanned. These missionsrequire increasing communication capabilities, proportional totheir complexity. Such missions require a reliable, scalableand easy to deploy common communication infrastructure totransmit scientific data from outer space to the earth and back.The advantages of such strategy are manifold: (1) Interoper-ability: It significantly reduces the cost of communication andfacilitates inter-agency cooperation. Future missions wouldbenefit from sharing resources. (2) Security: Organizationscan work together towards a reliable and secure infrastructure.(3) Increased bandwidth: Future missions will require ahigher bandwidth to provide increasing amounts of data to thegeneral audience. (4) Scalability: Since space exploration isan incremental process, it is more efficient to progressivelyscale up the network than deploying all the resources atonce. (5) Colonies: Several organizations envision definitiveMars colonies. Interconnecting both planets networks wouldfacilitate the expansion of human knowledge.

In this paper, we propose an evolutionary architecture foran Interplanetary (IPN) Internet [3]. Such an architectureinterconnects networks from various organizations to form aunified network. To address future space exploration missionsrequirements, we design 3 inter-compatible architectures, eachof which corresponding to a given exploration milestone: near-term (current missions), mid-term (human colony on Mars) andlong-term architectures (manned and unmanned colonizationof the complete solar system). We support these architectureswith the most effective technologies for long, dynamic andautonomous usage. Our contribution is threefold:• Proposition of a bottom-up scalable and integrated, ar-

chitecture for future IPN Internet, taking into account theincoming milestones of space exploration.

• Analysis of the implications of this architecture over thepath (Jupiter −→ Mars −→ Earth).

• Preliminary evaluation of the architecture.The remainder of this paper is organized as follows. Sec-

tion II outlines the related work towards an Interplanetary IPNInternet architecture. We then introduce the existing architec-tures in Section III. Afterwards, we propose our architecture,demonstrate the data flow, and discuss further implementationnotes. Finally, we support our findings with a preliminaryevaluation of the performance in Section V.

II. RELATED WORK

The earliest works in space communication are concurrentto space exploration itself. The ideas of mutualization andreutilization emerged as soon as 1982 [4]. In this section, wepresent the key studies related to our proposed architecture.

Many studies propose novel infrastructures for the nextmilestone of space exploration: Earth-Mars deep space com-munications. Wan et al. [5] propose a structure of satelliteorbits based on several 2D planes to optimize the point-to-point physical wireless link capacity. Gladden et al [6] discussthe current infrastructure and its limitations. The authorsconsider the necessary predecessor technologies to implementa delay and disruption tolerant network (DTN) with automated,in-situ communication scheduling. These pending technologiesenable such a network to be scalable by assimilating spacecraftfrom a wider cast of participating organizations in the nextdecades. At NASA, Bhasin et al [7] describe an architectureto support higher data volumes for Mars exploration alongsidethe spectrum (X, Ka-band). While proposing a scalable com-munication architecture to maximize data delivery (> 100Mb),they also define the requirements for such architectures: thearchitecture elements and interfaces, layered/integratedcommunication architecture, and the Communication nodes(rovers, satellites, spacecrafts etc.).

Regarding the currently deployed architecture, Mars orbitersoperate as relays, receiving data from a lander on Marssurface and sending them to a single destination: Earth. Theground operators on Earth inform these orbiters the identityof the asset they communicate prior. The ground station thenprocesses, depackages, resorts and deliver the data to itsfinal destination [6]. More recently, NASA experimented withsatellites acting as bent-pipe relays to land the InSight mission.These relays allowed to keep a line of sight communicationwith Earth during the critical steps of the landing, resultingin an end-to-end delivery delay of 8 minutes. Without thesesatellites, NASA would have had to wait for the MRO tocomplete its revolution around Mars to transmit data, threehours later [8].

Due to the long distances, space communication introducesunshrinkable latencies. Fraire et al [9] investigate the DTNtechnologies. They prove that DTN features could become avaluable means to achieve data delivery in future interstellarnetworks. Project Loon [10] uses TS-SDN for inter-operation

and coordination of aerospace networks. Finally, JPL deployedthe first DTN Gateway located 25 million kilometers fromthe Earth during the ”Deep Impact Network Experiment(DINET)” [11]. The currently operating architectures are lim-ited and cannot be scaled up to the intended future mis-sions. Moreover, many solutions are purely mathematical [5]without considering the concrete implementation, namely thehardware, the spectrum, the communication protocols, andthe prospective technologies highlighted in [6]. These aspectsneed to be complemented by the use of technologies adaptedto IPN conditions such as DTN [9], [11]. In this paper, weprovide a holistic view to a long term, scalable and evolutivearchitecture. This architecture can unfold over time to adapt tospace exploration missions from near-term to long-term future.It covers the milestones of solar system exploration through3 inter-compatible architectures, that we refer to as near-term (current missions), mid-term (human colony on Mars)and long-term (manned and unmanned colonization of thecomplete solar system). Our evolutive IPN architecture notonly takes into consideration the time frame of the futureexploration and science missions, but also NASA’s require-ments [7]. Moreover, we move further to shed the light on themost effective technologies to embed on IPN nodes towards along term architecture.

III. TOWARDS AN IPN INTERNET

Proposing any Inter-planet network demands studying thesupportive architectures, the challenges enforced in the en-vironment and the potential technologies to overcome thechallenges.

A. Satellite and Space Communication Networks

The current satellite infrastructure can be broken downinto space segment and ground segment. Nowadays, satellitesare distributed over 3 orbits [12]: Geostationary Earth Or-bit (GEO),Medium Earth Orbit (MEO), and Low EarthOrbit(LEO). The combination of these satellites covers thewhole surface of Earth thanks to Inter-Satellite Links (ISL).The current space communication architecture operated byNASA embraces 3 operational networks [13]: The Deep SpaceNetwork (DSN) is composed of 3 equidistant ground stationsto provide continuous coverage of GEO orbits, and unmannedspacecraft orbiting other planets of our solar system, theNear Earth Network (NEN) consists of both NASA andcommercial ground stations and finally, the Space Network(SN) is a constellation of geosynchronous relays, Tracking andData Relay Satellite System (TDRSS), and ground stations.

This infrastructure is very convenient for an IPN Internetand can be used as an access network between the Earth andother planets. Establishing colonies demands such constella-tions on the host planet to provide full surface coverage andinterconnect with other planets’ Internet.

B. IPN Challenges

Most of the nodes involved in an IPN Internet are revolvingaround other stellar objects: planets revolve around the Sun

with long distances, satellites orbit planets at a relatively closerange. This motion poses many challenges to the interplanetarycommunication [14]–[16]:

1) Extremely long and variable propagation delays: 3 to20 minutes from Mars to Earth, 4 to 7 hours from Plutoto Earth, depending on their relative positions.

2) Intermittent link connectivity: the Sun or other planetsmay temporarily obscure a given link between two stellarobjects. For instance, the Earth to Mars line of sightis regularly obstructed by the Sun when they reach theopposite position in their orbits [17].

3) Low and asymmetric bandwidth: the limited payloadof the satellites severely impacts their transmission powercompared to Earth transmission relays.

4) Absence of fixed infrastructure: Nodes are in constantmotion which leads to time-varying connectivity. In-stances of connectivity are planned and scheduled ratherthan opportunistic.

An IPN Internet architecture must address these constraints tooptimize the few resources available in the system. Contraryto other opportunistic networks, the motion of the nodes in thesolar system follows regular patterns that can be precalculated.As such, the sender can request a graph of contacts to deliverthe data to its next destination.

C. Delay Tolerant Networks (DTN)

The end-to-end latency in an IPN Internet can reach up toa day, and jitter is measured in hours. As such, conventionalInternet architectures based on the TCP/IP stack are not appli-cable. Delay-tolerant architectures and protocols are designedto withstand the extreme constrains of the system. Fraire atel [9] prove that DTN protocols can be conveniently combinedwith the infrastructure of interstellar relay systems to avoidretransmission of data on long distances and, thus, achievelower end-to-end latency while reducing the relay buffer re-quirements. Therefore, we consider the DTN architecture andprotocols to be the building blocks of our IPN network. Thefirst concepts of DTN were originally proposed to cope withthe characteristics of deep space communication (long delays,discontinuous network connectivity) before being extendedto other domains. The DTN nodes provide store-and-forwardcapabilities to handle the eventual link unavailability. DTN ar-chitectures insert an overlay network protocol called BundlingProtocol (BP), that provides end-to-end transmission betweenheterogeneous links [18]. This overlay complies with existingInternet infrastructures. At each point of the network, BPemploys the transport protocol adapted to the transmissionconditions. BP therefore operates operates over TCP, UDP andLicklider Transport Protocol (LTP) [14] to provide a point-to-point transmission protocol for intermittent links. Suchcharacteristics make BP particularly adapted for interplanetarytransmission, where the traditional TCP/IP paradigm cannot beapplied.

Fig. 1: IPN near-term Architecture. Reusing the existinginfrastructure with minimal addition.

IV. ARCHITECTURE & COMMUNICATIONINFRASTRUCTURE FOR FUTURE IPN

Deploying and operating a long-term architecture is ex-tremely intricate and demand a lot of time. In addition, the ar-chitecture should withstand the harsh constraints of deep spacecommunication. Given these constraints, an incremental andevolutionary architecture is required. As such, we define anevolutionary architecture which consists of 3 inter-compatiblesub-architectures, each of which is corresponding to differentmilestones in space exploration.

A. IPN Near-Term Communication Architecture

We propose an IPN near-term communication architecturefor the current missions targeting Mars and the Moon. Bothare accessible within a reasonable amount of time and severalorganizations are already planning manned missions within thenext 10 years. This architecture reuses a maximum amountof available technologies to interconnect the Earth, Mars andthe Moon in a short time frame. Figure 1 illustrates the IPNnear-term architecture. We separate this architecture in twosub-systems: the physical layer, that we will call Spectrum,and the upper layers, referred to as Network. The spectrumsub-system provides two bands in the microwave spectrumfor data: Ka (26.5-40GHz) and X (8-12.4GHz). These bandsprovide higher data rates than the conventional RF bands.The Ka-band allows for the communication in the backbonenetwork and for inter-satellite communication due to its higherfrequency (thus, higher data rates). X and Ka together allow forthe communication from satellites to the surface of the planet.In our architecture, we switch between both bands dependingon the weather as the Ka-band suffers from attenuation inpresence of humidity.

The network sub-system contains 3 sub-networks (seeFigure 1): The proximity network contains the inter-elementlinks relatively close to the planet or the Moon and the surfacenetworks. The access network consists of satellites orbitingthe planet or the Moon interconnected with each other. In

Fig. 2: End-to-End Data Transfer Using DTN. In near-termarchitecture, the DSN directly connects to the Mars Orbiter,which relays the bundle to the rover using BP over LTP.

our architecture, there are 3 access networks formed by thesatellites orbiting each planet and the Moon. The backbonenetwork interconnects the 3 access networks with the DSNstations on Earth. This network provides two kinds of linksfor interconnection: direct links and indirect links. Direct linksconnect Mars and lunar relay satellites directly to the DSN onEarth. Indirect links go from Mars and lunar relay satellites toEarth relay satellites where the data is then directed to DSNantennas. We propose to launch four Lunar Relay Satellites(LRS) and four Mars Relay Orbiters (MRO). Three relaysatellites are operating, while the fourth remains as a spare.On the Earth, we reuse TDRS satellites, currently operating ingeostationary orbit, to serve as relay of the data arriving on theindirect links. Each node along the path has SmartSSR tech-nology to support DTN functionality. Therefore, it is installedon DSN, relay satellites (LRS, MRO, TDRS), and even in theproximity networks: Landers, Robots, Rovers (data collectors),SN and mission centers (data destination). SmartSSR is asolid-state recorder (SSR) developed by the Applied PhysicsLaboratory (APL). Its small mass and size combined withits relative low cost make it easy to massively install on thepayload module of any spacecraft. These features make theSmartSSR an optimal choice to provide DTN capabilities. Itfeatures JPL’s DTN ION implementation and uses the SpaceFile System (SpaceFS) to manage spacecraft data. SpaceFS isadapted to meet the special requirements of space operationalenvironment [19]

Fig 2 outlines the data transmission process from themission center on the Earth to a Mars rover. BP functionsas an overlay over TCP between the mission center and theDSN antennas, and over LTP in second trunk (DSN antennas−→ Mars Orbiter −→ Mars Rover). On the latter, we utilize themicrowave band Ka and LTP to transmit the data over the longdistance – high latency – between the Earth and Mars. Thegreen continuous line depicts the data path from the missioncenter to the Rover. The purple dashed lines show the hopto hop acknowledgments between two neighboring elementsgained from custodian transfer property, a property of DTNfunctionality supported through utilizing SmartSSR hardware.Currently, Mars and Moon’s Orbiters use the Proximity-1data link protocol to communicate with the surface elementsand the Advanced Orbiting Systems (AOS) space data link

Fig. 3: IPN mid-term Communication Infrastructure. We startto deploy Lasercom for long distance links, and extend thearchitecture to other planets.

protocol for communication between orbiters and the DSNantennas on Earth.

B. IPN Mid-Term Communication Architecture

For our mid-term architecture, we consider the humancolonization of Mars and the further side of the moon, whichwill further lead to the long-term goal of colonizing the wholesolar system. As such, we expect an ever-increasing demand toexchange huge amounts of data in both directions. The lowerpower consumption, lower mass, higher range and higherbandwidth of optical communication compared to RF make itan auspicious technology to serve as a communication mediumin IPN [20]; Therefore, we propose using an onboard opticalmodule for spacecrafts and optical communication terminals(OCT) on the planet’s surface to support two-way communica-tion with high data rates. This design allows us to considerablyreduce the bandwidth asymmetry. These technologies requireless power and considerably reduce the payload. They are alsoable to reach longer distances and provide higher data rates,10X - 100X higher than RF.

Figure 3 illustrates the IPN mid-term architecture thatinterconnects the Earth with Mars and other planets. In thisarchitecture, we upgrade the transmission spectrum from mi-crowave (X, Ka) to optical communication (i.e, FSO). Opticalcommunication is an emerging technology in which data ismodulated onto a laser for transmission. The laser beam issignificantly narrower than a RF beam and thus promises todeliver more power and achieve higher data rates. In outerspace, the communication range of FSO is on the order ofthousands km. Optical telescopes, therefore, play a pivotalrole as beam expanders to bridge interplanetary distances ofmillions of kilometers. To this purpose, each spacecraft carriesa small (a dozen cm) Cassegrain reflector, a 22cm aperture,4 W laser and contains an isolation and pointing assembly

Fig. 4: Spacecrafts placement at Sun-Earth Lagrangian points.When the Mars-Earth LOS is obscured, data can go throughthe LDRS or FWRS to avoid service interruption.

(IPA) for operating in the presence of spacecraft vibrationaldisturbance, and a photon-counting camera to enable theacquisition, tracking and signal reception.

The planet’s ground optical terminals contain photon-counting ground detectors that can be integrated with largeaperture ground collecting apertures (telescopes) for detect-ing the faint downlink signal from deep space. The groundOCT contains six small (a dozen cm) refractive telescopesfor the transmitter and a single bigger reflective telescopeas a receiver. The latter is connected via optical fibers tothe destination. The operating constellations in this archi-tecture are Optical TDRS around Earth, Geostationary MarsOrbiters (GMOs) and Geostationary Planet Orbiter (GPOs).They provide relay services between nodes at the surface ofthe outer planet, in-between planets and between the accessnetwork from other planets. The Optical Deep Space Network(ODSN) substitutes the DSN ground stations by supportingtwo communication technologies: RF microwaves (X and Ka-band) and optical (Lasercom). This hybrid results in installingoptical mirrors in the inner 8m of a standard DSN 34m beamwaveguide antenna. The RF communication is kept in orderto maintain the operation in all weather conditions.

C. IPN Long-Term Communication Architecture

Optical communication in space is based on line of sight(LOS), which may experience obstruction or conjunction. Forinstance, Earth and Mars can be obscured from each other bythe Sun. This obstruction lasts for two weeks every 26 months.Moreover, LOS communication in space attenuates becauseof Free-Space Loss (FSL) which increases with distances.Therefore, communication between the Earth and further plan-ets experiences much more attenuation than communicationbetween the Earth and Mars. If we consider transmissionbetween the Earth and Pluto, the signal travels 38.44 AU =5,766 million km (0.52 AU for Earth to Mars) in space andneeds 5.4 hours to reach its destination.

We propose operating spacecrafts in Sun-Earth’s Lagrangianpoints to address these problems. Fig 4 shows the positions

of the five Lagrangian Points L1,L2,L3,L4,L5. At each point,the gravitational forces of two large bodies (Sun-Earth for in-stance) cancel the centrifugal force. A spacecraft can thereforeoccupy the point and move around the Sun without the needfor external intervention. These points are commonly used forobservation missions and are envisioned as relays for spacecolonization. In our architecture, we employ these points tooperate spacecraft as repeaters. These repeaters bridge thelong distances, relay the data between the planets and providealternative paths for the routing in the space. More specifically,we propose to operate spacecraft in points L4, Lead RelaySpacecraft (LDRS), and L5, Follow Relay Spacecraft (FWRS),to address obstruction and conjunction, and in L1, Front EndRelay Spacecraft (FERS), and L2, Back End Relay Spacecraft(BERS), to tackle attenuation. In addition, this will fragmentthe tremendously long path to to amplify the weak signals.

D. Autonomous Operation

Currently, space communication systems are mission-specific and point-to-point. Moreover, they are dependent onoperator-specific resources. Our approach aims at reducingthe dependency on resource scheduling provided by Earthoperators and interconnect the planets. To do so, our archi-tecture provides autonomous operation on each spacecraft toautonomously deliver the data in space.

To this end, we employ DTN alongside with allowingthe communication terminal of the satellite to control theantenna pointing, transmit power and data rates, and providesynchronization capabilities between the sender or the receiver.The next step is to provide interactive links between the nodesthat can be created and broken on demand at any time in thewhole IPN Network.

These on-demand features require specific hardware forpointing and focusing transmission. We propose to use CoarsePointing Assembly (CPA) and Fine Pointing Assembly (FPA)[21] to orient the antenna and the beam. Pointing synchro-nization between the sender and the receiver is very crucialto provide autonomous operation of the communication links.Therefore, we incorporate the following two subsystems:

1) Localization Subsystem: The HORIZONS System [22]provides the solar system’s spatiotemporal data and the accu-rate ephemerides for solar system objects. These ephemeridesprovide the sender spacecraft’s Optical Communication Termi-nal (OCT) with spatiotemporal information about the receiverspacecraft’s OCT location at a given time. The sender’s OCTlocates the receiver’s OCT by feeding the reference positionand the new position to the position controller and feedingthe reference velocity and the current velocity in the velocitycontroller. It then estimates the distance to its partner andadjusts the transmission power accordingly. The orbits of all ofthe planets are within a few degrees of the same plane whichmakes the solar system disc-shaped; therefore, the localizationprocess is achieved by adjusting both the azimuth (the horizon-tal orientation of the sender’s OCT in relation to Sun’s equator)and the elevation (the vertical tilt of the sender’s OCT) anglesof the pointing assembly. The expected localization accuracy

is Azimuth < 1 arcSecond, Elevation< 1 arcSecond, where 1arcSecond ≈ 4.85µrad [23]. Likewise, the expected angularaccuracy is < 5µrad. The angular disturbance can only reacha few hundreds of nanoradians to stabilize the laser beamin the presence of spacecraft base motion disturbances andvibrations. To achieve this accuracy, a Pointing and VibrationControl Platform (PVCP) can be employed. PVCP integratesthe pointing with vibration isolation to reduce the disturbanceand thus improve the pointing control accuracy [24], [25].

2) Pointing Control Subsystem: Once the sending space-craft’s OCT adjusts the orientation and the direction of thelaser beam, the next phase of pointing starts. The techniquehere employs the same laser as a beacon and for transmission.The beam-width is controlled from broad in the acquisitionstage (also referred to as Coarse Pointing), to narrow in thetracking stage (also referred to as Fine Pointing). The acqui-sition is achieved by the hardware 2-axis gimbals’ pointingand a Coarse Pointing Assembly (CPA) which allow contactwith broad beacon beam. When acquired, the beam focusingphase (Fine Pointing) progressively narrows the beam whilecorrecting the pointing accuracy up to sufficient level of beamconcentration to get maximum received power, thus high data-rates. This stage uses either the Fine Pointing Assembly(FPA) or beam control approach which includes 3 controlcomponents: Fast Steering Mirror (FSM), Point Ahead Mirror,and Laser Beam Defocus Mechanism (LBD).

In short, synchronizing the sender and the receiver requiresto incorporate these subsystems to quickly find the partner andreduce the off-line time. Afterwards, the sender starts sendingdata as bundles using DTN technology. Custody data transferensures end-to-end reliability on the hop to hop basis.

E. DataFlow and Bundle Delivery

In the light of the long-term architecture IV-C, we hereindiscuss the delivery of the scientific data on the return link( e.g. Mars lander −→ intermediate spacecraft −→ Earth’sstation ). This subsection demonstrates how to employ aDTN architecture to deliver the data packets on the returnlink. The Packet data unit (PDU), in the context of DTN,is called Bundle. In fact, using Bundle as PDU guaran-tees the application of custody data transfer which in turnguarantees end-to-end reliability on the basis of hop-to-hopcustody transfer [6]. We propose to install a SmartSSR oneach node (i.e. satellite, spacecraft, lander, and DSN) withthe Interplanetary Overlay Network (ION) software v3.7 orhigher to support the DTN capabilities, the Contact GraphRouting (CGR) implementation and the Consultative Com-mittee for Space Data Systems’s (CCSDS) Schedule-AwareBundle Routing (SABR). DTN functions enables the node tostore the bundles, carry them until a link to the next hopis available, and then forward them using CGR or SABRto an endpoint identified by a Compressed Bundle HeaderEncoding (CBHE)-conformant [26] Endpoint Identifier (EID)through the best route with the best delivery time (BDT). BothCGR and SABR rely on accurate contact plan information,provided in the form of contact plan messages [27]. Each DTN

Fig. 5: Mars Rover (#14) to Earth Mission Center (endpointidentifier:ipn://dest node number.service number). Data bun-dles flow through the Mars Geostationary orbiter (#13), whichforwards them to the Earth LDRS (#12). The LDRS thentransmits them to the DSN on Earth (#11), which finallyforwards the bundle to the Mission Center.

node contains a Bundle Protocol Agent (BPA) to originate,forward and deliver the bundles, and is assigned a number byan authority, from a range allocated by the Space AssignedNumber Authority (SANA). Each node must be registeredusing one or more EIDs. Each EID serves as a different DTNapplication operating at that node [28].

Let us consider that a rover on Mars is sending a file to theNASA Mission operation center. The Nasa Mission Centre’sEID is identified using the URI ipn://10.2 which conformsto the CBHE, where 10 is the receiving node number, and2 is the service number for a file transfer application. Theservice number functions as a de-multiplexing token. Assum-ing that the route shown in Figure 4, where the link betweenMars and the Earth is available through LDRS, is used as arelay between Mars and Earth to re-amplify the signal andcounteract eventual occlusion. This route ensures the bestdelivery time (BDT) since its contacts are available anytimeand features the shortest path of any possible routes. Otherpossible routes lead to longer forwarding times. These routesinclude R2=[C0:100

MarsRover,MarsGMO , C0:1000MarsGMO,LDRS , C0:700

LDRS,FERS, C0:100

FERS,DSN , C0:1DSN,NANA], where C is a contact with the

parameters (sender, receiver), (start time: stop time), and a starttime equal to 0 means the link is available for transmissionanytime, whereas the stop time value is set to any number ≥light seconds required to send the bundles.

We illustrate both the data flow and the process flow inFigure 5, where Mars rover sends the file using CCSDS FileDelivery Protocol (CFDP). The rover’s BPA encapsulates thefile into bundles, which are stored and forwarded when the linkwith Geostationary Mars Orbiter (GMO) becomes available.Once it is available, the BPA invokes LTP to transport eachbundle as segments. The segments are sent to the GMOusing either laser communication or Ultra High Frequency RF(UHF). The GMO then follows the same logic to forwardthe bundles to the Earth’s LDRS, using laser communicationand onboard autonomous operation. The LDRS stores thebundles on persistent storage (Flash NAND in SmartSSR) and

Application Layer

Control LayerTS-SDN

Controller

Spectrum Management

Mobility Management

TS-SDN

Application

Topology Management

Bundle Routing

API

Infrastructure Layer

CDPI

Fig. 6: Abstract overview of SDN architecture.

carries them until the link to the next hop becomes available.LDRS communicates with the DSN to forward the bundlesusing the same protocols and communication medium. TheDSN’s BPA invokes the underlying convergence layer agentto transform from LTP to TCP to transport the bundles. TheOCT on DSN directs the optics to the control room wherethe laser is demodulated digital data. This data is finallytransferred using unshielded twisted pairs cables (UTP) tothe mission center, which is the final destination identified byipn:10.2. After delivering data to the mission control center,the de-encapsulation process converts the TCP segments intobundles and delivers them into the CFDP to build the file.For each hop, the receiver’s BPA (Custodian endpoint) createsand replies with an acknowledgment to confirm data reception.This acknowledgment confirms the delivery on a hop-to-hopbasis (dotted lines on Figure 5). The BPA guarantees end-to-end reliability through custody transfers.

F. IPN Implementation Notes

The deployment of an IPN infrastructure takes severalyears. Afterwards, the deployed elements have to operateautonomously for decades. Besides, failure is unpredictableand may cause unforeseeable outages. Maintenance requiresan extremely long time. Therefore, a comprehensive proposalneeds to to address these issues. To this end, we propose tointegrate TemporoSpatial Software Defined Network (TS-SDN) [10], [29] with the autonomous operation IV-D, alignedwith the spacecraft at the Lagrangian points IV-C. TS-SDNdecouples radio control functions from radio data-plane. Itreplaces hardware with software to reduce deployment costs,limit the number of points of failure, and provides the ability toupdate the internal mechanisms without physical access to thehardware. TS-SDN Controllers use a predefined knowledge, asdiscussed in IV-D1, to predict the future state of the lower-levelnetwork. This setup provides a predictive, time-dynamic andholistic view of the topology that includes both the availablewireless links, the approximate accessible time interval ofthe current and the candidate wireless links, and the futurelinks. We thus replace fixing hardware failures with softwareupdates, and drastically reduce the maintenance time fromyears to hours. It also provides autonomous network topology

formation to deliver the data that overcomes any unpredictableproblem along the path.

In Figure 6, we demonstrate how TS-SDN architectureworks and how its components interact. Given the TS-SDNapplication is topology management, this application interactswith TS-SDN Controllers, namely Spectrum management andmobility management through API or so-called northboundinterface. Afterwards, the controller interacts with the under-lying network elements (satellites, spacecraft, rovers, landers,etc) through Control-to-Data-Plane Interface (CDPI). This in-teraction could update the transmit power, the beam width, theRF band used or any network configuration. Each planet hassuch architecture deployed; Therefore, the TS-SDN Controllerhas the knowledge about the position of each planet andits trajectory plus the relative position and trajectory of theorbiters.

Hardware softwarization and robotic missions, however, re-quire the communication systems to offer maximum reliabilitywith robust two-way links for software uploads and updates,virtual interactions and telemetry control. Establishing a bidi-rectional link requires an efficient pointing technology. Whena spacecraft points to another spacecraft, the receiving oneshould be able to infer the location of the sender from the lightsignal itself and respond by an uplink laser beacon to guide thetransmitter, then redirect the laser beam to the exact locationof the the receiving telescope. To this end, a spacecraft needto be provided by with multiple transceivers, each of whichis provided with telescope to bridge the interplanetary longdistances, as discussed in Mid-term Architecture Section IV-B.In addition, the spacecraft should have many capabilities (i.e,localization, coarse-grain pointing, fine-grain pointing), andthe communication terminal is capable of controlling theantenna pointing, transmit power, laser beam direction, asstated in Autonomous Operation Section IV-D.

V. PRELIMINARY EVALUATION OF THE ARCHITECTURE

During the motion of Earth and Mars, the minimum distancebetween them is 54.6e6 km, while the maximum is 401e6 kmwhich happens when they are in opposition relative to thesun. The minimum distance between Earth and Moon is363,104 km, while the maximum is 405,696 km. In the Nearand Mid-term architecture, other planet’s satellites communi-cate directly to TDRS orbiting Earth, so the communication issubject to latency (delay). The speed of light in a vacuum, c0,is 299,792,458 meters per second and the propagation delay,caused by the long distances, is delay = distance/speed;Therefore, the estimated latency for the minimum, maximumand average distance of Moon, Mars, and Jupiter are shown inFigure 7a. Since the IPN mid-term Architecture uses laserCominstead of X-band, we achieve a higher throughput that enablesmissions to transfer 10X → 100X more data from the planetof interest to Earth but same delay as Near-future architecture.

By adding spacecraft LDRS in Lagrangian point L4 forindirect communication (planet→LDRS →Earth), the over-all availability of LOS in Long-term Architecture is 100%

(a) Latency of direct communication, Near/Mid-term

(b) Latency using LDRS for Long-term

Fig. 7: Latency of communication with Moon, Mars andJupiter for Near-term and long-term.

Near-term architecture

104

106

Mars to Earth latencyJupiter to Earth latency

Mid-term architecture

104

106

Late

ncy

(s)

2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028Long-term architecture

Time (years)

103

Fig. 8: Predicted latency to Mars and Jupiter for the next 10years for near (top), mid (middle) and long (bottom) termarchitectures.

since it overcomes the blackout of communication. Com-munication thus follows the alternative path (Mars→LDRS→Earth). The estimated distance of Mars→LDRS path is300e6 km, Jupiter→LDRS is 800e6 km, and LDRS→Earth pathis 200e6 km. Therefore, the average latency for each segmentand the whole path are shown in Fig 7b.

In Figure 8, we estimate the latency of Earth to Mars

and Earth to Jupiter communication over 10 years for ournear-term, mid-term and long-term architecture. In the near-term architecture, the average latency between Earth andMars is varying periodically between 182 and 1337 secondsdue to their respective orbits around the sun. However, onceevery 26 months, the Earth and Mars are in opposition, andcommunication is interrupted for 14 days. During this period,the latency reaches up to 14 days or 1.21e6 s and decreaseslinearly with time until conjunction passes. As our near-termarchitecture only focuses on Mars and the Moon, we assumethat no new equipment is deployed on Jupiter. Currently, onlyone satellite – Juno – orbits around Jupiter, with a periodof 53 days. Consequently, communication is only possiblefor a duration of 26.5 days every 53 days. As such, theaverage latency varies heavily between 2600s and 2.6e6 s.In the mid-term architecture, we assume that Jupiter is nowcovered by a network of stationary satellites. Communicationto Jupiter is thus possible most of the time, except in thecase of opposition. When planets are in opposition, it becomespossible to use another planet as a relay. When Mars and Earthare in opposition, the latency drops to 5175 s, by using Jupiteras a relay. Similarly, Mars to Earth communication can useJupiter as a relay in case of conjunction. Nevertheless, the mid-term architecture does not take into account the case of bothJupiter and Mars being in opposition as it happens in 2027,which results in high latency. As stated previously, our long-term architecture uses LDRS at L4. Not only do these satellitesallow to keep constant communication with Earth, but theyalso considerably lower the latency in case of opposition. Thislatency drops to 1667 s for Mars and 3335 s for Jupiter. Thelong-term architecture thus decreases the maximum latency upto 700 times and allows for communication with Jupiter in lessthan 2 hours year-round.

VI. CONCLUSION

This paper proposes an evolutionary architecture for IPNInternet to migrate from mission-centric architectures to asingle common, scalable and reliable architecture.

In this paper, we first propose 3 inter-compatible architec-tures, each of which corresponds to a given milestone of spaceexploration, near-term (current missions), mid-term (humancolony on Mars) and long-term architectures (manned andunmanned colonization of the complete solar system). Wethen support this architecture with the prospective technologiesthat effectively address the IPN challenges, specifically DTNand the communication protocols that fit each environment.We also address the problem of point-to-point communicationand provide the hardware and subsystems for multipointcommunication and autonomous operation. Long run of sucharchitecture may cause unpredictible failure. We thus proposeto extend the usage of TemporoSpatial Software Definednetwork for efficient, time-less maintenance (e.x. softwareupdate). Afterwards, we demonstrate the sequence of theprotocol stack and the dataflow using a case study (send filefrom Jupiter to Earth through Mars as a relay planet). Finally,we evaluate the latency of this case study.

With this paper, we hope to have provided a novel point ofview for future IPN architectures, and set some foundationsfor an actual implementation within the next decades. In futureworks, we plan to focus on two concrete aspects of spacecommunication at the core of our architecture. First of all,we plan to develop a time-dynamic, predictive communicationlink modeling, and solar objects’ position and velocity mod-eling to support TS-SDN and operating links autonomouslyin deep space. Furthermore, we aim at providing a rigoroustechnical analysis of the costs and benefits of stationingmore inexpensive relay satellites (inspired by CubeSats) atthe Earth/Planet-Sun liberation points. This analysis needs toevaluate the performance in terms of boosting the signals andgained data rate.

ACKNOWLEDGMENT

We thank Dr. Scott Burleigh for reading our paper thor-oughly and providing feedback and comments. His suggestionshave greatly improved our work and broadened our insight topotential future research in the area of IPN Internet.

REFERENCES

[1] B. Dickey, “Benefits stemming from space exploration.” NASA,2013. [Online]. Available: https://www.nasa.gov/sites/default/files/files/Benefits-Stemming-from-Space-Exploration-2013-TAGGED.pdf

[2] NASA. Missions. [Online]. Available: https://www.nasa.gov/missions[3] S. Burleigh, V. Cerf, R. Durst, K. Fall, A. Hooke, K. Scott, and H. Weiss,

“The interplanetary internet: a communications infrastructure for marsexploration,” Acta astronautica, 2003.

[4] The consultative committee for space data systems. [Online]. Available:https://public.ccsds.org/default.aspx

[5] P. Wan and Y. Zhan, “A structured solar system satellite relay con-stellation network topology design for earth-mars deep space com-munications,” International Journal of Satellite Communications andNetworking.

[6] R. E. Gladden, G. Kazz, S. Burleigh, D. Wenkert, and C. D. Edwards,“Next generation relay services at mars via an international relaynetwork,” in 2018 SpaceOps Conference, 2018.

[7] K. Bhasin, J. Hayden, J. R. Agre, L. P. Clare, and T.-Y. Yan, “Advancedcommunication and networking technologies for mars exploration,”2001.

[8] M. Koziol. (2018) Insight mission shows that cube-sats can relay messages from deep space. [On-line]. Available: https://spectrum.ieee.org/tech-talk/aerospace/satellites/marco-demonstrate-that-cubesats-can-get-the-job-done-in-deep-space

[9] J. A. Fraire, M. Feldmann, F. Walter, E. Fantino, and S. C. Burleigh,“Networking in interstellar dimensions: Communicating with trappist-1,” IEEE Transactions on Aerospace and Electronic Systems, 2018.

[15] L. Shi, J. Jiao, A. Sabbagh, R. Wang, Q. Yu, J. Hu, H. Wang,S. C. Burleigh, and K. Zhao, “Integration of reed-solomon codes tolicklider transmission protocol (ltp) for space dtn,” IEEE Aerospace andElectronic Systems Magazine, vol. 32, no. 4, pp. 48–55, 2017.

[10] B. Barritt and V. Cerf, “Loon sdn: Applicability to nasa’s next-generationspace communications architecture,” in 2018 IEEE Aerospace Confer-ence. IEEE.

[11] V. Cerf, S. Burleigh, R. Jones, J. Wyatt, and A. Hooke, “First deepspace node on the interplanetary internet: the deep impact networkingexperiment (dinet),” in Ground System Architectures Workshop, 2009.

[12] L. J. Ippolito and L. J. Ippolito Jr, Satellite communications systemsengineering: atmospheric effects, satellite link design and system per-formance. John Wiley & Sons, 2017.

[13] J. Mukherjee and B. Ramamurthy, “Communication technologies andarchitectures for space network and interplanetary internet,” IEEE com-munications surveys & tutorials, 2013.

[14] G. Yang, R. Wang, K. Zhao, X. Zhang, W. Li, and X. He, “Queue-ing analysis of dtn protocols in deep-space communications,” IEEEAerospace and Electronic Systems Magazine, vol. 33, no. 12, pp. 40–48,2018.

[16] K. Zhao, R. Wang, S. C. Burleigh, A. Sabbagh, W. Wu, and M. De Sanc-tis, “Performance of bundle protocol for deep-space communications,”IEEE Transactions on Aerospace and Electronic Systems, vol. 52, no. 5,2016.

[17] (2017) Solar conjunction. [Online]. Available: https://mars.nasa.gov/allaboutmars/nightsky/solar-conjunction/

[18] K. Zhao, R. Wang, S. C. Burleigh, M. Qiu, A. Sabbagh, andJ. Hu, “Modeling memory-variation dynamics for the licklider trans-mission protocol in deep-space communications,” IEEE Transactions onAerospace and Electronic Systems, vol. 51, no. 4, pp. 2510–2524, 2015.

[19] C. J. Krupiarz, E. J. Birrane, B. W. Ballard, L. Benmohamed, A. A.Mick, K. A. Stambaugh, and E. W. Tunstel, “Enabling the interplanetaryinternet,” Johns Hopkins APL Technical Digest, 2011.

[20] H. Kaushal and G. Kaddoum, “Optical communication in space: Chal-lenges and mitigation techniques,” IEEE Communications Surveys &Tutorials, 2017.

[21] T. Yamashita, M. Morita, M. Shimizu, D. Eto, K. Shiratama, andS. Murata, “The new tracking control system for free-space opticalcommunications,” in Space Optical Systems and Applications (ICSOS),2011 International Conference on. IEEE, 2011.

[22] NASA. Horisons system. [Online]. Available: https://ssd.jpl.nasa.gov/?horizons

[23] J. Wang, Y. Zhou, R. Bai, and G. Wang, “Point-ahead angle and coalign-ment error measurement method for free-space optical communicationsystems,” Journal of Lightwave Technology, vol. 35, no. 18, pp. 3886–3893, 2017.

[24] M. Li, Y. Zhang, Y. Wang, Q. Hu, and R. Qi, “The pointing and vibrationisolation integrated control method for optical payload,” Journal ofSound and Vibration, vol. 438, pp. 441–456, 2019.

[25] J. W. Burnside, D. V. Murphy, F. K. Knight, and F. I. Khatri, “Ahybrid stabilization approach for deep-space optical communicationsterminals,” Proceedings of the IEEE, vol. 95, no. 10, pp. 2070–2081,2007.

[26] S. Burleigh, “Compressed bundle header encoding (cbhe),” Tech. Rep.,2011.

[27] C. .-R.-. DRAFT RECOMMENDED STANDARD, “Schedule-awarebundle routing,” July 2018.

[28] D. G. BOOK, “Solar system internetwork (ssi) architecture,” 2014.[29] B. Barritt and W. Eddy, “Temporospatial sdn for aerospace communi-cations,” in AIAA SPACE 2015 Conference and Exposition, 2015.