EDITOR: Schahram Dustdar, dustdar@dsg.tuwien.ac.at

DEPARTMENT: INTERNET OF THINGS, PEOPLE, AND PROCESSES

The Sky is the Edge—Toward MobileCoverage From the SkyNhu-Ngoc Dao , Sejong University, Seoul, 05006, South Korea

Quoc-Viet Pham , Pusan National University, Busan, 46241, South Korea

Dinh-Thuan Do , Asia University, Taichung, 41354, Taiwan

Schahram Dustdar , TU Wien, 1040, Vienna, Austria

Witnessing the recent rapid 5G commercialization with multiple advantages interms of high throughput, low latency, great serviceability, and extreme density,people look forward to seeing a new innovation in the next-generation mobilenetworks–6G. As a demand response, 6G additionally integrates artificialintelligence into all operational perspectives of the network while incorporating newinfrastructure to support service coverability to yet underserved areas. Since theartificial intelligence in 6G has been discussed significantly in the literature, thisarticle, on the other hand, provides major insights of the 6G aerial radio accessnetwork (ARAN) as an expansion of mobile coverage from the sky. First, we highlightthe distinct attributes of ARAN by constructing a comparative taxonomy amongmobile access infrastructures. Consequently, comprehensive ARAN architecture,reference model, and potential technologies are analyzed. Next, current researchtrends are investigated followed by future challenge discussions.

The research maturity and rapid commercializa-tion of the fifth generation (5G) have recentlyexposed multiple advantages for emerging

application scenarios. Three major service categoriessuch as enhanced mobile broadband, ultrareliable,and low-latency communications, and massivemachine type communications have benefited fromhigh throughput, low latency, great serviceability, andextreme density performance of the 5G networks.Although 5G has yielded an impressive success atimproving service experience for a massive number ofend-users, we have never stopped our efforts in fur-ther increasing and expanding its capabilities towardan innovative next-generation network: 6G. Based oninherited achievements from the predecessor, the 6G

networks are expected to target three foundationaldirections including i) ultrareal service experience, ii)intelligent networking and service provisioning plat-form, and iii) global mobile Internet coverability.1 Whilethe first two directions can be seen as evolutionarystrategies continuing current advanced features ofthe 5G networks as revealed from 6G literaturereview,2 the last direction is proposed to handlerecently emerging paradigms such as a new one–mul-tialtitude airborne transportation and the missingone–underserved/isolated terrestrial areas.

As a demand response, 6G introduces the aerialradio access network (ARAN) to complement itsaccess infrastructure for the aforementioned para-digms.3 The ARAN involves airborne objects equippedwith mobile transceiver antennas, a.k.a. aerial basestations (ABSs), to provide a radio access mediumfrom the sky to end-users for Internet service connec-tivity. Typical ABS consists of the low-altitude plat-forms (LAPs) and the high-altitude platforms (HAPs)such as drones, unmanned aerial vehicles (UAVs),

1089-7801 � 2020 IEEEDigital Object Identifier 10.1109/MIC.2020.3033976Date of current version 16 April 2021.

March/April 2021 Published by the IEEE Computer Society IEEE Internet Computing 93

Authorized licensed use limited to: TU Wien Bibliothek. Downloaded on May 03,2021 at 09:17:59 UTC from IEEE Xplore. Restrictions apply.

balloons, and flights. In ARAN, the fronthaul linksbetween ABSs and end- users are designed to adoptcommon modulation schemes at the frequencyspectrum assigned for mobile communications.Meanwhile, the backhaul links wirelessly interconnectthe ABSs to the core networks via either satelliteconstellation and ground station or terrestrial macrobase stations. The completely wireless infrastructuredesign provides ARAN with flexible deployment andquick adaptation abilities to serve diverse applicationparadigms such as search and rescue, aerial data har-vesting, and remote/isolated areas that are difficult tobe satisfactory by existing terrestrial infrastructure.

However, current proposals for aerial access haveonly focused on partially accommodating specificapplications; there is no complete and systematicaldesign. For instance, swarms of drones/UAVs arewidely utilized to temporarily assist communicationrelay, traffic alleviation, and sensory data harvesting interrestrial mobile networks such as postdisaster com-munication, aerial surveillance, and aerial crop moni-toring systems. While satellite constellation has beenexploited to provide users with expensive and limitedInternet access in underserved areas over the pastdecades. Fortunately, OneWeb (https://www.oneweb.world) and Startlink (https://www.starlink.com), twoambitious airborne mobile broadband projects, havebeen recently launching thousands of miniaturizedsatellites forming dense constellations at low Earthorbit (LEO) altitude to deliver Internet across the globecheaper and faster. Nevertheless, it is necessary toinvestigate these diverse and separate proposals andgather them into one comprehensive architecture, i.e.,ARAN.

Inspired from the above observation, this articleaims to clarify ARAN from multiple perspectives.First, we position ARAN in the 6G context through acomparative taxonomy among existing mobile accessinfrastructures within three categories: coveragearea, assisted technology, and architectural design.Consequently, the multitier architecture and refer-ence model are thoroughly analyzed for a comprehen-sive overview of ARAN with distinguished features.Then, potential technologies are deeply discussed toconcretize the advanced features of ARAN. Finally, weprovide statistical insights of recent ARAN researchoutcomes and draw future challenges.

AERIAL ACCESS INFRASTRUCTURETo be ready for a broad range of emerging paradigms,one of the efficient strategies in 6G is to expandcommunication vertically and horizontally forming a

ubiquitous 3-D mobile coverage.4 Here, we investigatetechnical aspects of the additional aerial access infra-structure with the aim of efficiently accommodatingemerging new aerial 6G communications.

ARAN PositioningAs the frontier of the network to interact with end-users, radio access networks (RANs) play a criticalrole in all mobile generations. Involving in the wholesystem maturity, mobile access infrastructure flexiblyintegrates recent advanced technologies and designconcepts into its organization. Accordingly, existingmobile access infrastructure can be classified intothree categories: architectural design, assistedtechnology, and coverage area, as demonstrated inFigure 1.

› Architectural design: This category regards tech-nical factors in the RAN design. In particular, thecognitive RAN improves frequency spectrum effi-ciency through vacant licensed/unlicensed spec-trum sensing mechanisms, the heterogeneousRAN jointly utilizes various access technologiesfor diverse users and services. Meanwhile, thevirtualized RAN softwarizes network functionsand operations in a serverless manner, and theultradense RAN exploits access point implemen-tation density to serve amassive number of usersconcurrently.

› Assisted technology: This category representsa utilization of advanced technologies to facili-tate user services with high performance. Forinstance, the cloud RAN introduces a high in-net-work computing capability at back-haul infra-structure for heavy offloading traffic, while thefog RAN provides low-latency response to userrequests at the edgewith localization. Fromotherperspectives, the blockchain RAN ensures asecure environment for sensitive user activitiesand data go through, while the intelligent RAN

FIGURE 1. Radio access network taxonomy from multiple

perspectives.

94 IEEE Internet Computing March/April 2021

INTERNET OF THINGS, PEOPLE, AND PROCESSES

Authorized licensed use limited to: TU Wien Bibliothek. Downloaded on May 03,2021 at 09:17:59 UTC from IEEE Xplore. Restrictions apply.

can learn user behaviors for service response per-sonality by integrating recent advanced artificialintelligent (AI) technology.

› Coverage area: This category considers RANbased on the coverage space. One of the mostwell-known systems in this category is the terres-trial RAN. The terrestrial RAN includes variousmobile and wireless access infrastructures toprovide user services on the ground such as cel-lular, radio, and WiFi systems. On the contrary,the satellite RAN incorporates satellite constella-tions into the access infrastructure to offer end-users Internet connection via satellite terminalstations on the global wide. In the middle, the fly-ing RAN defines access medium encompassedby LAPs equipped with mobile transceiver anten-nas to expand service coverage of the terrestrialinfrastructure. On the other hand, the aerialRAN (i.e., ARAN), as aforementioned, stands for aunified architecture harmonizing multitier aerialaccess infrastructures to support aerial usersand users at underserved areas.

The ARANs are distinguished from existing RANs infour areas. First, ARAN supplements existing accessinfrastructure with additional physical components(e.g., LAPs, HAPs, CubeSats, and ComSats). This com-plement is indispensable in the 6G context, where new

application scenarios are increasingly emerging butcurrent infrastructures fail to support them with highperformance and stable serviceability. Second, ARANhas a complete RAN architecture consisting of mobileaccess points (i.e., ABSs) at the fronthaul and trafficdistribution components at the backhaul to inter-connect with the core networks. The complete archi-tecture allows ARAN to operate independently of otheraccess infrastructures. Third, ARAN is definitely com-patible with additional assisted technologies (e.g.,cloudization, AI, and blockchain) and hybrid design con-cepts (e.g., cognitive radio, heterogeneous access, net-work virtualization, and access densification). Fourth,ARAN is a complete wireless architecture; in otherwords, the network components are dynamically inter-connected via time-varying wireless links. This propertyhelps ARAN with flexibility and quick adaptability toenvironmental changes and service requirements.

Multitier ARAN ArchitectureAs the aerial coverage is characterized by a 3-D mobilespace wherein users are distributed discretely, ARANarchitecture must adopt a multitier networking designto flexibly direct its communication resources fortargeted narrow service areas. Figure 2 illustrates anoverview of ARANs in the context of a complete user-core path. In detail, there are three tiers constituting

FIGURE 2. Overview of aerial radio access networks: Architecture and new emerging services.

March/April 2021 IEEE Internet Computing 95

INTERNET OF THINGS, PEOPLE, AND PROCESSES

Authorized licensed use limited to: TU Wien Bibliothek. Downloaded on May 03,2021 at 09:17:59 UTC from IEEE Xplore. Restrictions apply.

a typical ARAN; those are LAP communications atthe altitude of 0–10 Km, HAP communications at thealtitude of 20–50 Km, and LEO communications at thealtitude of 500–1500 Km above the sea level.3

In charge of front-end interfacing directly to end-users, LAP communication tier establishes swarms ofABSs (e.g., drones and UAVs) providing Internet serv-ices through wireless transmission technologies suchas 5G new radio and WiFi in the fronthaul. Dependingon application scenarios, ABS swarms can be orga-nized following different topologies including mesh,star, bus, chain, and hierarchical architectures. In allthe collaboration topologies, ARAN defines severalspecific ABSs acting as gateways to maintain connec-tion on the backhaul links to the core networks. Thereare two redundant backhaul links: one is via uppertiers of the ARAN (e.g., LEO satellites–ground station–core) and the other is via macro base stations of theterrestrial mobile networks (e.g., 5G gNB).

In the middle of the ARAN, HAP communicationtier develops a stable mesh of HAPs (e.g., aircraft andballoon) for an ultrawide coverage area. In the fron-thaul, the HAP mesh provides either direct accessinterface for end-users (here, HAPs are considered asABSs) or relay connection for LAP tiers to the corenetworks. In the backhaul, HAP communicationdevelops two redundant links to the core networks asthe LAP tier does. As specified by the InternationalTelecommunication Union (ITU) organization, 2 GHz, 6GHz, 27/31 GHz, and 47/48 GHz bands are assigned forHAP communications.5 Recently, some industrieshave successfully launched pilot projects to offerInternet connections to rural and remote areas at theHAP tier such as Loon (https://www.loon.com)and HAPSMobile (https://www.hapsmobile.com).

At the top of ARAN, LEO communication tier con-sists of multiple miniaturized satellite constellations

(e.g., CubeSats and ComSats) orbiting at LEO altitude.In particular, CubeSat deployment is designed for low-latency broadband Internet satellite (tens-of-mslatency at Mbps data rate) while ComSats aim at widecoverage with high serviceability, compared to othersatellite classes.6 Different from two lower tiers, LEOcommunications typically provide Internet service toend-users as well as LAP/HAP gateways via satelliteterminals (e.g., very small aperture terminals (VSATs))instead of a direct access. On the opposite side, allLEO satellites can connect to the core networksthrough dedicated ground stations. Ku, Ka, and Vbands are widely utilized on bidirectional satellite linksas recommended by the ITU. Recent years have wit-nessed emerging LEO satellite projects such as Star-link, OneWeb, and Telesat with thousands-of-satelliteconstellation launches.

Reference ModelThe reference model of ARAN is proposed by jointlyconsidering the 5G Release 167 and the satellite ter-restrial integrated network (STIN)8 standardized bythe third Generation Partnership Project (3GPP) orga-nization and European Telecommunications Stand-ards Institute (ETSI), respectively. Figure 3 shows thedetails of the ARAN reference model. Without loss ofgenerality, the model is designed to be adapted for acomplete integration into the recent 5G architecture.

From a functional perspective, it is observed thatLAP and HAP tiers share the same set of networkingoperations; hence, LAPs and HAPs use the samerepresentation–ABSs. Considered as a 5G point ofattachment, the ABSs utilize the Xn interface for theirpeer-to-peer communications. While the fronthauldefines NR Uu interfaces between ABSs and end-users(UE–user equipment), the backhaul connects ABSs tothe core through either 5G gNBs on the Xn interface

FIGURE 3. ARAN reference model adopting the 3GPP 5G Release 16 standards.

96 IEEE Internet Computing March/April 2021

INTERNET OF THINGS, PEOPLE, AND PROCESSES

Authorized licensed use limited to: TU Wien Bibliothek. Downloaded on May 03,2021 at 09:17:59 UTC from IEEE Xplore. Restrictions apply.

or VSATs on the Ymx interface, respectively. In 5G corenetworks, the user plane function (UPF) acts as a con-tact point bridging between internal mobile infrastruc-ture and the external networks (e.g., Internet andcontent providers). It is seen that the LAP/HAP tiersof ARAN are covered by the 3GPP 5G standard modelas internal parts using native interfaces; therefore,management protocols and resource orchestrationsharing with other 5G network functions are fullysupported.7

Different from the LAP/HAP tiers, the 3GPP 5Gstandard model considers the LEO tier as an externalsystem. For that, the ETSI TR 103.611 (V1.1.1) standard8

defines a seamless integration of satellite systemsinto existing 5G infrastructure. In particular, the exter-nal interface Ymx is used for interconnection betweenVSATs and 5G gNBs at the fronthaul. For the backhaullink, a gateway (GW) of the LEO systems, i.e., theground station, interacts with the UPF component ofthe core networks on the Ygw interface. Other linksamong VSATs, LEO satellites, and GWs use internalinterfaces specified by satellite standards.

In the control plane, the operational control mes-sages and data packets are delivered on the sameinterfaces but using different protocols. Meanwhile, inthe resource management plane, ARAN referencemodel mainly follows the 5G management and orches-tration (MANO) design, where the networking, com-puting, storage resources, and functional data areshared, exchanged, and collaborated among 5G net-work components.9 The resource management mes-sages are transferred on dedicated channels to/fromthe MANO systems.

POTENTIAL TECHNOLOGIESTo enable ARAN, several potential technologies arebeing applied to handle the distinct properties ofARAN architecture, especially the complete wirelessinfrastructure and various-altitude aerial communica-tion. Potential technologies include energy refills,intelligent beamforming, mobile cloudization, and traf-fic engineering. The details are described below.

Energy RefillsThe most important concern in ARAN is to replenishenergy for airborne components efficiently whileretaining the system stability. Although there are avariety of recharging solutions proposed in the litera-ture, it is widely acknowledged that radio frequency(RF) wireless charging and solar energy harvesting(EH) techniques10 are particularly appropriate owingto the ARAN features.

RF wireless charging is a far-field charging methodthat exploits electromagnetic radiation to transferenergy from the charging point to the receivers. Asthe RF wireless charging method can operate over along distance, it exposes (i) the dynamicity allowingboth charging points and receivers can move aroundduring the charging time and (ii) the simultaneityimplying multiple receivers can be powered at thesame time by one charging point, even under a non-line-of-sight transmission. Furthermore, to improvethe charging efficiency, directional antennas withintelligent beamforming technologies are imple-mented at the charging point to concentrate energytoward the targeted receivers. These advantagesconfirm the RF wireless charging method, an excel-lent candidate for energy refills in ARAN. Owing tothe great potential in many-field applications, RFwireless charging method has recently attracted vari-ous companies to develop and launch trial productsand services such as WiTricity(https://witricity.com),Ossia(https://www.ossia.com), and Energous Corpora-tion (https://www.energous.com).

On the other hand, solar EH is a method that useenergy-stored cells (e.g., photovoltaic) to capturesolar energy radiated from the sun, then convert theheat to electrical energy. Comparing to other sources(e.g., thermal, wind, and kinetic), solar energy hasbeen widely exploited to energize devices in variousreal-world applications because of the energy effi-ciency and reliability. Especially, the effectiveness ofintegrating solar energy harvester into power systemof aerial devices have been proved in many successstories such asNASA’sPathfinder(https://mars.nasa.gov/MPF/index1.html), Indian MARAAL (https://www.iitk.ac.in/aero/maraal), and PHASA-35 (https://www.baesystems.com/en/product/phasa-35). Obviously,solar-powered capability provides airborne compo-nents with the potential to fly for days and even for-ever during their operational life cycles. Finally yetimportantly, a hybrid solution between RF wirelesscharging and solar EH is definitely applicable forenergy refills in ARAN.

Intelligent BeamformingWith a note that end-user locations disperse geo-graphically in a wide 3-D space, efficiently configuringcommunication resources to focus on desired end-user(s) is critical for ARAN communication. In this cir-cumstance, intelligent beamforming is a potentialtechnology to resolve the requirement.11 Here, the dis-tribution of end-user locations in both time and spacedomains is learnt based on user behaviors and traffic

March/April 2021 IEEE Internet Computing 97

INTERNET OF THINGS, PEOPLE, AND PROCESSES

Authorized licensed use limited to: TU Wien Bibliothek. Downloaded on May 03,2021 at 09:17:59 UTC from IEEE Xplore. Restrictions apply.

changes. Consequently, communication requirementsof end-user(s) are predicted to drive radio resourceoptimization. Beam parameters such as activationstate, beam width, transmit power, azimuth, anddown-tilt are configured automatically to obtain anoptimal performance at certain narrow areas.

In addition, the utilization of dense distributedantenna array can significantly empower the intelli-gent beamforming capability with flexible transmitdirections on the surface of airborne components.12

Iteratively, the joint optimization model of antennaelements’ parameters and beam configurations areupdated to yield dynamic global adjustment for a max-imal serviceability. As a result, long-range low-powerwireless communication is natively featured on ARANfronthaul.

Mobile CloudizationIn-network computing service is amust-have capabilitythat was already specified as a native function in therecent 3GPP 5G standards7; therefore, ARAN cannotbe an exception. Moreover, the in-network computingservice is especially critical in ARAN to reduce the bur-dens of back-haul traffic over multihop wireless trans-mission. For that, mobile cloudization9 is the keyenabler, which aggregates computing resources fromall networking components on a virtual pool to opti-mally allocate these resources on demand. Imple-mented on the MANO system, collaboration betweenvirtual infrastructure managers and the central orches-trator tailors resources as requested by functional net-work operations as well as offloading services fromusers. The flexibility of mobile cloudization optimallyexploits distributed computing resources of airbornecomponents for advanced technologies such as rein-forcement learning and network slicing within aprompt response. As user requirements prioritizewhether performance or latency, appropriate comput-ing tiers (e.g., edge, fog, and cloud) can offer in-networkcomputing services with satisfaction, accordingly.

Traffic EngineeringAs ARAN is constituted by mesh communicationamong airborne components and hierarchical commu-nication among tiers, the complexity of such an archi-tecture makes traffic engineering essential to beoptimized globally. While network softwarization bringssoftware-defined networking ability for ARAN toabstract the physical infrastructure, the gathered net-work knowledge at the central controller is a greatsource to be exploited for traffic patternization andprediction.13 According to the traffic pattern, volume,

and location, whether intratier (mesh connection) orintertier (hierarchical connection) traffic engineeringmechanisms are implemented. To improve the systemstability and serviceability in ARAN, load balancing andredundancy are typically setup as the objectives tooptimize routing decision with latency commitments.A good traffic engineering mechanism facilitates userdata delivery with not only reliability but also securityand privacy against a wide variety of threats.

FUTURE RESEARCHFor ARAN maturity, a comprehensive knowledge ofcurrent trends and open challenges is necessary todirect the future research.

Current TrendsTo have a historical view of ARAN attraction in theresearch community, we investigated related studiesin three major publication databases: the Associationfor Computing Machinery Digital Library (ACM DL–https://dl.acm.org), the Institute of Electrical andElectronics Engineers Xplore (IEEE Xplore–https://ieeexplore.ieee.org), and the Elsevier Scopus (Scopus–https://www.scopus.com). The investigation was per-formed on September 30, 2020. To derive the statisticalinformation from the databases, we configured thequery syntax on Document title using keyword set of{drone, unmanned aerial vehicle, UAV, low altitude plat-form, LAP, high altitude platform, HAP, LEO satellite,CubeSat} during the publication date from 1/1/2010 to9/30/2020. Results are demonstrated in Figure 4. Basedon the given syntax configuration, the ACM DL, IEEEXplore, and Scopus returned 521, 12153, and 34726appropriate results, respectively. Since the year 2020 isnot complete, Figure 4(a) plots the observation inrecent 10 years (2010, 2019) with a note that there are4401 publications from January to September in 2020.According to the ACM DL database, the number ofpublications increases 6.03 times after 10 years, from1172 publications in 2010 to 7072 publications in 2019within an exponential trending. The statistical insightshows an expectation that ARAN is retaining attractivemegatrend to research community in the next years.

To discover recent research objectives in ARAN,individual keyword (and its variables) in the set {trajec-tory, energy, computing, routing, security, throughput,location, latency} is iteratively used to filter theobserved results above. The distribution of researchobjectives is plotted in Figure 4(b). In particular,Trajectory topic dominates the publications withapproximate 29% while the second place is for Energytopic with 21%. Here, Trajectory stands for mobility

98 IEEE Internet Computing March/April 2021

INTERNET OF THINGS, PEOPLE, AND PROCESSES

Authorized licensed use limited to: TU Wien Bibliothek. Downloaded on May 03,2021 at 09:17:59 UTC from IEEE Xplore. Restrictions apply.

and topology designs and Energy regards energy con-sumption and recharging scheme solutions. Together,the Trajectory and Energy concerns attract the mostattention from the research community wherein a halfof the studies have been dedicated to resolve theirissues. Next, Computing and Routing share approxi-mately a quarter of publications with respectively 12%and 11%. These topics represent the mobile cloudiza-tion and traffic engineering concerns in ARAN. Finally,Security, Throughput, Location, and Latency completethe research attractiveness with 6–8% for each.

Open ChallengesOpen challenges derived from the above researchtrend analyses are discussed to draw the future stud-ies in ARAN. First, although the mobility and flexibilityare auspicious capabilities of ARAN, they lead to chal-lenging designs for the network to quickly adapt tothe environmental and operational changes. As in anaerial wireless environment, the movement of individ-ual airborne components directly affects communica-tion states of adjacent components as well as thewhole network topology. Therefore, optimizing the tra-jectory is a foundational factor for the ARAN systemstability.

Second, energy efficiency must be in the focusbecause of the battery capacity limitation of airbornecomponents. In particular, several aspects of energyrefill techniques can be further improved such as charg-ing distance, charging time, and energy transfer rate. Onthe other hand, energy consumption in ARANoperation,communication, and computation should be jointlyminimizedwhen optimizing the systemperformance.

Third, computing resource constraints of airbornecomponents limit ARAN’s capability to provide high-performance in-network computing services to heavyuser applications. Although the mobile cloudization

introduces a promising platform to optimize comput-ing resource abstraction and orchestration, improvingthe global cloudization mechanism subject to variousenvironmental factors and diverse user requirementsstill remains challenges to resolve.

Fourth, since ARAN operates in multiple altitudesto support different classes of user services, multiob-jective communications should be satisfied to flexiblyallow various requirements. To this end, reliable/low-latency and long-range/low-power properties are con-sidered two critical metric pairs of an efficient trafficengineering mechanism for high serviceability inARAN. The former is necessary to support mobilebroadband users with high quality-of-service while thelatter is vital for IoT sensory data harvesting applica-tions. Nevertheless, long transmission distance in thefronthaul links is one of the most difficult obstaclestoward user satisfaction in ARAN.

Fifth, as ARAN architecture is built from a grid ofwireless link interconnections, this open infrastructureis vulnerable to a variety of common cyberattackssuch as packet sniffing, signal jamming, man-in-the-middle, and denial-of- service. In addition, multitiercommunications lead to multiple protocols used inARAN; hence, it is highly possible to be exploited byfault injection attacks. On the other hand, user dataharvesting and offloading traffic are recommended touse in-network computing service in ARAN to process.In this circumstance, privacy and integrity are themost concerns to protect essential user information.

Sixth, owing to the distinct characteristics ofARAN, existing evaluation frameworks may not accu-rately configure multialtitude aerial wireless environ-ment as in certain application scenarios. In particular,such a simulation-based evaluation typically simplifiesenvironmental changes as well as mechanical con-straints of airborne components resulting in reality

FIGURE 4. Research trends in ARAN-related publications from 2010 to September 30, 2020.

March/April 2021 IEEE Internet Computing 99

INTERNET OF THINGS, PEOPLE, AND PROCESSES

Authorized licensed use limited to: TU Wien Bibliothek. Downloaded on May 03,2021 at 09:17:59 UTC from IEEE Xplore. Restrictions apply.

reduction. Therefore, a dedicated ARAN emulationplatform is essential for researchers to precisely vali-date their proposed solutions regarding performancemetrics and applicability evaluation.

CONCLUSIONIn this article, we provided a comprehensive overviewof ARAN in the context of 6G development. Majoraspects of the ARAN were discussed including thenetwork identification, system architecture, referencemodel, potential technologies, emerging trends, andfuture research. As the ARAN is a vital complement in6G access infrastructure, this article is considered tobe a convenient facility for the readers to quicklytouch on current states and future visions of ARAN.Furthermore, through the detailed technical investiga-tion as well as pros and cons analyses, this articleis expected to engage multidisciplinary research tointegrate and exploit the ARAN in various fields.

ACKNOWLEDGMENTSThis work was supported by the National ResearchFoundation of Korea (NRF) grant funded by theKorean government (MSIT) under Grant NRF-2021R1G1A1008105.

REFERENCES1. X. Qiao, Y. Huang, S. Dustdar, and J. Chen, “6G vision:

An AI- driven decentralized network and service

architecture,” IEEE Int. Comput., vol. 24, no. 4, pp. 33–

40, Jul./Aug. 2020.

2. G. Gui, M. Liu, F. Tang, N. Kato, and F. Adachi, “6G:

Opening new horizons for integration of comfort,

security and intelligence,” IEEE Wireless Commun.,

vol. 27, no. 5, pp. 126–132, Oct. 2020.

3. L. Song, Z. Han, and B. Di, “Aerial access networks for

6G: From UAV, HAP, to satellite communication

networks,” in Proc. IEEE Int. Conf. Commun., Virtual

Program, 2020, pp. 1–1.

4. Z. Zhang et al., “6G wireless networks: Vision,

requirements, architecture, and key technologies,” IEEE

Veh. Technol. Mag., vol. 14, no. 3, pp. 28–41, Sep. 2019.

5. S. C. Arum, D. Grace, and P. D. Mitchell, “A review of

wireless communication using high-altitude platforms

for extended coverage and capacity,” Comput.

Commun., vol. 157, no. 1, pp. 232–256, 2020.

6. Y. Su, Y. Liu, Y. Zhou, J. Yuan, H. Cao, and J. Shi,

“Broadband LEO satellite communications:

Architectures and key technologies,” IEEE Wireless

Commun., vol. 26, no. 2, pp. 55–61, Apr. 2019.

7. 3GPP TS 23.501 V16.5.0 Release 16, “5G; System

architecture for the 5G system (5GS),” Jul. 2020.

8. ETSI TR 103 611 V1.1.1, “Satellite Earth stations and

systems (SES); Seamless integration of satellite and/or

HAPS (high altitude platform station) systems into 5G

and related architecture options,” Jun. 2020.

9. N.-N. Dao, W. Na, and S. Cho, “Mobile cloudization

storytelling: Current issues from an optimization

perspective,” IEEE Int. Comput., vol. 24, no. 1, pp. 39–47,

Jan./Feb. 2020.

10. J. Huang, Y. Zhou, Z. Ning, and H. Gharavi, “Wireless

power transfer and energy harvesting: Current status

and future prospects,” IEEE Wireless Commun., vol. 26,

no. 4, pp. 163–169, Aug. 2019.

11. Y. Huang, X. Xu, N. Li, H. Ding, and X. Tang, “Prospect of

5G intelligent networks,” IEEE Wireless Commun.,

vol. 27, no. 4, pp. 4/5, Aug. 2020.

12. Y. Li and G. Amarasuriya, “NOMA-aided massive MIMO

downlink with distributed antenna arrays,” in Proc. IEEE

Int. Conf. Commun., 2019, pp. 1–7.

13. Q. Zhang, X. Wang, J. Lv, and M. Huang, “Intelligent

content-aware traffic engineering for SDN: An AI-driven

approach,” IEEE Netw., vol. 34, no. 3, pp. 186–193, May/

Jun. 2020.

NHU-NGOC DAO is an Assistant Professor with the

Department of Computer Science and Engineering, Sejong

University, Seoul, South Korea. His research interests include

network softwarization, mobile cloudization, and the Internet

of Things. Contact him at nndao@sejong.ac.kr.

QUOC-VIET PHAM is a Research Professor with the Research

Institute of Computer, Information and Communication,

Pusan National University, Busan, South Korea. His research

interests include network optimization, mobile edge/cloud

computing, and resource allocation for 5G wireless networks

and beyond. Contact him at vietpq@pusan.ac.kr.

DINH-THUAN DO is an Assistant Professor with the Depart-

ment of Computer Science and Information Engineering, Asia

University, Taichung, Taiwan. His research interests include

signal processing in wireless communications networks,

NOMA, full-duplex transmission, and energy harvesting. Con-

tact him at dodinhthuan@asia.edu.tw.

SCHAHRAM DUSTDAR is a Full Professor of Computer Sci-

ence (Informatics) with a focus on Internet Technologies head-

ing the Distributed Systems Group, TU Wien, Vienna, Austria.

He is a member of the Academia Europaea: The Academy of

Europe. He is a Fellow of the IEEE. He is the corresponding

author of this article. Contact him at dustdar@dsg.tuwien.ac.at.

100 IEEE Internet Computing March/April 2021

INTERNET OF THINGS, PEOPLE, AND PROCESSES

Authorized licensed use limited to: TU Wien Bibliothek. Downloaded on May 03,2021 at 09:17:59 UTC from IEEE Xplore. Restrictions apply.