IEEEAcceSSSPECIAL SECTION ON GREENINTERNETOF THINGSaaidshlany:BReceived November 15,2019, accepted November 29,2019, date of publication December 4, 2019dateofcurrentversionDecember18,2019Digital Objecr Identifier 10. 1109/ACCESS.2019.2957648ASurveyonGreen6GNetwork:Architecture and TechnologiesTONGYI HUANG1,WUYANG2,JUN WU1,(Member,IEEE)JINMA',XIAOFEI ZHANG3,ANDDAOYINZHANG3'School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China2Information Security Research Center, Harbin Engincering University, Harbin 150001, China3State Grid Electric Power Research Institute,Nanjing 211000, ChinaCorresponding authors: Wu Yang (yangwu@hrbeu.edu.cn)and Jun Wu (junwuhn@sjtu.edu.cn)This work was supported in part by the National Natural Science Foundation of China under Grant 61831007 and Grant 61972255, in partby the State Key Laboratory of Smart Grid Protection and Control under Grant SGNROOOOGZJS1808084, and in part by the Research onKey Technologies of Security Vulnerabilities and Risk Experiment Capabilities of State Grid Headquarters Science and TechnologyProjects.:ABSTRACT While 5G is being commercialized worldwide, research institutions around the world havestarted to look beyond 5G and 6G is expected to evolve into green networks, which deliver high Qualityof Service and energy efficiency.To meet the demands of future applications,significant improvementsneed to be made in mobile network architecture.We envision 6G undergoing unprecedented breakthroughand integrating traditional terrestrialmobile networks with emerging space,aerial and underwater networksto provide anytime anywhere network access.This paperpresents a detailed survey on wireless evolutiontowards 6G networks.In this survey,the primefocus is on the newarchitectural changes associated with 6Gnetworks,characterized by ubiquitous 3D coverage,introduction of pervasive AI and enhanced networkprotocol stack.Along with this, we discuss related potential technologies that are helpful in formingsustainableand socially seamless networks,encompassing terahertz and visiblelight communication,newcommunication paradigm,blockchain and symbioticradio.Our work aims to provide enlighteningguidancefor subsequentresearch of green 6G.:INDEX TERMS 6G,architecture,green networks, VLC, blockchain.LINTRODUCTIONacomplete6Gecosystem[3].TheU.K.andGermangov-With the completion of the first full set of 5G standards,ernments have invested in some potential technologiesforthe initial commercial deploymentof 5Gwireless networks6G such as quantum technology,and theUnited Statesbeganhas begun in 2019.5G wireless network marks the begin-research on terahertz-based 6Gmobile networks.The Min-ning of a true digital society and achieves significant break-ister of Industry and Information Technologyin China hasmadetheofficial pronouncementthatthecountryhas focusedthroughsintermsoflatency,datarates.mobilityandnumberof connected devices in contrast to previous generations [1]onthedevelopmentof6GLooking back at the evolution of mobile communication,Novel service requirements and scale increases are theit takes about onedecade from the initial concept researchdrivingforcebehindtheevolutionofwirelessnetwork.Theto the commercial deployment, while its subsequent usagerapid development of emerging applications results in alasts for at least another 10 years. That is, when the previ-never-ending growth in mobile data traffic.According to theforecast byInternational Telecommunication Union (ITU)ousgenerationmobilenetworkentersthecommercialphase.the next generation begins concept research.As 5G is in theglobal mobiledatatrafficwill reach5zettabytesby2030initial stages of commercialization,nowis the righttime to[4], as shown in Fig.1. Upcoming applications (e.g.e-healthlaunch research on 5G's successor.andautonomousdriving)havemorestringentrequirementsIn the past few years, some countries have issued stratefor latencyand throughput,which will eventuallyexceed thegic plans for the development of 6G [2]. In 2018, Finlandlimits of 5Gnetworks.It is expected that5Gwill reach itsannounced the 6Genesis Flagship program, an eight-yearlimitsinadecadeorsoandtomeetthesedemands.themainprogram with the overall volume of s290 million to developtechnical objectivesfor 6Gnetworks will be. Ultra-high data rate (up to ITbps) and ultra-low latency.The associate editor coordinating the review of this manuscript andapproving it for publication was Zhenyu Zhou.High energy efficiency for resource-constraineddevices.This work is licensed under a Creative Commons Attribetion 4.0 License.For more information,seehttp://creativecommons.org/licenses/by/4.0/175758VOLUME7,2019
SPECIAL SECTION ON GREEN INTERNET OF THINGS Received November 15, 2019, accepted November 29, 2019, date of publication December 4, 2019, date of current version December 18, 2019. Digital Object Identifier 10.1109/ACCESS.2019.2957648 A Survey on Green 6G Network: Architecture and Technologies TONGYI HUANG 1 , WU YANG2 , JUN WU 1 , (Member, IEEE), JIN MA1 , XIAOFEI ZHANG3 , AND DAOYIN ZHANG3 1School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China 2 Information Security Research Center, Harbin Engineering University, Harbin 150001, China 3State Grid Electric Power Research Institute, Nanjing 211000, China Corresponding authors: Wu Yang (yangwu@hrbeu.edu.cn) and Jun Wu (junwuhn@sjtu.edu.cn) This work was supported in part by the National Natural Science Foundation of China under Grant 61831007 and Grant 61972255, in part by the State Key Laboratory of Smart Grid Protection and Control under Grant SGNROOOOGZJS1808084, and in part by the Research on Key Technologies of Security Vulnerabilities and Risk Experiment Capabilities of State Grid Headquarters Science and Technology Projects. ABSTRACT While 5G is being commercialized worldwide, research institutions around the world have started to look beyond 5G and 6G is expected to evolve into green networks, which deliver high Quality of Service and energy efficiency. To meet the demands of future applications, significant improvements need to be made in mobile network architecture. We envision 6G undergoing unprecedented breakthrough and integrating traditional terrestrial mobile networks with emerging space, aerial and underwater networks to provide anytime anywhere network access. This paper presents a detailed survey on wireless evolution towards 6G networks. In this survey, the prime focus is on the new architectural changes associated with 6G networks, characterized by ubiquitous 3D coverage, introduction of pervasive AI and enhanced network protocol stack. Along with this, we discuss related potential technologies that are helpful in forming sustainable and socially seamless networks, encompassing terahertz and visible light communication, new communication paradigm, blockchain and symbiotic radio. Our work aims to provide enlightening guidance for subsequent research of green 6G. INDEX TERMS 6G, architecture, green networks, VLC, blockchain. I. INTRODUCTION With the completion of the first full set of 5G standards, the initial commercial deployment of 5G wireless networks has begun in 2019. 5G wireless network marks the beginning of a true digital society and achieves significant breakthroughs in terms of latency, data rates, mobility and number of connected devices in contrast to previous generations [1]. Looking back at the evolution of mobile communication, it takes about one decade from the initial concept research to the commercial deployment, while its subsequent usage lasts for at least another 10 years. That is, when the previous generation mobile network enters the commercial phase, the next generation begins concept research. As 5G is in the initial stages of commercialization, now is the right time to launch research on 5G’s successor. In the past few years, some countries have issued strategic plans for the development of 6G [2]. In 2018, Finland announced the 6Genesis Flagship program, an eight-year program with the overall volume of $290 million to develop The associate editor coordinating the review of this manuscript and approving it for publication was Zhenyu Zhou . a complete 6G ecosystem [3]. The U.K. and German governments have invested in some potential technologies for 6G such as quantum technology, and the United States began research on terahertz-based 6G mobile networks. The Minister of Industry and Information Technology in China has made the official pronouncement that the country has focused on the development of 6G. Novel service requirements and scale increases are the driving force behind the evolution of wireless network. The rapid development of emerging applications results in a never-ending growth in mobile data traffic. According to the forecast by International Telecommunication Union (ITU), global mobile data traffic will reach 5 zettabytes by 2030 [4], as shown in Fig. 1. Upcoming applications (e.g. e-health and autonomous driving) have more stringent requirements for latency and throughput, which will eventually exceed the limits of 5G networks. It is expected that 5G will reach its limits in a decade or so and to meet these demands, the main technical objectives for 6G networks will be • Ultra-high data rate (up to 1Tbps) and ultra-low latency. • High energy efficiency for resource-constrained devices. 175758 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/ VOLUME 7, 2019
IEEEAcceSST. Huang et al: Survey on Green 6G Network: Architecture and TechnologiesTraffic/monthA.FROM1GTO3G6000The first generation mobile network was introduced in the50150001980s,which was designedfor voiceservices,with adata rateup to2.4kbps.It used analog signal totransmit information3327and there was no universal wireless standard, leading to many194Adrawbacks such asproblematichand-off,lowtransmission2000efficiency and no security [6]. Compared to first-generation1000systems, 2G was based on digital modulation technologies5791158249such as Time Division Multiple Access(TDMA) and Code20202021202220232024202520262027202820292030Division Multiple Access(CDMA). It has a data rate up tonon-M2MtrafficM2Mtrafic64kbps,supporting notonlybettervoiceservices,butalsoFIGURE 1.Global mobile data traffic in 2020-2030 forecast by ITU.services like Short Message Service (SMS).The all-dominantmobile communication standard in the 2G era was theGSM(Global System for MobileCommunication)[7].The third.Ubiquitous global network coverage.generation was proposed in 2000 with the goal of offering.Trusted and intelligent connectivity across the wholehigh-speed datatransmission.3G network provides a datanetwork.transfer rate of at least 2 Mbps as well as high speed accessto Internet [8].It enables advanced services not supported byIn this paper, we discuss some emerging ideas about1Gand2G networks,includingWebbrowsing,TVstreaming.potential architecture and technologies of 6G. The rest ofnavigational maps and video services. In order to achievethe paper is organized as follows. Section II presents theglobal roaming, an organization called 3rd Generation Part-evolution of mobile communication networks. Section IIInership Project (3GPP) was established to define technicalgives the detailed description of architectural changes of 6G.specifications and continue the work by defining mobileSectionIV provides a brief overview of some visionarytech-standardsand systems[9]nologies that may be key parts of 6G. Finally, this paper isconcludedin Section V.B.4GII.EVOLUTIONOFMOBILECOMMUNICATIONNETWORK4G is an all IP based network introduced in the lateThere has been a phenomenal advancement inmobile com-2000s,which is capable of providing high-speed datamunication network since the first emergence of analogrates up to1Gbits/s inthe downlink and500Mbits/scommunications networkinthe198Os.Thisadvancementin the uplink. It apparently improves spectral efficiencyis not a one-step process, but consists of several generaand reduces latency,accommodating the requirements settions which have different standards, capacities and tech-by advanced applications like Digital Video Broadcasting(DVB), High Definition TV content and video chat. More-niques.New generation have been introduced nearly everyten years [5]. The evolution of mobile network is shownover, 4G enables terminal mobility to provide wireless ser-in Fig. 2.vices at anytime and anywhere, through automatic roaming4BlockchainTerahertzonnona~TbpsBDMASDN6GMm-WaveWiMAX~20GbpsUMTSLTE-ATD-SCDMA5GTDMA<2GbpsCDMA2000GSMWCDMA4GEDGEFDMA2-100MbpsGRPSAnalog3G<64Kbps草<2.4KbpsP2G好1GED国YHNCMobile InternetFullyintelligentandWeblloTSMSVoice serviceofapplicationscoveredconnection198019902000201020202030YearsFIGURE 2. Evolution of mobile wireless systems.175759VOLUME 7, 2019
T. Huang et al.: Survey on Green 6G Network: Architecture and Technologies FIGURE 1. Global mobile data traffic in 2020-2030 forecast by ITU. • Ubiquitous global network coverage. • Trusted and intelligent connectivity across the whole network. In this paper, we discuss some emerging ideas about potential architecture and technologies of 6G. The rest of the paper is organized as follows. Section II presents the evolution of mobile communication networks. Section III gives the detailed description of architectural changes of 6G. Section IV provides a brief overview of some visionary technologies that may be key parts of 6G. Finally, this paper is concluded in Section V. II. EVOLUTION OF MOBILE COMMUNICATION NETWORK There has been a phenomenal advancement in mobile communication network since the first emergence of analog communications network in the 1980s. This advancement is not a one-step process, but consists of several generations which have different standards, capacities and techniques. New generation have been introduced nearly every ten years [5]. The evolution of mobile network is shown in Fig. 2. A. FROM 1G TO 3G The first generation mobile network was introduced in the 1980s, which was designed for voice services, with a data rate up to 2.4 kbps. It used analog signal to transmit information and there was no universal wireless standard, leading to many drawbacks such as problematic hand-off, low transmission efficiency and no security [6]. Compared to first-generation systems, 2G was based on digital modulation technologies such as Time Division Multiple Access(TDMA) and Code Division Multiple Access(CDMA). It has a data rate up to 64kbps, supporting not only better voice services, but also services like Short Message Service (SMS). The all-dominant mobile communication standard in the 2G era was the GSM (Global System for Mobile Communication) [7]. The third generation was proposed in 2000 with the goal of offering high-speed data transmission. 3G network provides a data transfer rate of at least 2 Mbps as well as high speed access to Internet [8]. It enables advanced services not supported by 1G and 2G networks, including Web browsing, TV streaming, navigational maps and video services. In order to achieve global roaming, an organization called 3rd Generation Partnership Project (3GPP) was established to define technical specifications and continue the work by defining mobile standards and systems [9]. B. 4G 4G is an all IP based network introduced in the late 2000s, which is capable of providing high-speed data rates up to 1Gbits/s in the downlink and 500Mbits/s in the uplink. It apparently improves spectral efficiency and reduces latency, accommodating the requirements set by advanced applications like Digital Video Broadcasting (DVB), High Definition TV content and video chat. Moreover, 4G enables terminal mobility to provide wireless services at anytime and anywhere, through automatic roaming FIGURE 2. Evolution of mobile wireless systems. VOLUME 7, 2019 175759
IEEEAcceSST. Huang et al.: Survey on Green 6G Network: Architecture and Technologiesacross geographic boundaries of wireless networks. Longthroughput, ultra-low latency and reliability. ThereforeTerm Evolution-Advanced (LTE-A) and Wireless Interoper-forward-lookingresearchon futurenetwork frameworks isnecessary. FG NET-2030 has established Sub-Group 3 toabilityforMicrowaveAccess(WiMAX)areconsideredas4G standards [10].LTE integrates existing and newtech-formulate architectureof Network2030.However,it is unre-nologies suchas coordinatedmultipletransmission/receptionalistic to accurately illustrate what the future network archi-(CoMP), multiple-input multiple-output (MIMO), orthogo-tecturewill be.Sub-Group3has reached a compromisenal frequency division multiplexing (OFDM).interpreting the architecture from different dimensions ratherthan defining a unified framework. In this section, we intro-C.5Gducethe architectural changes associated with 6G from threeThe fifth generation mobile communication network hasdimensions, as shown in Fig. 3.almost completed the initial basictests,hardwarefacilitiesconstruction and standardization process,and will soon be putA.FROMTERRESTRIALTOUBIQUITOUS3DCOVERAGEintocommercial use.Thegoal of 5G is tomakerevolutionaryOne target of the next generation network architecture is toadvances in data rates, latency, network reliability, energyexpandthebreadth and depthof communication coverageefficiency,and massive connectivity[11].It not onlyusesThe current network architecture based on legacy terres-the new spectrum of the microwave band (3.3-4.2 GHz),trialcellularinfrastructurehasthefollowingtwodrawbacks:but also innovatively uses themillimeter-wave band for theinability to meet the high-altitude and deep-sea commu-first time, greatly increasing data rates (up to 10 Gbps).nication scenarios.which is an inevitable requirementfor5G applies advanced accesstechnologies,includingBeamfuture services; prohibitively expensive provisioning cost forDivisionMultiple Access (BDMA)and Filter Bank multidense cellular networks to provide connectivity in theglobalcarrier (FBMC). Many emerging technologies are integratedscale.In order to cover the above drawbacks, 6G will inte-into5Gtoimprovenetworkperformance:MassiveMIMOgrate non-terrestrial networks to provide full wireless cover-forcapacityincreaseSoftwareDefinedNetworks(SDN)forage [16].Preliminary envision about Space-Air-Ground-Seaflexibility in network,device-to-device(D2D)for spectralintegrated communication has been discussed in [17]efficiency.InformationCentricNetworking(ICN)forreduction in network traffic and network slicing for quickdeploy-1)SPACENETWORK:LEOSATELLITESYSTEMment of various services [12]-[14]. IMT 2020 proposedHigh throughput satellite (HTS) systems are capable ofthreemajor 5G usage scenarios:Enhanced mobile broad-broadband Internet access service comparableto terrestrialband(eMBB),Ultra-reliableandlowlatencycommunica-services interms of pricing and bandwidth.Most commu-tions (URLLC)and Massivemachinetype communicationsnications satellites arein geostationaryorbit(GEO)atan(mMTC).altitudeof35.786km,naturallyleadingtoexcessivedelayand infeasibility of integration with terrestrial mobile net-D.VISIONOFGREEN6Gwork.Non-geostationary orbit(NGSO)satellitesystemisAs5Gisenteringthecommercialdeploymentphase,researchproposed to provide low-latency,high-bitrateglobal Internetinstitutions around the world have begun to pay attention toconnectivity and several satellite constellations areabout to6G, which is considered to be deployed in about 2030. Greenbegin commercialization:6G is expectedto enhancetheperformanceof information: Starlink: American company SpaceX plans to launchtransmission - peak data rates up to 1 Tbps and ultra-lowStarlink,a constellationof 4.425lowEarthorbitLEO)latency inmicroseconds.Itfeaturesterahertzfrequency com-satellites and 7518 VLEO satellites in approximatelymunication and spatial multiplexing,providing as muchas340km orbits.The plan was authorized by theFederal1000timeshighercapacitythan5Gnetworks.Onegoal of6GCommunicationsCommission(FCC)[18]andwillbeis to achieve ubiquitous connectivity by integrating satellitefullydeployed in2027.communication networksandunderwater communicationsto.OneWeb:OnFebruary27,2019,OneWeb successfullyprovideglobal coverage[19].Energy harvestingtechnologieslauncheditsfirst six satelliteintoorbits.Theconstel-and the use of new materials will greatly improve the systemlation consists of 720LEO satellite[20] and hasgotenergy efficiency and realize sustainable green networks.authorization fromUK andFCCThreenew6Gserviceclassesweredescribedin[15]:ubiqui.Hongyan: China Aerospace Science and Technologytousmobileultrabroadband(uMUB),ultrahigh-speed-with-Corporation(CASC)willlaunchnineLEOsatellitesaslow-latencycommunications(uHSLLC)andultrahighdataa pilot demonstration for the Hongyan system, whichdensity (uHDD)ultimatelywillcomprise320satellitesandbecompletedby 2025.III.ARCHITECTURESOFGREEN6GNETWORKGreen 6G networks are expected to achieve energy-efficientAlthough there is still a long time before overall deploy-and socially seamless wireless connections in a global scope,mentof NGSOsatellite systems and convergenceof satellitewhile the existing network architecture is nuable to guar-communications and mobile wireless networks,advantagesof LEO satellite networks have been confirmed in theoryantee future application delivery constraints ultra-high175760VOLUME7,2019
T. Huang et al.: Survey on Green 6G Network: Architecture and Technologies across geographic boundaries of wireless networks. Long Term Evolution-Advanced (LTE-A) and Wireless Interoperability for Microwave Access (WiMAX) are considered as 4G standards [10]. LTE integrates existing and new technologies such as coordinated multiple transmission/reception (CoMP), multiple-input multiple-output (MIMO), orthogonal frequency division multiplexing (OFDM). C. 5G The fifth generation mobile communication network has almost completed the initial basic tests, hardware facilities construction and standardization process, and will soon be put into commercial use. The goal of 5G is to make revolutionary advances in data rates, latency, network reliability, energy efficiency, and massive connectivity [11]. It not only uses the new spectrum of the microwave band (3.3-4.2 GHz), but also innovatively uses the millimeter-wave band for the first time, greatly increasing data rates (up to 10 Gbps). 5G applies advanced access technologies, including Beam Division Multiple Access (BDMA) and Filter Bank multi carrier (FBMC). Many emerging technologies are integrated into 5G to improve network performance: Massive MIMO for capacity increase, Software Defined Networks (SDN) for flexibility in network, device-to-device (D2D) for spectral efficiency, Information Centric Networking (ICN) for reduction in network traffic and network slicing for quick deployment of various services [12]–[14]. IMT 2020 proposed three major 5G usage scenarios: Enhanced mobile broadband (eMBB), Ultra-reliable and low latency communications (URLLC) and Massive machine type communications (mMTC). D. VISION OF GREEN 6G As 5G is entering the commercial deployment phase, research institutions around the world have begun to pay attention to 6G, which is considered to be deployed in about 2030. Green 6G is expected to enhance the performance of information transmission - peak data rates up to 1 Tbps and ultra-low latency in microseconds. It features terahertz frequency communication and spatial multiplexing, providing as much as 1000 times higher capacity than 5G networks. One goal of 6G is to achieve ubiquitous connectivity by integrating satellite communication networks and underwater communications to provide global coverage [19]. Energy harvesting technologies and the use of new materials will greatly improve the system energy efficiency and realize sustainable green networks. Three new 6G service classes were described in [15]: ubiquitous mobile ultrabroadband (uMUB), ultrahigh-speed-withlow-latency communications (uHSLLC) and ultrahigh data density (uHDD). III. ARCHITECTURES OF GREEN 6G NETWORK Green 6G networks are expected to achieve energy-efficient and socially seamless wireless connections in a global scope, while the existing network architecture is nuable to guarantee future application delivery constraints — ultra-high throughput, ultra-low latency and reliability. Therefore, forward-looking research on future network frameworks is necessary. FG NET-2030 has established Sub-Group 3 to formulate architecture of Network 2030. However, it is unrealistic to accurately illustrate what the future network architecture will be. Sub-Group 3 has reached a compromise — interpreting the architecture from different dimensions rather than defining a unified framework. In this section, we introduce the architectural changes associated with 6G from three dimensions, as shown in Fig. 3. A. FROM TERRESTRIAL TO UBIQUITOUS 3D COVERAGE One target of the next generation network architecture is to expand the breadth and depth of communication coverage. The current network architecture based on legacy terrestrial cellular infrastructure has the following two drawbacks: inability to meet the high-altitude and deep-sea communication scenarios, which is an inevitable requirement for future services; prohibitively expensive provisioning cost for dense cellular networks to provide connectivity in the global scale. In order to cover the above drawbacks, 6G will integrate non-terrestrial networks to provide full wireless coverage [16]. Preliminary envision about Space-Air-Ground-Sea integrated communication has been discussed in [17]. 1) SPACE NETWORK: LEO SATELLITE SYSTEM High throughput satellite (HTS) systems are capable of broadband Internet access service comparable to terrestrial services in terms of pricing and bandwidth. Most communications satellites are in geostationary orbit (GEO) at an altitude of 35,786 km, naturally leading to excessive delay and infeasibility of integration with terrestrial mobile network. Non-geostationary orbit (NGSO) satellite system is proposed to provide low-latency, high-bitrate global Internet connectivity and several satellite constellations are about to begin commercialization: • Starlink: American company SpaceX plans to launch Starlink, a constellation of 4,425 low Earth orbit LEO) satellites and 7518 VLEO satellites in approximately 340 km orbits. The plan was authorized by the Federal Communications Commission (FCC) [18] and will be fully deployed in 2027. • OneWeb: On February 27, 2019, OneWeb successfully launched its first six satellite into orbits. The constellation consists of 720 LEO satellite [20] and has got authorization from UK and FCC. • Hongyan: China Aerospace Science and Technology Corporation (CASC) will launch nine LEO satellites as a pilot demonstration for the Hongyan system, which ultimately will comprise 320 satellites and be completed by 2025. Although there is still a long time before overall deployment of NGSO satellite systems and convergence of satellite communications and mobile wireless networks, advantages of LEO satellite networks have been confirmed in theory 175760 VOLUME 7, 2019
IEEEAcceSST. Huang et al: Survey on Green 6G Network: Architecture and TechnologiesControl view:IntelligentconnectionDistributedArtificial IntelligenceReal-timeIntelligentEdge6G applicationsIntelligent RadioContent-driven routingGEOXQuantunLEOateNetworkrepeaterSatellitFSORrlintSpaceview:networkEnhancedStratificationQuantumHAPBlockchannelntchainAirInfrastructureAerial4BackhoneXMannetworkview:planeUAVLUbiquitousArtuida3D coverageVLCSateliDParTerrestrialStetonnetwork611THZ4commkAetrumacsPUnderseanetworkUaderseaRF/optical/aconensotFIGURE 3.Different dimensions of the architecture of green 6Gand simulation environment.LEO network with laser andan altitude of no more than several kilometers.Comparedradio frequency(RF)co-routing mechanism can provideto LAP,HAP networks arecapable of widercoverageandlowerlatency communicationsthanterrestrial optical fiberlonger endurance, but the advantages of HAP overlapwithnetworks when communication distances are greater thanLEO satellitenetworkto someextent.On theotherhand,about3000km[21].LAP networks based on unmanned aerial vehicle (UAV)Apotential architecture of the space-terrestrial integratedcan be swifter to deploy, more flexibly reconfigured tonetwork (STIN)has shown in [22].comprised of thebest suit the communication environment,and present betterspace-basedbackbonenetwork (SBN)ofGEO satellites andperformance in short-range communication[25].Besides,the inter-satellite links (ISLs)connecting them,terrestrial net-flying base stations like UAV can work as relay nodes inworks (TN)and space-based accessnetworks (SAN)of LEOlong-distance communication to promote the integration ofand medium earth orbit(MEO)satellites.SBNis capableofterrestrial and non-terrestrial networks.Thesefeatures makeUAV-based wireless network a potential integral componentextending coverage and ensuring reliable space-ground con-nectivity while SAN is essential for integration with terres-of next-generation mobile communication system.In[26]a fully integrated, multi-layer vertical architecture for 6Gtrial and HTS networks to support ubiquitous global wirelessaccess. Several emerging technologies are embraced to facil-network was presented, including heterogeneous terrestrialitate the integration of terrestrial and satellite networks.SDNnetworks,UAV-basedLAPs,HAPs,LEO and GEO satelliteand ICNhavebeen introduced into STIN with the advantagesnetworks.The most attracting characteristic of UAV wireless net-of flexiblenetwork control,efficientnetwork configurationand small requestdelay[23].Theperformance of Multipathwork is that it enables mobile communication in situationsTCP (MPTCP) was evaluated in [24] and the result indicateswhere there are heavily compromised infrastructures or eventhat MPTCP strategy improves throughput and provides unin-no infrastructures,especially in catastrophic and emergencyterrupted connections duringhandover.situation. UAV network has been applied to temporary emer-gency communication services, however, there are issues to2)AERIALNETWORK:FLEXIBLERELAYSERVICESbetackled before stable and reliable UAVnetwork can beAerial network can be broadly classified into two categories,introduced into common application scenarios [27].First ofall,energyefficiency is critical tolong-term network servicehigh altitude platforms (HAP)which generally operate in thestratosphere and low altitude platforms (LAP)typically atPropulsion and directional adjustment consume most of theVOLUME7,2019175761
T. Huang et al.: Survey on Green 6G Network: Architecture and Technologies FIGURE 3. Different dimensions of the architecture of green 6G. and simulation environment. LEO network with laser and radio frequency (RF) co-routing mechanism can provide lower latency communications than terrestrial optical fiber networks when communication distances are greater than about 3000 km [21]. A potential architecture of the space-terrestrial integrated network (STIN) has shown in [22], comprised of the space-based backbone network (SBN) of GEO satellites and the inter-satellite links (ISLs) connecting them, terrestrial networks (TN) and space-based access networks (SAN) of LEO and medium earth orbit (MEO) satellites. SBN is capable of extending coverage and ensuring reliable space-ground connectivity while SAN is essential for integration with terrestrial and HTS networks to support ubiquitous global wireless access. Several emerging technologies are embraced to facilitate the integration of terrestrial and satellite networks. SDN and ICN have been introduced into STIN with the advantages of flexible network control, efficient network configuration and small request delay [23]. The performance of Multipath TCP (MPTCP) was evaluated in [24] and the result indicates that MPTCP strategy improves throughput and provides uninterrupted connections during handover. 2) AERIAL NETWORK: FLEXIBLE RELAY SERVICES Aerial network can be broadly classified into two categories, high altitude platforms (HAP) which generally operate in the stratosphere and low altitude platforms (LAP) typically at an altitude of no more than several kilometers. Compared to LAP, HAP networks are capable of wider coverage and longer endurance, but the advantages of HAP overlap with LEO satellite network to some extent. On the other hand, LAP networks based on unmanned aerial vehicle (UAV) can be swifter to deploy, more flexibly reconfigured to best suit the communication environment, and present better performance in short-range communication [25]. Besides, flying base stations like UAV can work as relay nodes in long-distance communication to promote the integration of terrestrial and non-terrestrial networks. These features make UAV-based wireless network a potential integral component of next-generation mobile communication system. In [26], a fully integrated, multi-layer vertical architecture for 6G network was presented, including heterogeneous terrestrial networks, UAV-based LAPs, HAPs, LEO and GEO satellite networks. The most attracting characteristic of UAV wireless network is that it enables mobile communication in situations where there are heavily compromised infrastructures or even no infrastructures, especially in catastrophic and emergency situation. UAV network has been applied to temporary emergency communication services, however, there are issues to be tackled before stable and reliable UAV network can be introduced into common application scenarios [27]. First of all, energy efficiency is critical to long-term network service. Propulsion and directional adjustment consume most of the VOLUME 7, 2019 175761
IEEEAcceSST. Huang et al: Survey on Green 6G Network: Architecture and TechnologiesTABLE 1.Comparison of diffirent undersea wireless communication technologies.RFAcousticOpticalHighAttenuationLowestxturbidityData rates [34]~Mbps~Kbps~GbpsModerateHighLowLatency [34]Transmission distance<10m<100km<100mHighPowerconsumptionModerateLowenergyin UAV communications,thereforenew trajectoryAl-driven approach where intelligence will be an endogenousoptimizationandrouteplanning schemes areproposed andcharacteristic of 6G architecture.Initial intelligence is that a relatively isolated networksignificantly improve energy efficiency [28], [29]. In [30]a learning-based computing offloading approach was pre-entity can intelligently adjust the configuration based onsentedtoreduceenergyconsumption andtotalcost.Secondly,multiple predefined options in a different yet determinis-extreme weather condition should be considered in UAVtic manner [35], which is actually an implementation ofcommunication.Free spaceoptics(FSO)is introduced intoperceptual AI without the capability to respond to unin-backhaulframework where UAVstransmit informationviatended scenarios. As the network is evolving into an extremepoint-point FSO links, enabling UAV network to offer highcomplex and heterogeneous system because of diversifieddata rates in different weather condition [31].Thirdly, dueservice requirements and explosively growing number ofto frequent topology change caused by high-speed mobility,connected devices, a novel AI paradigm of self-aware, self-more advanced mobile ad hoc protocols are demanded. Theadaptive,self-interpretive and prescriptive networking ismuch needed [38]. It requires not only embedding intelli-authors in [32] proposed adaptivehybrid communication pro-tocols which outperform existingprotocols.gence across whole network,but also embedding the logicof AI into the network structure,in which perception andinference interact in a systematic way,eventually enabling3)UNDERSEANETWORKallnetworkcomponentstoautonomouslyconnectandcontrolThere is a lot of controversy about whether undersea networkwith the ability to recognize unexpected situations and adaptis able to become a part of the future 6G network [17]. Under-to them.The ultimate expectation of intelligent networks issea wireless communication mainly involves RE, acousticthe autonomous evolution of networks.Wehighlight threeand optical communication.The comparison between thekey enablers for intelligent network, as described below.above three communicationtechnologies is shown in Table1.Unpredictableand complexunderwater environmentleadsto1)REAL-TIMEINTELLIGENTEDGE（RTIE)intricate network deployment, severe signal attenuation andNext generation network will require the support of interac-physical damage to equipment [33], leaving plenty of issuesto beresolved.tive AI-powered services and some services like autonomousvehicles are sensitive to response latency,which needs tointeract intelligently with their environments in real timeB.TOWARDSINTELLIGENTNETWORKCentralized cloud AI dealing with static data is incapable ofArtificial intelligence (AI), more specifically machine learn-achieving such services and there is an urgent need for theing (ML), has attracted a lot of attention from industry andRTIE,where intelligentprediction,inference anddecision areacademia in recentyearsand initial intelligence has beenmade onlive data.Major academiaand industryhavebegunapplied to many aspects of 5G cellular networks [35],fromtodeveloptechnologiesand softwarecomponentsthatmeetphysical layer applications such as channel coding and estithe real-time requirements in collaborative research labs suchmation,toMAClayerapplicationssuchasmultipleaccess,as theBerkeleyRISELab[39].High-performancehardwaretonetworklayerapplications suchas resourceallocation andis another driving factor for RTE and a specialized real-timeerror correction.and etc.In addition,the combination ofAI processor has been designed in [40].artificial intelligence and edge computing proves to improveQuality of Experience andreduce costs[36].Edgelearn2)INTELLIGENTRADIO(IR)ing also provides new possibilities for the implementationIn contrast to deployed physical (PHY) layer with initial intel-of many applications, such as healthcare[37].However,the application of AI in 5G networks is limited to the opti-ligence,IR is a broader and deeper conception that separatesmizationoftraditional networkarchitectureanditisdifficulthardware and transceiver algorithms. It operates as a unifiedto fully realize the potential of AI in the 5G era since theframework where hardware capabilities are estimated and5G networkdid not take AI into account at the beginningtransceiver algorithms can dynamically configure themselvesof architecture design. To fulfill the vision of intelligentaccording to the hardware information.From the perspectiveof PHY layer,IR is able to access theavailable spectrum,network,the design of 6G architecture should considerpos-sibilities of Al in network comprehensively and follow ancontrol transmission powerand adjusttransmission protocols175762VOLUME7,2019
T. Huang et al.: Survey on Green 6G Network: Architecture and Technologies TABLE 1. Comparison of diffirent undersea wireless communication technologies. energy in UAV communications, therefore new trajectory optimization and route planning schemes are proposed and significantly improve energy efficiency [28], [29]. In [30], a learning-based computing offloading approach was presented to reduce energy consumption and total cost. Secondly, extreme weather condition should be considered in UAV communication. Free space optics (FSO) is introduced into backhaul framework where UAVs transmit information via point-point FSO links, enabling UAV network to offer high data rates in different weather condition [31]. Thirdly, due to frequent topology change caused by high-speed mobility, more advanced mobile ad hoc protocols are demanded. The authors in [32] proposed adaptive hybrid communication protocols which outperform existing protocols. 3) UNDERSEA NETWORK There is a lot of controversy about whether undersea network is able to become a part of the future 6G network [17]. Undersea wireless communication mainly involves RF, acoustic and optical communication. The comparison between the above three communication technologies is shown in Table 1. Unpredictable and complex underwater environment leads to intricate network deployment, severe signal attenuation and physical damage to equipment [33], leaving plenty of issues to be resolved. B. TOWARDS INTELLIGENT NETWORK Artificial intelligence (AI), more specifically machine learning (ML), has attracted a lot of attention from industry and academia in recent years and initial intelligence has been applied to many aspects of 5G cellular networks [35], from physical layer applications such as channel coding and estimation, to MAC layer applications such as multiple access, to network layer applications such as resource allocation and error correction, and etc. In addition, the combination of artificial intelligence and edge computing proves to improve Quality of Experience and reduce costs [36]. Edge learning also provides new possibilities for the implementation of many applications, such as healthcare [37]. However, the application of AI in 5G networks is limited to the optimization of traditional network architecture and it is difficult to fully realize the potential of AI in the 5G era since the 5G network did not take AI into account at the beginning of architecture design. To fulfill the vision of intelligent network, the design of 6G architecture should consider possibilities of AI in network comprehensively and follow an AI-driven approach where intelligence will be an endogenous characteristic of 6G architecture. Initial intelligence is that a relatively isolated network entity can intelligently adjust the configuration based on multiple predefined options in a different yet deterministic manner [35], which is actually an implementation of perceptual AI without the capability to respond to unintended scenarios. As the network is evolving into an extreme complex and heterogeneous system because of diversified service requirements and explosively growing number of connected devices, a novel AI paradigm of self-aware, selfadaptive, self-interpretive and prescriptive networking is much needed [38]. It requires not only embedding intelligence across whole network, but also embedding the logic of AI into the network structure, in which perception and inference interact in a systematic way, eventually enabling all network components to autonomously connect and control with the ability to recognize unexpected situations and adapt to them. The ultimate expectation of intelligent networks is the autonomous evolution of networks. We highlight three key enablers for intelligent network, as described below. 1) REAL-TIME INTELLIGENT EDGE (RTIE) Next generation network will require the support of interactive AI-powered services and some services like autonomous vehicles are sensitive to response latency, which needs to interact intelligently with their environments in real time. Centralized cloud AI dealing with static data is incapable of achieving such services and there is an urgent need for the RTIE, where intelligent prediction, inference and decision are made on live data. Major academia and industry have begun to develop technologies and software components that meet the real-time requirements in collaborative research labs such as the Berkeley RISELab [39]. High-performance hardware is another driving factor for RTE and a specialized real-time AI processor has been designed in [40]. 2) INTELLIGENT RADIO (IR) In contrast to deployed physical (PHY) layer with initial intelligence, IR is a broader and deeper conception that separates hardware and transceiver algorithms. It operates as a unified framework where hardware capabilities are estimated and transceiver algorithms can dynamically configure themselves according to the hardware information. From the perspective of PHY layer, IR is able to access the available spectrum, control transmission power and adjust transmission protocols 175762 VOLUME 7, 2019