《全球6G技術大會：2024年10.0C6G無線智能云網絡白皮書（英文版）.pdf》由會員分享，可在線閱讀，更多相關《全球6G技術大會：2024年10.0C6G無線智能云網絡白皮書（英文版）.pdf（19頁珍藏版）》請在三個皮匠報告上搜索。

1、1/18Table of Contents1.Abstract.22.O-RAN Architecture.23.6G Vision and design target.34.Key technical pillars and considerations.44.1Network architecture considerations.44.2Service based RAN.64.3AI.84.3.1 Cross-domainAI collaboration.104.3.2 Large Model.114.4 Spectrum sharing.124.5 Sustainability an

2、d Energy saving.145.Forward-Looking.16Reference.16Abbreviation.17Authors.182/181.AbstractO-RAN ALLIANCE has been founded in 2018 by AT&T,China Mobile,Deutsche Telekom,NTT DOCOMO and Orange.Since then,O-RAN ALLIANCE has become a world-widecommunity of mobile network operators,vendors,and research&aca

3、demic institutions operatingin the Radio Access Network(RAN)industry.The mission is to re-shape the RAN industrytowards more intelligent,open,virtualized and fully interoperable mobile networks.O-RANleverages most of the physical features defined in 3GPP,which maintains a unified and healthyecosyste

4、m.O-RAN specifications splits the network entities and defines the interfaces to facilitatethe multi-vendor jointly develop and interoperate test the products.ITU defined IMT-2030 Framework and related timeline,and the industries initialized the 6Gstudy according.O-RAN also kicked off the 6G study i

5、n nGRG,which is to formulate the 6Grelated views before standard.Beyond the advanced features,O-RANs flexible architecture couldprovide some unique advantages for future 6G networks,which includes programable architecturefor network intelligence,service-based RAN design,sufficient network power opti

6、mization,flexible spectrum sharing and etc.This whitepaper briefly introduces some key 6G technical pillars based on O-RAN nGRGdiscussion.To facilitate the reader to understand the technical issues and considerations,theO-RAN architecture is introduced in section 2.We also provide forward looking fo

7、r O-RAN in 6Gera at the end of the whitepaper.2.O-RAN ArchitectureBelow is the O-RAN architecture overview defined by O-RAN alliance 1.O-RANleverages the 3GPP defined interface and also defines some new interfaces as it splits the RANfunctions into O-CU,O-DU,and O-RU.3/18Figure 1 O-RAN Architecture

8、overview O-RANWithin the architecture,RAN Intelligent Controller(RIC)is the logical functions to enablethe controls in a near real time or non-real time manner.The network control functions are splittedinto the two entities based on required time scale.E2 is the interface connecting the NR RIC andO-

9、CU,and most of the network intelligent functions are connected via E2.O-DU and O-RU aretwo remarkable entities to represent the network openness.There are several split options based onthe supported functions on O-RU.Operators could program the RIC functions with different APPs,and multiple APPs cou

10、ld flexibly enable the different network functions.3.6G Vision and design targetITU-R defines the IMT-2030 Framework 2,which includes the usage scenarios andcapabilities of 6G.This framework recommendation is one of the most important guidance for 6Gand would be referred as design guidance for 3GPP

11、and other SDO to specify the 6G standard.Usage scenarios of IMT-2030 are envisaged to expand on those of IMT-2020(i.e.eMBB,URLLC,and mMTC introduced in Recommendation ITU-R M.2083)into broader use requiringevolved and new capabilities.In addition to expanded IMT-2020 usage scenarios,IMT-2030 isenvis

12、aged to enable new usage scenarios arising from capabilities,such as artificial intelligenceand sensing,which previous generations of IMT were not designed to support.The usage scenarios of IMT-2030 include Immersive Communication,Hyper Reliable andLow-Latency Communication,Massive Communication,Ubi

13、quitous Connectivity,ArtificialIntelligence and Communication,and Integrated Sensing and Communication.4/18Figure 2 IMT-2030 Usage Scenarios2O-RANs network architecture provides most flexibility by splitting RAN functions anddefining standard interface.As the 6G usage scenarios are doubled compared

14、to 5G,the O-RANsflexibility would be a good foundation for further innovation.In section 3 of the paper,wediscussed several highlighted technical pillars for 6G design and analyze the challenges andpotential solutions.4.Key technical pillars and considerations4.1Network architecture considerations6G

15、 network bridges the physical and digital worlds.An increasing number of traffic willoccur on the edge of the 6G network.The potential features of future 6G network are intelligence,programmability and resource pooling.Intelligence is the key enabler technology for 6G architecture,and native AI has

16、arousedmore attention from academia and industry.In order to achieve the native AI,the related interface(e.g.,E2)and procedure(e.g.,AI/ML flow)should be considered in 6G architecture.Thesub-section 3.3 describes the native AI in detail.In the context of 6G,the integration of native AI needs an effic

17、ient and convenient approachto incorporate AI elements seamlessly.In addition,the difference 6G service requires differentnetwork resources.Therefore,programmability emerges as a promising solution to drive thedevelopment of 6G architecture.5/18Programmability encompasses three key components:Parame

18、ter,Data,and Algorithm.Programmable parameter facilitates the seamless adaptation of parameters of 6G network througha programmable framework and general interface.Programmable data involves the construction ofdata sets for AI algorithm training and the exploration of data relationships within the n

19、etworkfunctions.Additionally,data can be securely provided to third parties through relevant securedmethods.The programmable algorithms define the input and output data format with differentscenarios.The network functions and third parties can embed or replace the AI algorithms via aprogrammable fra

20、mework under the subjects of the aforementioned input and output and asecurity check.To facilitate this process,a programmable framework is required to deploy andmanage these algorithms effectively.The framework should encompass a comprehensive set ofprogrammable interfaces and function modules,enab

21、ling seamless integration and operation.Additionally,the programmable algorithm ensures that the 6G network dynamically adapts tovarious scenarios requirements.For instance,if consumers seek high throughput from the RAN,the AI algorithm incorporates the RAN slice.Similarly,for consumers prioritizing

22、 Quality ofService(QoS),theAI algorithm integrates the QoS optimization.To enable the implementation of programmable RAN,it is essential to progressively open thetraditionally closed protocol stack within the RAN.This involves enhancing the functionality atthe protocol stack level,and standardizing

23、and generalizing the newly opened interfaces.In thecontext of the ongoing evolution of native AI,programmable RAN catalyzes advancement towarda more open and intelligent RAN vision.By embracing programmability,the RAN can effectivelyadapt to dynamic network requirements,foster innovation,and leverag

24、e the full potential of nativeAI.Resource pooling plays an important role in 6G architecture.The resource is stillheterogeneous,it consists of common and dedicated resources.General resources are common,standardized hardware(i.e.,industrial servers based on X86 or ARM CPUs),and diversifiedhardware c

25、hips with scalability,including acceleration and clock resource chips,and graphicsprocessing unit(GPU)for AI model training.For RAN,high-speed processing and a large numberof dedicated resources are required,such as Field Programmable Gate Array(FPGA)for codingand encoding.The clock resources are ap

26、plied to fulfill synchronization accuracy among networkelements and UEs.Dedicated resources(e.g.,ASIC chips)provide specialized services for a smallnumber of facilities with large capacity and ultra-high-performance requirements.6/18Figure 3 Programmable RAN4.2Service based RANHistorically,RAN archi

27、tecture was mainly designed to guarantee the connection service fortraditional ToC business,using a relatively closed protocol but with the performance advantagesof specialization.As more scenarios and services introduced in 5G and 6G,and IT technologies isintegrated in mobile network,RAN architectu

28、re need to evolve to provide more flexible,adaptablenetwork.O-RAN is defining a more open architecture,building a unified cloud platform for RAN,standardizing more open interfaces,and introducing an intelligent function.The implementationof current cloud RAN only changes the running platform for sof

29、tware instead of changingsoftware architecture of the RAN.This RAN architecture is not cloud-friendly and cannot makefull use of the advantages of cloud-native.Cloud-RAN is the first step,RAN system can be further evolved.SBA in 5G Core Networkcan be used as reference.The goal of Service-based RAN i

30、s achieving a fully cloud-nativearchitecture by rebuilding RAN functions into combinable and reusable network services andusing unified interface with RAN internal services and CN.The advantages of the service-based RAN include:1)Flexible and elastic deployment of network functions,rapid upgrading a

31、ndexpansion of network capabilities,enabling more business scenarios;2)Bring new end-to-end network interaction ways without reducing the impact of7/18cross-domain new functions introduction on existing services;3)More timely and multi-dimensional opening of wireless network capabilities;4)Integrate

32、d managementand orchestrationwithCNservices,reducing thecomplexity of network operation and maintenance,improving the network of adaptability tonew businesses.Figure 4 The concept of service-based RANThe design of the service-based RAN architecture needs to consider the following aspects:Service Gra

33、nularityThe granularity for RAN services which is rebuilt form the originally RAN functions iscrucial.The smaller the granularity,the more flexible it is,but it may bring performance andefficiency issues.The implementation of the 5G core networks service-based architecture includestwo levels:NF and

34、NF Services.NFs communicate with each other,and internal NF services canshare databases which reduces complexity and is also different from Microservices architecture.Considering the internal functional correlations and complexity,RAN can initially be rebuilt in asimilar way to the Core Network.Serv

35、ice-based RAN functions5G RAN can be functionally divided into control plane and user plane,and there is also theconcept of separation of control plane and user plane,but in the deployment layer,it still adopts asingle mode.The control plane mainly includes functions such as connection management,se

36、ssionmanagement,mobility,and measurement,and the user plane includes the processing of datapackets.There are different considerations for on different functional planes.For the control plane,the servitization can rebuild the existing control plane functions intofiner grained services according to th

37、e degree of coupling,and different services can be combinedand flexibly deployed in different scenarios and regions on demand.For example,in the scenarioof the Internet of vehicles,the mobility management service is suitable for centralized deploymentto optimize the mobility experience.At the same t

38、ime,the service-based functions of the controlplane can realize direct access to the Core Network control plane,reduce unnecessary signalingforwarding,and the interaction with other core network services can be changed from serialinteraction to parallel interaction,optimizing the signaling process o

39、f the control plane.The8/18optimization of signaling processes helps improve network performance,such as delay andefficiency.Besides,for extreme requirements of specific services,it also helps RAN and CNintegrates at the edge,simplifying deployment complexity and improving performance.Finally,for th

40、e more complex functional configuration and parameter configuration of the future network,the service-based control plane can be executed and updated at a smaller granularity withoutaffecting the operation of other services.For the user plane,the traditional mobile communication protocols all follow

41、 the OSIhierarchical protocol design concept.Each layer receives specific services provided by the lowerlayer and is responsible for the upper layer.The upper and lower layers interact with each otheraccording to the interface agreement,and the same layer interacts with each other according to thepr

42、otocol agreement.The problem of this layered design concept is that the protocol and servicemodel are fixed,and flexible cross-layer signaling interaction and cross-layer functioncombination cannot be realized.The diversified characteristics of future applications will bringmore differences in data

43、packet processing,such as small data packets for industrial control,whichrequire higher reliability and need to utilize the PDCP replication function in user plane;Immersive interactive applications have different processing requirements for I-frame,P-frame.The user plane needs function combination

44、and arrangement for the new QoS guarantee.Inaddition to the current types of existing applications,sensing,AI and other new applications havealso brought new data packet models,requiring the user plane to be able to match the processingof different data packets,as well as forwarding.The service-base

45、d user plane has advantages inflexible combination,deployment,and rapid update.For scenarios with different user data packetprocessing requirements,the service-based user plane can be preferred.Besides,the new services such as AI,computing,sensing will be provided by the futurewireless system,on the

46、 one hand,this can enable the enhancement of the existing functional plane,such as introducing control functions for sensing and computing power and introducing new userpacket processing mode in user plane.On the other hand,RAN may also introduce new functionalplanes,such as data plane,bringing new

47、functional interactive ways,that will raise more demandson network flexibility and rapid update.Service-based architecture has certain advantages in theseaspects.Service-based interfaceAt present,the RAN and the Core Network interact through the point-to-point N2 interface.For service-based RAN,a se

48、rvice-based N2 interface can be considered,and the RAN is still anindependent whole,RAN services can interact with each other through an internal efficientinterface.This approach is relatively easy to implement and can be adopted during the initialphase of service orientation.The other way is to use

49、 a consistent interface between RAN internalservices and the core network,RAN services and core network services are in a peer position andcan achieve direct interaction,this approach has more advantages,but at the same time will bringmore issues related to network security,ecological change.4.3AIAr

50、tificial Intelligence(AI)has been proposed as one of the most powerful technologies that9/18improves system performance and enables new features in the wireless communication network,by analyzing the data collected and autonomously processed that can yield further insights.3GPP introduced a new logi

51、cal function entity,named NWDAF,to the 5GC to providemultiple types of network data analytic services.The network data analytic services include:Observed Service Experience related network data analytics,to provide average ofobserved Service MoS and/or variance of observed Service MoS indicating ser

52、viceMOS distribution for services such as audio-visual streaming as well as services that arenot audio-visual streaming such as V2X and Web Browsing services;NF Load Analytics,to provide the average load of the NF instance;Network Performance Analytics,to provide the base station status information,

53、resourceusage,communication performance and mobility performance in an area of interest;UE related analytics,to provide the UE mobility analytics,UE communication analytics,expected UE behavioral parameters related network data analytics and abnormalbehavior related network data analytics;User Data

54、Congestion Analytics,to provide congestion experienced while transferringuser data over the control plane or user plane or both;QoS Sustainability Analytics,to provide the QoS change statistics or likelihood of aQoS change for an analytics target period in a certain area.In RAN,3GPP also conducted s

55、everal studies on the AI-enabled network.In Release 17,3GPP conducted a study on AI-enabled RAN intelligence,which defined a reference functionalframework and identified a set of high-level principles to guide the standards work.The study onAI-enabled RAN intelligence focused on three use cases:Netw

56、ork EnergySaving,tooptimizetheenergysavingdecisions(e.g.,cellactivation/deactivation)by predicting the energy efficiency and load state of the nextperiod;Load Balancing,to provide higher quality user experience and to improve systemcapacity by based on collection of various measurements and feedback

57、s from UEs andnetwork nodes;Mobility Optimization,to reduce the probability of unintended events associated withmobility,to predict UE location,mobility and performance,and to steer traffic toachieve efficient resource handling.In Release 18,3GPP conducted a study on AI for NR air interface,to explo

58、re the 3GPPframework for AI for air-interface corresponding to each target use case regarding aspects such asperformance,complexity,and potential specification impact.The study on AI for NR air interfacealso adopted a use case centric approach,focusing on three selective use cases,namely CSIfeedback

59、 enhancement,beam management and positioning accuracy enhancement.In Release 19,3GPP conducted a study on AI for mobility,to improve handover and/or RRM performance bypredicting cell level measurement,handover failure/radio link failure,and measurement events.10/18In the O-RAN architecture,the intro

60、duction of the RIC has been an important development,making it possible to introduce AI based solutions to a widely use cases.Enabling AI drivennetworking requires a paradigm shift in the architectural blueprint.In the 6G,there are threeimportant features of AI need to be considered,namely,native AI

61、,cross domain AI,and networklarge model.Native AI refers to embedding AI into functionalities supported by various nodes/endpointsand interfaces within a network architecture 8.Considering the four key components of AI,i.e.,computing power,data,AI algorithms and functionalities,a native AI network s

62、hould be with ahybrid centralized/distributed AI architecture.The centralized AI entities run for orchestration,managing,deploying and controlling all the distributed AI entities,e.g.,on the SMO platform thatinteracts with other domain-specific AI entities.The distributed AI entities run for serving

63、functions of the local network and receiving commands from the centralized AI entities,e.g.,onthe CN,TN,BS and UE respectively.The 6G wireless network will natively integrate communication capabilities with AI.On theone hand,end-to-end AI may leverage massive amounts of data produced by air interfac

64、es andnetworks to optimize 6G networks and offer consumers customized network services.On the otherhand,as the computing power of infrastructure and terminal devices enhances,future networkswill be able to offer a distributed deployment environment for AI,delivering more flexible andreal-time AI ser

65、vices at the network edge for users.Firstly,it is essential that support for AI betaken into consideration from the beginning when designing network architecture.Thisconsideration must ensure the seamless integration of traditional communication interactions,while the metrics for training and infere

66、nce of AI are converged into the control and data flow.Collaboration within AI is also a critical factor to take into account.This includes cooperationbetween centralized and distributed AI deployment,cooperation between large network modelsand other specialized models,and the cross-domain AI collab

67、oration among RAN,CN,andmanagement systems.Hence,it is imperative to design efficient AI collaboration mechanismsfrom the perspectives of AI orchestration and management,data interaction,distributed learningalgorithms,and computing power scheduling.Last but not least,the problem of AI security hasco

68、ntinuously presented a major obstacle to the use of AI technologies,requiring protections intrustworthy AI,data security,and privacy to guarantee the dependability and security of 6G AIapplications.4.3.1 Cross-domain AI collaborationCross domain AI refers to collaboration and integration of AI-enabl

69、ed functionalities acrossdifferent domains,where the domains can map to networks domains(e.g.,RAN,CN,TN,networkapplications,network digital twins)or other domains 8.In order to enable coordinated AIcapabilities across different network domains,the centralized AI entities(e.g.,on the SMOplatform)shou

70、ld handle the end-to-end AI management and orchestration capability,such ascross-domain data arrangement and mapping,AI task identification and decomposition,mappingAI tasks with computing resources.Figure 3 provides a potential architecture for native and cross-domain AI,where thecentralized AI ent

71、ity is located in the management domain.For E2E intelligent scenario,across-domain AI management function should be added in the SMO as a centralized AI entity to11/18coordinate the AI capabilities from other domains.This module is required to handle AI serviceorchestration,network computing resourc

72、e management,model storage and management,cross-domain AI lifecycle management,and other related functions.Figure 5 Native and cross-domain AI network architecture.For the native and cross-domain AI network architecture,the collaboration control betweendifferent network domains is a new challenge 9.

73、Firstly,the collaboration control functionlocated in the centralized AI entity will decompose intents into service requirements on evolvednetwork domains,where the service requirements will affect connection requirements,AIalgorithm requirements,data requirements and computing requirements.Based on

74、the servicerequirements,the distributed AI entity located in the network domain will decompose the servicerequirements into the network function requirements,connection requirements and resourcerequirements.Secondly,to provide a more real-time management capabilities,the services withhigh real-time

75、requirements and low complexity will be processed by the distributed AI entities,while the services with low real-time requirements,large areas and high complexity will beprocessed by the centralized AI entity.Therefore,the collaborative control methods should beconsidered,e.g.,federated learning,sp

76、lit learning and transfer learning.4.3.2 Large ModelAs a breakthrough development of AI techniques,Network Large Model(NetLM)haveattracted attention from both the scientific community and industry alike.Compared withtraditional AI models that optimize networks under predefined operations,the NetLM l

77、everagesgenerative AI algorithms,e.g.,generative adversarial network and transformer,to automaticallyand creatively generate customized network solutions.For example,for the joint communicationand sensing,the NetLM could help in generating relevant rays(e.g.,generating the rightdistribution of Azimu

78、th and Elevation angles of the radio frequency beam transmitted out of thenode)to capture and sense the surrounding.Another example is the NetLM for digital twin,where12/18the NetLM is able to learn the tail behavior of the train dataset distribution with only a fewsamples,and generate the new behav

79、ior based on what was learned that is consistent with reality.Since NetLM usually consists of billions of parameters,it is difficult to deploy the NetLM onthe edge directly due to the limited computing,communication and storage resources.In addition,the deployment of the NetLM in the core network wi

80、ll also cause tremendous transmission latencydue to the huge and distributed data to be collected and trained in the cloud.Therefore,thedeployment of the NetLM will emphasize the collaboration between NetLM with various scales,including the collaborative training and inference.A standardized collabo

81、ration mechanism needsto be defined,such as network architecture with new network element,large-scale data distributedstorage and real-time provision mechanism,and model-based collaborative interface.4.4 Spectrum sharingFrom 4G era,mobile operators jointly deploy the RAN networks to extend the cover

82、age.Thiscould largely reduce the CAPEX and becomes the major trends when 5G comes.In China andother regions/countries,operators share the frequencies and cooperate on the RAN networkconstruction.CMCC&CBN jointly deploy the 5G NR network on band n28,CT&CU jointlydeploy the 5G NR network on band n79 a

83、nd other frequencies.3GPP specified the RAN sharing mechanism,which is known as MOCN,standing forMulti-Operator Core Network.In a MOCN set-up,one radio access network provides access tothe network of multiple operators.Each operator runs her own core network,but the radio accessnetwork,including car

84、rier signals,is the same for all partners in a certain region.One drawback isonly the operator who owns the RAN network could optimize the scheduler&configuration basedon service characters,and other operators would not have such flexibility.If the other operatorsnew use cases(e.g.XR)are different t

85、han the original optimization,they have to suffer the poorperformance of radio link.O-RAN defines the standard interface between O-CU/O-DU/O-RU.A framework based onsharing O-RUs among operators for spectrum sharing,which builds upon the O-RAN OpenFronthauls innate shared O-RU capabilities,addresses

86、the above inter operator sharing issue andalso many of these limitations and the drawbacks associated with unlicensed spectrum or 3GPPRAN sharing.Under the new sharing framework,unlike unlicensed spectrum sharing,powerlimits can be relaxed,and quality of service can be optimized for each operator.Th

87、e result is auser experience that is comparable to the exclusively licensed spectrum deployments.A“neutral host”O-RU deployment could support shared use from multiple independentO-DUs accessed dynamically via standardized fronthaul interfaces.In addition,it retains all of thecloud-based,complete,cen

88、tralized control capabilities of the Open RAN system.This includes allof the functionality of the O-DU,O-CU,RIC,and advanced network automation capabilitiesinherent in the O-RAN architecture.A shared O-RU architecture relies on prioritized use of resources to guarantee service qualityand allows stat

89、istical multiplexing of traffic among operators and other spectrum stakeholders,13/18resulting in more efficient overall use of spectrum resources.In order for such a scheme to work effectively for all of the cooperating spectrum users,theimprovement in efficiency is critically dependent on how fast

90、 the shared O-RU procedure toallocate idle resource among spectrum users is.The proposed shared O-RU spectrum sharingframework is applicable to sharing between spectrum users including public,private,andgovernment.Additionally,it is radio technology agnostic,i.e.,not all of the cooperating systemsne

91、ed to deploy the same 3GPP radio technology versions,and non-3GPP radio technologies canpotentially be accommodated using the same O-RAN Open Fronthaul-based shared O-RUmechanisms.Shared O-RU with spectrum sharing feature can create opportunities for newspectrum for the next generation networks.In a

92、ddition,it can lead to sustainability improvementsand reduction to CAPEX and OPEX for operators.Figure 6 Shared O-RU network architecture.This O-RU based spectrum sharing scheme could fit the requirements of local network whichis usually deployed in a limited area and sometimes with different spectr

93、um access priorities.Oneexample case is the spectrum for an Industrial Park where a few companies require separatenetwork with diverged use cases&network requirements and government users are with higherspectrum access priorities.Shared O-RU spectrum sharing requires management of coordinated priori

94、ty access tospectrum resources.As it can be inferred from Figure 6,the O-RU is in a unique position tocoordinate access since it is connected to multiple O-DUs over low latency fronthaul networkconnections.While the shared O-RU is the node most suitable to manage coordinated access,coordination can

95、be executed elsewhere in the network,as long as the low latency link towards allO-DU is available.Execution of the coordination function his does not mean that the O-RU mustbe responsible for complex radio resource management and scheduling decisions.On the contrary,in the context of priority-based

96、licensing,the role of the O-RU can be limited to a simple resourceaccess arbiter function that determines which operator has access to resources based on14/18provisioned policy.That policy can be orchestrated(dynamically)by the O&M and RANautomation systems of the primary owner or manager of the O-R

97、U.A typical basic rule is toalways grant access to the priority spectrum license holder,while secondary licensees/users aregranted resources,only if the priority licensee is not using them(subject to business agreements).As illustrated in Figure 6 shared O-RU based spectrum sharing is not limited to

98、 one primary andone secondary user.The benefits of spectrum and O-RU sharing increase as the number operatorsgrow.Therefore,the scalability of O-RAN procedures for spectrum sharing based on shared O-RUneeds to be taken into account.4.5 Sustainability and Energy savingAs stated in IMT-2030 Framework,

99、“sustainability is a foundational aspiration of futureIMT systems.IMT-2030 is expected to help address the need for increased environmental,social,and economic sustainability.IMT-2030 implementations are expected to be designed to achievethe least possible environmental impact and to use resources e

100、fficiently by minimizing powerconsumption,using energy efficiently and reducing greenhouse gas emissions.”Sustainability,or more specifically environmental sustainability,refers to the ability of boththe network and devices to minimize greenhouse gas emissions and other environmental impactsthrougho

101、ut their life cycle.Among the factors of sustainability,energy efficiency is a quantifiablemetric.It refers to the quantity of information bits transmitted or received,per unit of energyconsumption,which is usually measured in bit/Joule.Energy efficiency is expected to beimproved appropriately with

102、the capacity increase in order to minimize overall powerconsumption.Around 2010,CMCC and other major operators shared the observation on the mobilenetwork energy consumption.The observation revealed that half of the energy is consumed on airconditioner and other facilities.They further proposed cent

103、ralizing the RAN equipment to reducethe energy consumption,which is called as Centralized,Cooperative,Cloud and Clean RAN(C-RAN).CNMaincontrolBasebandDigital IFTransmitter/ReceiverPA&LNAAntennaFigure 7 Function split of Cloud-RANBased on CMCCs analysis 5,C-RAN is an eco-friendly infrastructure.First

104、ly,withcentralized processing of the C-RAN architecture,the number of BS sites can be reduced severalfolds.The air conditioning and other site support equipments power consumption can be largelyreduced.Secondly,the distance from the RRHs to the UEs can be decreased since the cooperativeradio technol

105、ogy can reduce the interference among RRHs and allow a higher density of RRHs.Smaller cells with lower transmission power can be deployed while the network coverage quality15/18is not affected.The energy used for signal transmission will be reduced,which is especiallyhelpful for the reduction of pow

106、er consumption in the RAN and extend the UE battery standbytime.Lastly,because the BBU pool is a shared resource among a large number of virtual BS,itmeans a much higher utilization rate of processing resources and lower power consumption can beachieved.When a virtual BS is idle at night and most of

107、 the processing power is not needed,theycan be selectively turned off(or be taken to a lower power state)without affecting the 7x24service commitment.O-RANinheritstheadvantageofC-RANanddefinestheopeninterfaceofO-CU/O-DU/O-RU,which provides sufficient flexibility to analyze the power consumption ofdi

108、fferent network entities.Ericssons report 6 indicates that“For most mobile networks over 80%of energy isconsumed in the radio access network,and the remainder in the core network,support systems andassociated cloud infrastructure.”and“of that estimated 80 percent used by the RAN,around 80percent of

109、that energy is used to power the radios,with the other 20 percent going to thedistributed unit(DU)this represents a major percentage of the total energy consumption of thenetwork.”With adaptation of the new technologies including Micro Sleep Tx,multi-band radiocomponents design&integration,the radio

110、 energy consumption could be saved largely.Meanwhile,cloud DUs are constructed based on high performance generic processors,which hasoptimized energy consumption and good support on software development eco-system.To fullyutilize the potentials,the software of SMO/RIC has been carefully designed.rAp

111、ps that addressthe need for automated non-Real Time network management,and these can be used to reduceoperational costs,improve network performance and reduce energy consumption.In 3GPP,Release 18 studied Network Energy Saving which models the system energyconsumptions and identified several possibl

112、e candidate technologies to reduce the powerconsumptions on feature level.10 As the follow-up,3GPP approved the following working scopefor Release 19 work item.11On frequency domain,SCell is not required to transmit the SSB all the time.UE in connectedmode and configured with the SCell could trigger

113、 the on-demand SSB only when needed.The SCell SSB might be triggered and used by UE for at least SCell time/frequencysynchronization,L1/L3 measurements and SCell activation,and is supported for FR1 andFR2 in non-shared spectrum.On time domain,on-demand SIB1 could achieve some energy reduction for UE

114、s inidle/inactive mode.UE could trigger the SIB1 via uplink wake-up-signal using an existingsignal/channel.The signal of SSB,PRACH and paging are usually uniformly configured in time and spatialdomain.The adaptive periodicity could reduce the network energy consumption,but theperiodicity needs to be

115、 studied to avoid the performance loss or latency increasing.3GPP optimization points out the potential directions to optimize the RAN energyconsumption and O-RAN architecture provides most flexibility via split network entities&programmable rAPPs.With the combination of the O-RAN architecture and t

116、he new energy16/18saving features,a bright future is expected in 6G era.5.Forward-LookingO-RAN has been founded to reformulate the ecosystem and promote the network openness.By fully utilizing the splitted network entities and standardized interfaces,O-RAN could largelyincrease the network flexibili

117、ty and serve some 6G usage scenarios like AI,service-based network,sustainability and Energy saving.Based on the analysis in this whitepaper,we could foresee theunique benefits of open network architecture in 6G era.Beyond the new technologies,an important issue for future network is the interplay b

118、etweenO-RAN and 3GPP.In 5G system,3GPP defines the general network entities and functions,O-RAN makes further split on RAN entities and defines the interfaces.Once 6G study starts,theeco-system should discuss and define the work split between the two organizations,especiallywhether 3GPP could take m

119、ore roles on network openness.Reference1O-RAN alliance,O-RAN Architecture Description,version 11.002ITU-R,Recommendation“ITU-R M.2160 Framework and overall objectives of the futuredevelopment of IMT for 2030 and beyond”,Nov.20233TR 25.927,Solutions for energy saving within UTRA Node B,Release 13.4Ch

120、ih-Lin I,Jinri HUANG,Ran DUAN,Chunfeng CUI,Jesse JIANG,and Lei LI,RecentProgress on C-RAN Centralization and Cloudification,IEEE Access,20145CMCC,C-RAN The Road Towards Green RAN6Ericsson,Open RAN architecture:Embracing energy efficiency-Ericsson7nGRG-RR-2023-03,Research report on native and cross-d

121、omain AI:state of the art andfuture outlook.8nGRG-RR-2023-02,Research report on O-RAN native AI architecture description.93GPP TR38.864,Study on network energy saving for NR,Release 18,Mar.2023.10 3GPP RP-234065,New WID:Enhancements of network energy savings for NR,Dec.2023.17/18Abbreviation5GThe Fi

122、fth-Generation Mobile Communications6GThe Sixth-Generation Mobile CommunicationsO-CUOpen Central UnitO-DUOpen Distributed UnitO-RANOpen RANO-RUOpen Radio UnitRICRAN Intelligent ControllerSMOService Management and OrchestrationMOCNMulti-Operator Core Network18/18AuthorsThis whitepaper is with jointly input from CTC,CUCC,CMCC,Lenovo,and Qualcomm,andreceived valuable guidance from Dr.Chih-Lin I.