1、The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0WHITE PAPER V9.0C2023.03-01-Executive SummaryBased on the previous study of RAN in white paper,this white paper facing 6G gives a potential solution for the architecture of RAN and framework of the protocol stack.In t

2、his solution,the technology of digital twin is introduced to support AI to embed into functionalities of RAN,and the native-AI computing power for native-AI RAN is scheduled on demand.The white paper hopes to arouse the continuous attention and thinking from the the academic and industrial on the re

3、search of the architecture of RAN and the framework of protocol stack for 6G,so that the deep integration of ICDT and promote the comprehensive development of 6G are achieved.01 01 03 0303 0917171923252931 IntroductionNative AI RAN ArchitectureCloud and Terminal Flexible Architecture for Mobile Netw

4、orks ICDT RAN Function of Native AISummaryIntent-Driven Protocol Management Control ArchitectureDevelopment of Network Management ProtocolsIntent-Driven Next-Generation RAN Control ArchitectureKey TechnologiesIntent Lifecycle ManagementSummaryTable of ContentsExecutive Summary Introduction 1 22.1 2.

5、22.333.13.23.33.43.5Reference AbbreviationAcknowledgement3233 341.IntroductionThe white paper proposes 6G protocol stack views and thinking for 2030+,based on the published versions such as The Next-generation Protocol Stack over Air Interface 4.0 and The Next-generation Protocol Stack over Air Inte

6、rface 3.0.We hope to provide reference to study the 6G-oriented protocol stack architecture and functions for the industry.The CU-DU separation architecture of 5G provides flexibility for the distributed deployment of base stations.Based on the CU-DU separation architecture,on-demand functional conf

7、iguration of 6G RAN is achieved,thus realizing a flexible architecture of cloud and terminal,which provides architectural support for native AI and digital twins.By defining the functions of the core network and separation mode base stations on the cloud,the functions of the remaining base stations

8、are deployed on the terminal.The cloud and terminal are wirelessly connected,and the terminal side provides wireless coverage over air interface.Through the control on the cloud side,the terminal side realizes wireless coverage,wireless shutdown,interference coordination and load balancing.The next-

9、generation RAN relies on the application intent mining capability provided by intent-driven networks,the global sensing capability of network status,and the real-time optimization capability of network configuration.It proposes an intent-driven next-generation RAN management protocol architecture,in

10、cluding modules such as intent translation,policy configuration,and resource orchestration.Through real-time status monitoring,it realizes the full life cycle management of user intent,providing a new feasible approach for the next-generation RAN management protocol.2.Native AI RAN Architecture2.1 C

11、loud and Terminal Flexible Architecture for Mobile NetworksUnder the unified control of the cloud side,each terminal side or service access point can act as a terminal.The core cloud side uses AI modeling and digital twins to achieve full-scale digital control of physical space,thereby achieving pre

12、cise dynamic networking and coverage expansion,and achieving flexible networking.-03-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0The cloud side digitizes the physical space of the network coverage,establishes a digital twin warehouse for big data,and introduces

13、 AI tools to control the coverage on the terminal side,achieving the self-organized wireless coverage on the terminal side under the unified control of cloud side,and achieving flexible network coverage on-demand,based on scenarios and service needs.1.Functional entities on the cloud side.Communicat

14、ion network functions:including the functions of the core network(CN)and some radio access network(RAN)functions.These functional entities are software-defined,including packet processing functions and signaling control parts.The core network includes all functional entities of the core network.Some

15、 RAN functions include layer 3 control plane and user plane functional entities,as well as layer 2 data packet processing parts.For example,the UPF,AMF,and SMF functional entities of the core network,as well as the RRC,SDAP,and PDCP signaling and data plane functional entities of the RAN are include

16、d.In addition,there are interface functions connected to the terminal side,including initial establishment,modification,and deletion of interface connections.AI and big data functions(AI&BD)refer to digital modeling information for the entire network coverage space,such as the coverage range,buildin

17、g features,spatial height,user service and behavioral characteristics within the coverage range,and three-dimensional coordinate information of the coverage space.State parameters of the entire network operation,various state parameters related to the network itself,service,and users generated durin

18、g network function operation.Storage function,computing function,and training functions of AI models for the above data.Based on the above data,a digital twin(digital mirror:by sampling the physical environment feature parameters,the digital description of the physical environment is achieved.)envir

19、onment of the physical coverage environment is constructed.Based on this digital twin environment,AI tools are used for network self-evolution,self-generation,and self-maintenance,to carry out functional testing,operation verification,error testing,and deployment of new features based on the network

20、s new functions and characteristics.The AI and big data functions on the cloud side model,predict,and control the network coverage on the terminal side.When it is necessary to turn on one or several wireless devices on the terminal side(which can be various terminals or wireless access points specia

21、lly used for coverage enhancement)according to terminals or wireless coverage devices,it provides a networking solution for terminal devices in the area,including the connection relationship,-04-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0power,interference coo

22、rdination strategy,maximum service volume,and wireless air interface networking strategy for each terminal and wireless coverage device.2.Functional entities on the terminal side:mainly including the low-level processing part of the RAN facing the air interface.For example,the MAC layer functional e

23、ntity,the physical layer(PHY)functional entity,and the radio frequency part of the air interface.In addition,there are interface functions connected to the cloud side,including initial establishment,modification,and deletion of interface connections.The two parts can be connected by wired or wireles

24、s means.If it is a wireless connection,the terminal side device has two types of wireless interfaces:one is a wireless frequency point and corresponding frequency point search function that is connected to the cloud side device,and the other is a wireless frequency point function that can receive wi

25、reless air interface coverage of terminal access.The terminal side functional device has the function of searching for cloud side functional device;after the terminal side functional device is powered on,it actively searches for the cloud side functional device and initiates the connection establish

26、ment.Figure 1 Cloud and Terminal Flexible Architecture for Mobile Networks-05-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0Figure 2 Initial Access Process for Cloud and Terminal DevicesIn Figure 1,the cloud side devices run on the cloud platform,and the terminal

27、 side devices are deployed in remote locations to achieve air interface coverage.Initial Access Process:Figure 2 shows the process of initial access of the terminal device to the cloud device.Step 1:After the terminal device is powered on,it only goes through the initial access process of the cloud

28、device and does not transmit signals to accept terminal access.If it is a wireless connection,the basic information of the air interface access of the cloud device is obtained by receiving the broadcast sent by the cloud device.-06-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack a

29、nd Function 5.0Step 2:Start searching for the cloud device.If it is a wired connection(such as the terminal device detecting that the network port or other cable connection interface is already connected),the terminal device initiates a connection request on the wired port.If it is a wireless connec

30、tion,the terminal device searches for the cloud device signal,completes frequency locking,and initiates a connection request on the selected frequency point.The terminal device has the ability to provide air interface coverage of a base station and to connect to the cloud device via wired or wireles

31、s means.The terminal device also has the ability to connect to the wireless frequency points and frequency points of air interface coverage for the cloud device.The wireless frequency point connected to the cloud device can be the same system bandwidth as the air interface coverage frequency domain(

32、3G,4G,or 5G frequency points),or it can be a completely independent different frequency point(3G,4G,or 5G frequency points),or it can even be microwave or Wifi.Step 3:Complete air interface synchronization and clock synchronization of the wireless system through a random access.The air interface syn

33、chronization is for the uplink channel synchronization.The terminal device completes the uplink channel synchronization through random access by receiving the broadcast of the cloud device for downlink synchronization.At the same time,the system clock is obtained during the random access,and the clo

34、ck of the cloud device is carried in the random-accessed response message sent from the cloud device to the terminal device.After completing the random access process,the terminal device sends a connection establishment request message to the interface of the terminal device and the cloud device.The

35、 message for connection establishment request contains relevant information such as the port and bearer protocol of the terminal device functionalities of interface.Step 4:After the cloud device receives the message for connection establishment request from the terminal device,it sends a connection

36、establishment message.This message contains information about the port and bearer protocol of the terminal device that corresponds to the functionalities of the interface of the cloud device.Step 5:After the terminal device successfully establishes the connection,it replies with a message indicating

37、 connection establishment completion.-07-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0Step 6:The terminal device reports its capabilities to the cloud device.This includes the maximum number of users the terminal device can support,the QoS level of user services

38、,the maximum coverage of air interface,the frequency point list of all air interface coverage supported,the version of base station protocol supported,the functions of the protocols on wireless access layer supported,including one or several protocol layer functions of Layer 1,Layer 2,and Layer 3,th

39、e functional parameters supported by the protocol layer,such as the RLC layer only supporting the UM mode,and the PHY layer only supporting the 16 Ports capability,and the configurable wireless access layer functions that can be supported in the flexible network.Step 7:After receiving the capabiliti

40、es reported by the terminal device,the cloud device obtains the existing cells at the location of the terminal device,and configures the neighbor cells,including the frequency,maximum transmission power,maximum coverage radius,average coverage radius,and other parameters,to the terminal device.Step

41、8:After receiving the neighbor cell list configured by the cloud device,the terminal device initiates neighbor cell measurement.It will obtain the signal strength,received interference level,and other measurement parameters of each cell in the list at the location of the terminal device.Step 9:The e

42、nd device reports the measurement parameters of the neighbor cells power,interference,and other identifying signal strengths to the cloud device.Step 10:After receiving the measurement,the cloud device calculates the signal strength around the terminal device,and decides whether to allow terminal ac

43、cess as a base station device based on the signal condition and the current load of the cell.If allowed,the transmission power,signal coverage range,system load(number of users,data throughput,etc.),and access layer protocol functions of the terminal device will be configured.If not allowed,the term

44、inal device will be configured to be silent.Step 11:The terminal device executes the cloud configuration after receiving the cloud devices configuration,and continues to report measurements to the cloud device.Step 12:The cloud device periodically or event-triggeredly performs clock synchronization

45、with the terminal device,and the terminal device may also initiate clock synchronization requests-08-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0based on its own operation.Step 13:Based on the big data information in the digital environment of the terminal devi

46、ce,the cloud device obtains the interconnection and networking scheme of the terminal device with the existing terminal devices,and configures it to the relevant terminal devices to trigger the update of the interconnection relationship of the terminal device,thereby achieving flexible update of RAN

47、 coverage.Through the above steps and the unified control of the cloud device,mutual interference of the terminal devices is avoided.One application scenario is:Installing an SoC that supports terminal devices on a laptop in the office.When everyone is having a meeting in a conference room,each lapt

48、op can function as both a terminal and an access point(AP)for the base station.Each laptop automatically connects to the cloud device,and through the control of the cloud device,it is determined which laptop can be a base station and which one is a terminal.This enhances indoor coverage while avoidi

49、ng interference.Service process:When the terminal performs service through the terminal device,the transmission capacity of the connection interface between the cloud device and the terminal device becomes the bottleneck of user service capabilities.At this time,traffic control is introduced on the

50、connection interface.The terminal device sends a traffic request to the cloud device,which improves the bandwidth utilization according to the request.2.2 ICDT RAN Function of Native AIIn 4G/5G networks,AI is an external AI.That is,various information(such as MDT measurement enhancements currently b

51、eing studied and promoted in 3GPP R17)required by AI functional nodes is reported to AI through the network side(base stations and core networks)and terminal side.-09-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0AI-related processing is carried out outside of ne

52、twork elements,and AI functional nodes collect,process,and train AI models,and then send the results or generated policies of AI operation back to the network.In the research of 6G networks,native AI and digital twins have become the core features of 6G networks.In the Native AI 6G network,AI is no

53、longer simply optimizing wireless resources of the RAN,but has become an AI system integrated with the core network,transmission network,and wireless access.The 6G digital twin system provides a fundamental running environment for the native AI of 6G networks,which provides basic support for AI-rela

54、ted processing and computing,while simplifying the physical networks operating load and complexity.In other words,the digital twin system and the native AI system of 6G networks together form a series of online operations for physical network operation,maintenance,and application-oriented control ca

55、lculations,becoming the brain of the physical network and commanding every part of the physical network to complete the service capability required by the protocol or operator.Currently,the AI in 4G/5G networks is external AI,and there are two challenges that cannot be overcome.1)In order to make AI

56、 results more accurate or effective,a large amount of real-time and fine-grained measurement information needs to be reported to external AI centers,which has high costs and interoperability issues(interconnection between different manufacturers),making it impossible to commercialize in commercial n

57、etworks.2)The effectiveness of AI on the network relies on the accuracy of measurement data,which can lead to AI-generated policies that cannot match the needs of the network and fail to reflect the gains AI brings to the network,making it impossible to achieve the vision of an intelligent network.T

58、he ICDT RAN solution of Native AI and Digital Twin(NAS:Non Access Stratum,relative to AS:Access Stratum;the stratum is divided according to NAS and AS on the terminal side,and according to CN and RAN on the network side).ICDT stands for Intelligence,Communication,Data,and Technology.In this solution

59、,multi-level control of RAN is supported by introducing online digital twin and distributed AI functions into different functional parts of RAN,thereby realizing the ICDT RAN solution of Native AI and Digital Twin.-10-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.

60、0Online digital twin and distributed AI functions are introduced into cloud units and edge access points of RAN to achieve flexible control of the cloud and edge.Figure 2 shows the ICDT RAN solution of Native AI.For Online DT and Offline DT,offline and online are not related to traditional high/low

61、real-timeliness or high/low speed and reflect the synchronization relationship between the corresponding functional entities.For example,Offline DT can obtain high real-time data for a period of time through recording and then process the data to obtain simulated results.Distributed AI is a logical

62、constraint relationship between AI algorithms or AI functions that exist between various functional entities in the network.The distributed logical constraint relationship defines the different AI algorithms and AI functions as a whole to complete different network-related tasks,and defines the task

63、s that each separated AI algorithm or AI function needs to independently undertake under the unified requirements of the overall network function,including the calculation results directly output to the network function,the calculation results output to other AI algorithms or AI function entities,an

64、d the acceptance of calculation results from other AI algorithms or AI function entities.This logical constraint relationship can be a tight or loose constraint relationship.Through this logical constraint relationship,AI algorithms or AI functions collaborate to complete specific tasks,and differen

65、t AI algorithms or AI functions can directly exchange messages,measurements,or data information.Figure 2 ICDT RAN solution of Native AI-11-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.01.For the inputs of each AS functional entity,they are also synchronously inpu

66、t into the corresponding online simulation functional entity in the DTL.In order to reduce the overhead of running the DTL,critical information can be input into the DTL,while non-essential information,such as data packets(PUD),can be discarded.Effective PUD headers can be input into DTL,while the n

67、et payload of the data packet is not a must.The input information includes signaling,data packets,measurement information,operational maintenance instructions,etc.2.Each functional entity of the AS reports its own key information during its operation to the DTL.For example,the error rate and delay o

68、f each PUD received or sent.Additionally,each functional entity provides its own signaling control information,such as UE status changes and measurement information about link quality during its operation.3.Based on the input information,the DTL simulates the operation of each AS functional entity o

69、nline and generates signaling,policies,or commands for different AS functional entities under the drive of AI algorithms.4.The signaling,policies,or commands generated by the DTL are fed back as inputs to the AS functional entities.Each NAS functional entity performs low-cost or low-overhead fast pr

70、ocessing based on the inputs.5.The results are output by each AS functional entity.6.The output results are simultaneously fed back to the DTL for verification or correction of each output result.On the RAN side,changes in the configuration and updates of the distributed AI(AIL)and online DT(DTL)for

71、 the UE side are carried through RRC signaling,MAC Control Element(MAC CE),or MAC PDU.On the UE side,the operation status parameters,measurements,establishment,or update requests for AIL and DTL are carried through RRC measurements or MAC CE or MAC PDU and sent to the RAN side.The AIL and DTL on the

72、 UE side are entirely established,changed,or deleted under the control of the RAN side.-12-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0OAM controls the AIL and DTL on the UE side through the above RAN methods.On the CU(cloud unit),RRC,RRM,L3UP,and C-MAC(cloud M

73、AC)are running.On the EU(edge unit),D-MAC(dedicated MAC,dedicated to the MAC function within the cell)and PHY are running.SDAP/PDCP/RLC can run independently on the CU and EU(if these three protocol function layers still exist in the future).For RRC and RRM,they are classified into two major functio

74、ns according to UE-level or related functions and Cell-level or related functions.UE-level or related functions refer to the management,control,and maintenance of the context information,link status and service status of the UE accessed to the RAN.Cell-level or related functions refer to the managem

75、ent,control,and maintenance of all cells established by the RAN,such as the maximum number of users the cell can accommodate,the air interface bandwidth of the cell,the maximum transmission power of the cell,the maximum throughput of the cell,the coverage area of the cell,the identity information of

76、 the cell,the information of adjacent cells,etc.L3UP has two types of functions:1.L3UP PDU data processing;2.Associated information collection and processing corresponding to each PDU processed.For example,shaping the received upper-layer data(i.e.,restoring the regularity of the received upper-laye

77、r data based on the known service characteristics,so as to send more accurately over the air interface),controlling the traffic of data sent to the lower layer,state information of each data packet sent or received at the lower layer,and statistical information of received or sent data packets(size,

78、interval,sending delay pattern).C-MAC has two categories of functions:1.Control of D-MAC;2.CU-EU interface control.The control of D-MAC mainly includes:SDU shaping of data sent to each D-MAC:According to the QoS requirements of the service and the transmission status of each D-MAC data,control the t

79、raffic of data sent to each D-MAC,eliminate the transmission jitter of data packtets,schedule data packets carrying associated control information,reconstruct the sending of important or critical data packets,and send data packets while restoring the correct order of packets.-13-The Next-generation

80、Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0SDU shaping of received data on each D-MAC:except for traditional sorting,elimination of repeated data packets,deconstruction and reconstruction of important or key data packets,traffic control and the sensing of data receiving state.Col

81、laborative control between different D-MACs:including coordination of PRBs(Physical Resource Blocks)within overlapping coverage,power allocation or control.Orchestration of D-MAC functions:based on the characteristics of the data to be transmitted(service,traffic),specific MAC functions are customiz

82、ed,including scheduling functions,measurement of resource allocation,HARQ mode selection,mode selection of HARQ process,and HARQ process selection,rate or modulation method or transport block(TB)selection.User-level D-MAC selection:for one or more D-MAC functional units belonging to a UE,that is,the

83、 link between C-MAC and D-MAC,dynamic selection or scheduling is performed as needed.Based on the quality monitoring of data reception and transmission on each D-MAC of the UE,one or more suitable D-MAC links for data reception and transmission are selected,including activating new D-MAC links,deact

84、ivating inappropriate D-MAC links,improving or reducing the service capability or quality of one or several D-MACs.User-level D-MAC link switching:based on the quality of reception and transmission of each D-MAC link of the UE,as well as the UEs possible movement trend,flexible selection of D-MAC li

85、nks is achieved.Plug-and-play control between CU and EU:Plug-and-play between CU and EU has two levels:1.plug-and-play for transmission connection between CU and EU;2.plug-and-play between wireless protocol stacks.This invention mainly focuses on the second level,that is,plug-and-play between wirele

86、ss protocol stacks.1.Introduction of the concept of common bearer(CB)between C-MAC and D-MAC:CB is the bearer for the connection between C-MAC and D-MAC,and is responsible for establishing and identifying connections between C-MAC and each D-MAC.C-MAC is the MAC at the cell level,and D-MAC is the MA

87、C at the user level.-14-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0When a cell is established,C-MAC is established.A cell can have one or more C-MACs.A D-MAC of a UE can only connect to one C-MAC at the same time,that is,the UE managed by the C-MAC is orthogon

88、al at the same time.When the cell is released,D-MAC is released.When a UE is established,D-MAC is established.D-MAC is established when SRB1 is established for the UE,and D-MAC is reconfigured for subsequent SRB2/3 and DRB.A UE can have one or more D-MACs,but each D-MAC corresponds to a non-overlapp

89、ing or orthogonal air interface.When the UE is deleted,D-MAC is deleted.The CB ID is the same as the cell ID.When C-MAC is established,CB is established,and CB includes at least:CB ID,the list of D-MACs connected to CB(initially empty),and the routing information of each D-MAC connected to CB(based

90、on this CB,each D-MAC connected to the C-MAC can be found).Information carried on CB includes:(1)C-MAC needs to transmit cell-level information through each D-MAC in the air interface(such as MIB,SIB,and uplink and downlink time slot adjusting methods).(2)C-MAC configures control information of D-MA

91、C(such as D-MACs maximum downlink transmission power,the target power of the uplink power control,QoS requirements for the data sent by D-MAC(packet sending delay,error rate,retransmission times,retransmission rate,coding rate,etc),data type indication for multiple data packets multiplexing into one

92、 transmission,data type indication for multiple data being transmitted concurrently,maximum or minimum coding rate when D-MAC sends data,transmission characteristics of the data being transmitted,time cycle for sending mode,DRX configuration,and DTX configuration).(3)D-MAC reports the state informat

93、ion of data reception and transmission to C-MAC(including error rate,retransmission times,residual error rate in real transmission,probability of segmental or cascading transmission,idle rate of data cache,overflow rate of data cache,etc.).(4)Other information such as heartbeat information reported

94、by the D-MAC.-15-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0(4)Other information such as heartbeat information reported by the D-MAC.Figure 3 shows a diagram of the connection between C-MAC and D-MAC on the common bearer.Wireless carriers such as SRB2 and DRB

95、established after D-MAC being established belong to a carrier included in CB,that is,a CB contains multiple UEs,and each UE contains multiple SRBs or DRBs.After D-MAC is established through SRB1,D-MAC actively sends an access request to C-MAC according to the CB ID,carrying the identity recognition

96、information of the D-MAC.C-MAC obtains the routing information of the D-MAC according to the identity recognition information of the D-MAC and records the D-MAC as an active state.Then C-MAC gives a response to the D-MAC with an establishment confirmation message and carries configuration informatio

97、n(all or part of the information in(1)or(2)in yellow of C-MAC configuration).In this way,when SRB1 is established,D-MAC is already active,realizing the plug-and-play of UE as a unit.Subsequently,default bearer and even no-bearer connection oriented to QoS can be introduced between C-MAC and D-MAC ba

98、sed on CB,thus achieving the plug-and-play of the data plane.Figure 3 ICDT RAN solution of Native AI-16-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0If the UE still uses traditional SRB and DRB for data transmission,then subsequent SRBs or DRBs established for t

99、he UE will be based on the CB of the C-MAC.2.3 SummaryThe RAN for native AI in 6G is based on the computational power and flexible scalability provided by cloud platforms to achieve on-demand scheduling.The network needs to sense terminal information,including location,service,terminal running state

100、,control requirements,mobility status,potential behaviors,etc.,and achieve native AI of the network through the processing capability of the network side.3.Intent-Driven Protocol Management Control ArchitectureThe intelligent information society in 2030 will be highly digital,intelligently inspired,

101、and globally data-driven,enabled by near-instant and unlimited full wireless connectivity.6G will be an autonomous ecosystem with artificial intelligence and consciousness.It will evolve from a human-centric approach to a machine-centric one,and provide multiple ways to communicate and interact with

102、 smart terminals,such as through fingers,voice,eyes,and brain waves(or neural signals).As mobile communication networks rapidly develop,on the one hand,the types,quantities,and services of terminals have dramatically increased,making it difficult for existing network architectures to expand new func

103、tions and services,and the performance requirements of networks are becoming increasingly higher,beyond the capabilities of the current 5G architecture.On the other hand,based on the current 5G network architecture,network management protocols are difficult to adapt to the highly elastic and dynamic

104、 service needs of future 6G networks.The main problems and difficulties are as follows:(1)Traditional network management protocols are difficult to obtain dynamic data of the entire networks resources in real-time,and they cannot monitor the communication system comprehensively.(2)Existing network p

105、rotocols have poor scalability,and service capabilities are limited,making it difficult to manage complex multi-scene objects.(3)Large-scale manual configuration and decision-making make the network prone to errors,and when problems arise,it is necessary to judge the causes of the errors manually,an

106、d the error correction process is complicated,leading to slow recovery of network speed.-17-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0Therefore,in the era of the sixth generation of mobile communication technology,a global control technology on network resour

107、ces is urgently needed for full coverage,full spectrum,and full-scene services.Such technology can achieve O&P automation,break through the contradiction between transmission costs and personalized services,provide users with a simple network architecture,and provide vertical industries with communi

108、cation networks featuring customized services,operational automation,interface standardization,and intelligent simplification of control,and also solve problems on the service experience of operators and users in an effective manner.The emergence of Intent-Driven Network(IDN)provides a new idea for

109、users personalized service needs.At the same time,IDN can be applied to autonomous management,self-optimization,and automatic configuration in various network scenarios,such as the Internet and data center networks,reducing manual intervention in network management and providing ordinary users with

110、higher participation.It is a programmable and customizable automated network that combines application intent mining,global sensing of network state,and real-time optimization capabilities of network configuration.The intent is a declarative description of the system state.It abstracts the objects a

111、nd capabilities of the network from the perspective of requirements and can be transformed into advanced policies.In the intent-driven network,based on the operators intent,it can be automatically translated,verified,deployed,configured,and optimized to achieve the target network state.At the same t

112、ime,relying on the networks holographic sensing and feedback optimization loop,automatic resolution of closed-loop abnormal events is realized to ensure network reliability.IDN is applied to network management protocols,mainly emphasizing the use of advanced user intent to specify network behavior,r

113、ather than low-level configuration.Through the intent-driven network management protocol,users can specify what they want the network to do,and the network management system will translate that intent into low-level configuration that the network can understand.Overall,as the representative of the m

114、ost cutting-edge network technology,IDN mainly uses intent translation and policy configuration technologies to separate ordinary users from the underlying network technology.Users only need to propose service requirements,and the network automatically translates intent into policies,and policy conf

115、iguration forms a new mode of automated network management for services.IDN includes several key steps such as intent-18-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0acquisition,intent translation,policy delivery,real-time monitoring,and closed-loop feedback.Fir

116、stly,it acquires the users intent and translates it into network policies,and verifies the feasibility of policies based on the current network state.Next,the verified policies are delivered to the underlying infrastructure.Finally,the system must monitor the network state in real-time to ensure tha

117、t the users intent is correctly implemented and feedback the results to the user.IDN can greatly improve network management efficiency.As an intent-driven network,IDN can improve network management efficiency,and the intelligent operation of network systems reduces the labor cost of network manageme

118、nt,making intent-driven technology the key to achieving network autonomy in the future.3.1 Development of Network Management ProtocolsWith the development of network management protocols,different protocols have emerged.Figure 1 shows the development stages of network management protocols,which are

119、TMN,PBNM,and IDNM.The following provides a detailed introduction to each.1TMNTMN(Telecommunication Management Network)is an organized network that provides a range of management functions and allows communication among various types of operating systems through standard interfaces.The network manage

120、ment of 3G systems is mainly based on the TMN framework.TMN provides five management functional domains for telecommunication network and telecommunication services:performance management,configuration management,accounting management,fault management,and security management.TMN management services

121、can be broadly classified into three categories:(1)daily service of communication network and management service of network operation;(2)network maintenance services such as monitoring,testing,and fault handling of communication networks;(3)network control service such as network control and abnorma

122、l service processing.However,in the case of increasingly complex mobile communication networks with growing scale and constantly increasing services,on the one hand,the burden on administrators is increased based on the TMN network management model,making it difficult to support the management tasks

123、 of the increasingly complex network service.On the other hand,the large number of minor configuration and query-19-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0commands for devices leads to a high rate of human error,reducing the overall efficiency of network m

124、anagement.2PBNMTo overcome the limitations of TMN,the Internet Engineering Task Force(IETF)standardized Policy-based Network Management(PBNM)and applied it to the management and control of large-scale networks.The basic idea of PBNM is to allow network administrators to focus less on the fine config

125、uration and management operations of network devices and more on the goals of network management and the network abstract behaviors needed to achieve those goals.Administrators are responsible only for policy formulation and their operations can more directly reflect decision-making and goal orienta

126、tion.Various management tasks such as data collection,network monitoring,data analysis,and device control are automatically implemented through predefined management policies and can dynamically adapt to various complex environmental changes.In PBNM,administrators define and characterize users manag

127、ement goals or service needs in the form of policies.The control system queries the relevant policies and calculates the feasibility of the policies,converts the policies into commands that devices can recognize,and performs real-time specific configurations and operations.In PBNM,policies have diff

128、erent manifestations at the high,medium,and low levels for different hierarchies and stages of the network.The differences in the hierarchical and usage aspects of network policies largely result in that the description and expression forms of policies are diverse.This makes the policy-based network

129、 management system complex and lacks a refined and unified standard expression method.In order to unify the expression of policies,the policy is further aggregated and abstracted to form an intent model that supports network policies.3IDNMIntent-driven network management(IDNM)is an advanced extensio

130、n of PBNM.It utilizes a standardized highly abstract expression similar to natural language,known as intent.This intent is input into the network and subsequently transformed into a low-level abstraction policy through processes such as intent translation and policy generation.Ultimately,the policy

131、is converted into a configuration command recognizable by the device for specific network operations.By combining-20-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0PBNM,IDNM avoids the complex and diverse technical details of the underlying network and achieves th

132、e goal of automatic network management.In response to the shortcomings of PBNM,IDNM uses a high-level abstract language to input intentions and then achieves automatic network management by converting intentions into implementable policies and combining policy-based network management.Figure 1 Relat

133、ionship of Network Management ProtocolsAs shown in Figure 1,the relationship among intent,policy,configuration,IDNM,PBNM,and TMN can be illustrated.From top to bottom,the refinement process of intent can be divided into three levels:high level of abstract intent,lower level of abstract policy,and co

134、nfigurable configuration.From top to bottom,as the network scale and technical details expand,the level of abstraction can shield the complicated technical details.With the intent which has the highest degree of abstraction,the network can be managed,forming a mapping from business intent to system

135、policy and finally to detailed configuration,achieving flexible network management.-21-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0Policy is a form that the network control system can understand and operate.Policies can be divided into three types.Command polic

136、y refers to the state that explicitly defines specific goals and uses intent-action-state to operate a set of service in the correct order.Declarative policy expresses the goals of policies rather than how to achieve them.Intent policy refers to the high-level abstract expression of policy goals.Fur

137、thermore,intent can be understood as an intent-based policy.Intent is a language form that users can express and understand.Configuration is a form that can be configured and operated on the underlying device and can be understood as a command policy.The use of intent can allow network administrator

138、s not need to create detailed policies,nor do they need to understand the overall situation of the network,such as topology and link information.They only need to express the goals of the service,and through translation,multiple network policies can be obtained to achieve the goal.These network poli

139、cies can then be transformed into commands for configuring and operating devices through the existing network management architecture.Overall,IDNM means that the network O&P mode is gradually shifting from a device-and network-centric to a user-and service-centric mode.In summary,IDN has the followi

140、ng advantages as a network management tool:(1)High compatibility and utilization:IDN does not rely on various interfaces of different manufactures devices at the physical level.For O&P,IDN can solve network management problems in heterogeneous environments and allow users to achieve centralized cont

141、rol in the form of policy-based management.(2)Dimensionality reduction of static parameters and dynamic network optimization:IDN can reduce manual configuration workload,automatically simplify parameter configuration,and improve the reliability of network services.In IDN,network failures can be quic

142、kly and accurately located.In addition,IDN achieves network self-optimization and self-recovery.(3)Multi-party win-win network:For network administrators,they can focus more on network services rather than specific and intricate network configurations,making network delivery closer to service.For us

143、ers,the network can provide a variety of services and a better user experience.For network service providers,network configuration and operation are more concise,efficient,and cost-effective.-22-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.03.2 Intent-Driven Next

144、-Generation RAN Control ArchitectureThe 6G network requires further enhanced sensing capabilities,including sensing of behavior,service,and intent,and allocation of network resources based on user service needs to provide experience assurance and on-demand services.IDN provides an effective way to m

145、eet these requirements.Declarative intent can automatically translate,validate,publish,configure,and ensure that the network state meets user expectations.At last,it achieves the automation and closed-loop optimization of network services.In 6G,network management needs to achieve integrated manageme

146、nt and intelligent connection,and the scenario-driven network resource scheduling and management of terminal services.Therefore,the intent-driven next-generation RAN control architecture is proposed,including the application layer,orchestration and control layer,and infrastructure layer,as shown in

147、Figure 2.Figure 2 Intent-Driven Next-Generation RAN Control Architecture-23-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0The protocol control architecture diagram shown in Figure 2 consists of three layers and two interfaces:the application layer,orchestration a

148、nd control layer,and infrastructure layer.To successfully implement these emerging services,the application layer allows users to declare intent on-demand in various ways,providing a recognizable,translatable,and reusable specification language.In addition,intent can be input externally,which means

149、that users can input intent in the form of text,voice,and other forms.Another way of intent can be internal intent,which means that when a fault occurs during network operation,the underlying system will feedback to the user interface,and the intent needs to be retranslated again to generate new pol

150、icies that conform to the current network state.Next,different service requirements are mapped to specific communication metrics(QoS metrics such as latency and bandwidth).Then,various basic resources are coordinated and controlled,such as wireless resources,computing resources,and storage resources

151、.Finally,network slicing is dynamically constructed,and a deterministic network is established to provide users with differentiated service experiences.The orchestration and control layer includes a resource control module,an intent management module,and a service orchestration module.The users serv

152、ice intent is sent from the northbound interface to the intent management module,which translates the users natural language intent expressed in the application layer into an intent consisting of objects,operations,and results.The translated intent is then sent to the service orchestration module,wh

153、ich collaboratively orchestrates and allocates resources based on the specific communication indicators translated.Finally,the resource control module manages and controls the virtual resources at the bottom layer based on the orchestration results,while monitoring the bottom layer resources in real

154、-time and feeding back the results of the intent to the orchestration module.In this case,DTN technology is adopted to deal with virtual resources.A unified data model is constructed by using AI/ML algorithms and big data analysis.This enables efficient network information collection,which is then s

155、tored in a unified data warehouse for effective management.The infrastructure layer should achieve integrated space-air-sea-ground global wireless access to achieve the performance goals of full coverage,huge capacity,huge connections,ultra-low latency,and high self-organization for 6G networks.A sp

156、ace network based on satellite communication provides wireless coverage for areas without ground network coverage through the dense deployment of orbital satellites.Low-altitude platforms of air networks can be deployed faster-24-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and

157、 Function 5.0and reconfigured more flexibly to best suit the communication environment and exhibit better performance in short-distance communication.High-altitude platforms of air networks can serve as relay nodes in long-distance communication to facilitate the integration of ground and non-ground

158、 networks.The ground network will support the terahertz frequency band,and its extremely small network coverage range will reach the limit of system capacity.The network architecture of going beyond cellular and user-centric ultra-dense networks will emerge.The underwater network will provide covera

159、ge and Internet services for wide-ranging and deep-sea military or commercial activities,but there is controversy over whether the underwater network can be a part of future 6G networks.Therefore,the designed infrastructure layer of the intent-driven 6G network currently adopts an air-space-ground n

160、etwork.3.3 Key Technologies1.Intent TranslationIn the intent translation module,the management system will convert tenants high-level and abstract slice intents into corresponding slice policies.The intent translation module is divided into three parts to better explain the intent translation proces

161、s:1)Intent Manager Front End;2)Intent Manager Back End and Policy Library;3)Northbound Interface of RAN-side controller FlexRAN and core network-side orchestrator OSM.See Figure 3.Figure 3 Intent Translation Process-25-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5

162、.0As shown in Figure 3,the intent translation and interaction process of the intent layer exists between the application(including Intent Manager Front End GUI,Intent Manager Back End module,and policy library)and the open NBI of FlexRAN/OSM,mainly involving the use of intent for service scenario sl

163、icing,QoS based on service scenarios(latency,throughput),and customized requirements.(1)Tenants provide their desired leased service scenarios(video streaming services,VR,etc.)on the GUI.(2)Based on this information,the back end retrieves the standard QoS requirements that meet this service scenario

164、 in the policy library,which includes end-to-end network-related QoS information such as latency and throughput for the service provided.(3)The back end passes the obtained information to the policy configurator,which has a slice template for interacting with the NBI of FlexRAN/OSM.The RAN/core netw

165、ork-side policy configurator is responsible for initializing this JSON file and issuing initial static configuration through the slice API of the unit configuration API of FlexRAN/OSM.Example JSON request message of slice API:dl:algorithm:Static,slices:id:0,static:posLow:4,posHigh:12.(4)After obtain

166、ing the required information,the controller deploys the initial static parameters to the underlying resources through FlexRAN proxy,slices the MAC layer of a single base station logically,and completes the initial slice creation.2 Smart Policy MappingThe smart policy mapping module is the process of

167、 achieving intents,consisting of a policy library,fuzzy decision tree,and deep Q network(DQN).The policy library stores a large number of atomic policies established based on historical policies and administrator O&P experience.Atomic policies refer to the smallest policies that can be executed and

168、cannot be further divided.For example,it can implement configuration policies for some functional nodes(such as routing selection,selection of service provider node,etc.).The function of the fuzzy decision tree is to generate new atomic policies for the policy library or adjust existing atomic polic

169、ies in the policy library.DQN is a combination of neural networks and reinforcement learning,used to recombine atomic policies into new policies that meet the current intent requirements.The neural networks function is to calculate the corresponding configuration scores of state-action pairs and out

170、put all actions.Then,according to the Q-learning principle,select the action with the highest configuration score as the-26-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0configuration action.When the intent is input from the intent translation module to the modul

171、e of the smart policy mapping,its working mechanism is described as follows.1)First,check whether there is an atomic policy combination in the policy library that meets the current intent requirements.If it exists,proceed to the next step;otherwise,use the fuzzy decision tree to generate new atomic

172、policies or adjust existing atomic policies to expand the policy library.2)The network state is used as the input S of the neural network,and DQN analyzes the configuration rewards of the combination policy through the neural network.Then,according to the Q-learning principle,the action with the max

173、imum value is output as the next action to be taken.3)Validate the combined policy.If the selected policy can be successfully executed under intent constraints,modify the network environment state,and the agent receives a reward R;if it cannot be successfully executed,the state S maintains its stage

174、 and repeats step 2 until the reward converges.4)The network gives feedback on the action and gets the next state s.Figure 4 Smart Policy GenerationThe above process is iterated continuously to obtain the optimal policy that achieves the goal.Combined with the policy library model,this policy is sup

175、plemented to the policy library,making the policy library more complete.In summary,the advantage of machine-driven smart policy-27-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0Figure 5 Standard Process of Intent-driven Networksgeneration is that administrators d

176、o not need to manually test in a large number of parameter and state space,and it improves network management efficiency.3 Intent Verification TechnologyIDN provides full lifecycle management for network elements under the condition of collecting network state.Unlike traditional networks that rely o

177、n specific command-line inputs,under the driving of intents,people do not need to directly input policy commands,but only input the desired business intent,that is,I hope the network can reach a certain state or provide high-quality service.Network administrators do not need to focus on network deta

178、ils or implementation technology,but only express their requirements.IDN can automatically translate the intent and complete subsequent operations.Then,IDN will real-time verify whether the network state meets the expected business intent.Therefore,verification is one of the core technologies in int

179、ent-driven networks.The intention is translated to obtain the corresponding network policy,which cannot be directly issued to the controller for execution.In order to ensure the continuous operation of the network,the correctness of the policy must be verified before the policy is issued.The network

180、s complexity and the policy execution guarantee require that intent-driven networks have the function of verification,and each language expression of intent requires verification technology to ensure correctness.The intent verification module will check the accuracy and completeness of parsing the i

181、ntent,and continuously monitor the possible conflicts in parsing the intent.For example,the network cannot meet the resource requirements of multiple intents issued simultaneously,and logical errors such as error endpoint groups and wrong time requirements in parsing the intent occur.Furthermore,the

182、 intent verification verifies whether the translated results can fully match the users original intent and feeds back the verification results to the user.Verification is not a one-step process.Intent has different expressions in different stages.As shown in Figure 5,high-level policies undergo mult

183、iple conversions to lower-level policies,and there are intent verification issues for each conversion.-28-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0After the intent application of the application layer initiates the request,the formal intent is issued to the

184、controller through the intent northbound interface(Intent NBI),and the intent verification module also collects the intent through the Intent NBI to form the intent table I composed of historical intents.After the controller receives the intent,it formulates the policy.On the one hand,the policy nee

185、ds to be generated based on the intent,and on the other hand,it needs to be modified based on the feedback from the intent verification module.The generated policy is issued to the underlying switch through the southbound interface(SBI),and the policy and the current network state issued during this

186、 policy deployment process are also collected into the intent verification module through SBI to form policy tables and resource tables,respectively.The switch can be divided into a switch software layer(OpenFlow agent)and a switch hardware layer(OpenFlow pipeline).A traffic collection module is add

187、ed in the OpenFlow pipeline to collect the actual flow forwarding behavior of the network in real-time after the policy is issued and report it to the intent verification module to form a behavior table.In the intent verification module,various collection behaviors in the above process form the inte

188、nt table I that records intent requests,the resource table that records network state,the policy table that records each policy deployment,and the behavior table that records the forwarding behavior actual network packet.Proposed verification method:(1)feasibility verification:the verification of an

189、d ensures the correctness of intent translation,the verification of and ensures that the policy issued meets the underlying network constraints,and the verification of ensures that the policy is conflict-free;(2)validity verification:the verification of and,and the verification of and guarantees tha

190、t the actual network flow forwarding behavior conforms to the specified policy and ensures the effectiveness of the intent.In the verification process,if verification failure occurs,the policy will be modified and reissued through policy feedback.3.4 Intent Lifecycle ManagementFigure 6 illustrates t

191、he intent lifecycle management,which includes two loops that jointly realize the creation to deletion of user intents,and ensure the continuity of user intents through intent assurance in case of faults.This dual-loop structure translates the users natural language into a configuration language that

192、 the underlying system can recognize through intent translation,and monitors the operation of the service intent in real-time through state management,and finally-29-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0changes and protects the intent based on the monito

193、ring results.Through continuous closed-loop management of the intent lifecycle,the generation of effective policies and the reliability of intent are ensured.Figure 6 Intent Lifecycle ManagementIn the loop above,policy configuration is used to determine how the system should respond to user inputs r

194、elated to intent.For example,policy configuration can specify how to retrieve and present information related to user requests.Then,the policy can be instantiated on the infrastructure layer to implement the intent.State monitoring is used to ensure that the intent is executed as expected and that p

195、olicies provide accurate and useful information to the user.In the loop below,the same modules are used to update and improve intent based on user feedback and performance indicators.The policy configuration module is used to adjust user intent-30-The Next-generation Protocol Stack-RAN Architecture,

196、Protocol Stack and Function 5.0in response to changes in underlying state information.Distribution and instantiation are used to make any changes or updates to the intent in the infrastructure layer.State monitoring is used to track the impact of these changes and ensure that the intent continues to

197、 meet user requirements.The lifecycle management of intent(dual-loop method)has several advantages:1.Enhanced user experience:By monitoring user feedback and updating policies and logic accordingly,the lifecycle of intent can be optimized to provide more tailored and effective user experiences.2.Eff

198、icient intent maintenance:The dual-loop method allows for continuous monitoring and maintenance of intent,reducing the possibility of errors and ensuring that intent continues to meet user needs.For example,if there are errors in the intent,they can be quickly identified and fixed through continuous

199、 monitoring.3.Improved intent accuracy:The dual-loop method enables continuous feedback and updates to improve intent accuracy and relevance.This can lead to more accurate and relevant responses to user input,improving the overall user experience.4.Better resource allocation:By tracking intent perfo

200、rmance,resource allocation can be optimized to focus on the most frequently used or impactful intent.Overall,the dual-loop lifecycle management of intent helps ensure that intent is optimized for user needs and provides a better overall user experience.3.5 SummaryFor the next generation RAN,intent-d

201、riven networks provide the capability of deep mining of application intent,global sensing of network state,and real-time optimization of network configuration,realizing a programmable and customizable automated network.By proposing an intent-driven next-generation RAN control protocol architecture,i

202、ncluding modules such as intent translation,policy configuration,and resource orchestration,and implementing real-time state monitoring,it achieves the full life cycle management of user intent and provides a new feasible approach for next-generation RAN control protocols.-31-The Next-generation Pro

203、tocol Stack-RAN Architecture,Protocol Stack and Function 5.0Reference1 NGMN White Paper,6G Drivers and Vision.2 3GPP TS 36.300:NR and NG-RAN Overall Description 3 3GPP TS 38.300:Evolved Universal Terrestrial Radio Access(E-UTRA)and Evolved Universal Terrestrial Radio Access Network 4 3GPP TS 38.300:

204、System architecture for the 5G System(5GS)5 TS 38.455,NG-RAN;NR Positioning Protocol A(NRPPa),3GPP6 IEEE 802.11bf:Toward Ubiquitous Wi-Fi Sensing7 Open Networking Foundation,Intent NBI:Definition and Principles.8 IEEE Wireless Communications,Vision,requirements,and technology trend of 6G:How to tack

205、le the challenges of system coverage,capacity,user data-rate and movement speed.9 3GPP TR 28.812,Telecommunication management;Study on scenarios for Intent driven management services for mobile networks.10 Internet research task force,draft-irtf-nmrg-ibn-concepts-definitions.-32-The Next-generation

206、Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0Abbreviation 3G 5G 6G AI BWP CU GoS HTC IoT IP ISLs MAC MCG ML NRSP NAS OSFC PHY PDCP PDU RB RLC SCG PCell SDN UE UDN IDN TMN PBNM IDNM IETF QoS DTN NBI SBI DQN the Third-Generation Mobile Communicationsthe Fifth-Generation Mobile Commun

207、ications the Sixth-Generation Mobile Communications Artificial intelligence Bandwidth Part Central Unit Grade of servic Holographic-type communication Internet-of-things Internet Protocol Inter-satellite links Medium Access Control Master Cell Group Machine learning NR sensing protocol Non-access la

208、yer Orchestratable Stack-Free Component-based packet forwarding protocol Physical Packet Data Convergence Protocol Protocol Data UnitRadio Bearer Radio link control Secondary Cell Group Primary CellSoft Definition Network User Equipment Ultra-Dense Network Intent-driven Netwok Telecommunication Mana

209、gement Network Policy-based Network ManagementIntent-driven Networks Management Internet Engineering Task Force Quality of Service Digital Twins NetworkNorthBound Interface SouthBound Interface Deep Q-Network -33-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0Ackn

210、owledgementGrateful thanks to the following contributors for their wonderful work on this whitepaper:Editors:CSCN XDU CMCC Contributors:Sun JunshuaiChungang YangChih-Lin I,Qixing Wang-34-The Next-generation Protocol Stack-RAN Architecture,Protocol Stack and Function 5.0FuTURE FORUM is committed to c

211、utting edge technologies study and applications.Controversies on some technical road-maps and methodologies may arise from time to time.FuTURE FORUM encourages open discussion and exchange of ideas at all levels.The White Paper released by FuTURE FORUM represents the opinions which were agreed upon

212、by all participating organizations and were supported by the majority of FuTURE FORUM members.The opinions contained in the White Paper does not necessarily represent a unanimous agreement of all FuTURE FORUM members.FuTURE FORUM welcomes all experts and scholars active participation in follow-on working group meetings and workshops.we also highly appreciate your valuable contribution to the FuTURE White Paper series.