1EvolvedRandomAccessandMultipleAccessTransmission TechnologiesWhite Paper2021.12.212摘要摘要6G 将通过万物互联/智联真正全面实现数字化社会。面向 6G 的多项关键技术研究工作正在开展中，作为空口增强的核心技术点，多址接入技术是 6G 研究热点之一。多址接入技术包括随机接入技术和多址传输技术。随机接入技术和多址传输技术分别服务于无线通信中的初始接入过程和数据传输过程，在 6G 系统中都将进一步演进，甚至融合形成新型的多址接入技术。面向多址接入技术及其演进，这本白皮书将分析多个 6G 中可能出现的新应用场景，将分析多址接入技术的基础理论进展。基于上述 6G 新场景和多址接入新理论，为了满足 6G 多元化的性能指标，包括提升多址接入的终端数量、提高接入和传输的效率、降低终端的功耗、降低通信的时延等，这本白皮书将着重研究以下演进的多址接入技术：分析现有初始接入过程的限制性因素，基于 4-Step RACH 技术、2-StepRACH 技术等，研究演进的随机接入技术；分析现有数据传输过程的限制性因素，基于 OFDMA 技术、NOMA 技术等，研究演进的多址传输技术；把初始接入过程和数据传输过程进行融合，研究随机接入技术和多址传输技术的新型融合技术。目前 6G 研究仍在探索之中，6G 系统将会拥有新资源、新能力，这本白皮书的主要思路，是基于经典的通信系统设计思想，对 6G 系统开展基础性研究，探索提升 6G 系统通信能力的关键技术。我们希望这种思路和这本白皮书能对 6G研究和 6G 产业有所贡献。3Executive Summary6G will truly and comprehensively realize a digital society through the Internet ofeverything or intelligent connection of everything.A number of 6G-oriented key technologyresearch work is ongoing.As the core technology point of air interface enhancement,randomaccess and multiple access transmission technology is one of the 6G research hotspots.For random access and multiple access transmission technology,random accesstechnology and multiple access transmission technology serve the initial access process anddata transmission process in wireless communication respectively.They will further evolveand even integrate into a new random access and multiple access transmission technology in6G system.For random access and multiple access transmission technology and its evolution,thiswhite paper will analyze multiple new application scenarios that may appear in 6G,and willanalyze the basic theory progress.Based on the above-mentioned new 6G scenarios and thenew theory progress,in order to meet the diversified key performance indicators of 6G,including increasing the number of terminals,improving access and transmission efficiency,reducing terminal power consumption,and reducing communication delay,etc.,this whitepaper will focus on the following evolved random access and multiple access transmissiontechnologies:Analyze the restrictive factors of the existing initial access process,and study theevolved random access technologies based on the current 4-step RACH technologyand 2-step RACH technology.Analyze the restrictive factors of the existing data transmission process,and studythe evolved multiple access transmission technology based on OFDMA technologyand NOMAtechnology.Integrate the initial access process and the data transmission process,and study thenew convergence technology of random access technology and multiple accesstransmission technology.At present,6G research is still being explored.The 6G system will have new resourcesand new capabilities.The main philosophy of this white paper is to explore the keytechnologies to improve the communication capability of 6G system based on the classiccommunication system design idea.We hope this philosophy and this white paper cancontribute to 6G research and 6G industry.4Table of Contents1.INTRODUCTION.51.1 BACKGROUND.51.2 NEWAPPLICATIONSCENARIOS.51.3ADVANCES INFUNDAMENTALTHEORY.91.4 EVOLUTIONROUTES.112.EVOLVED RANDOM ACCESS TECHNOLOGIES.132.1 LIMITATION OFEXISTINGRANDOMACCESSTECHNOLOGIES.132.2 FASTCONNECTION-LIKESTATEESTABLISHMENT.142.3 SLOTTEDALOHA ENHANCEDTECHNOLOGIES.172.4 BRIEFSUMMARY.203.EVOLVED MULTIPLE ACCESS TRANSMISSION TECHNOLOGIES.213.1 THEPROBLEMS ANDEVOLUTION OF5G NOMA.213.2 MILLIMETERWAVEMIMO-NOMATRANSMISSIONTECHNOLOGY.243.3 TANDEMSPREADINGMULTIPLEACCESS(TSMA)TECHNOLOGY.263.4 LATTICEPARTITIONMULTIPLEACCESS(LPMA)TECHNOLOGY.293.5 MCFTN-NOMATECHNOLOGY.303.6 EVOLVEDSCMATECHNOLOGY.313.7 BRIEFSUMMARY.344.INTEGRATION OF RANDOM ACCESS AND MULTIPLE ACCESS TRANSMISSION.354.1 NECESSITY OFINTEGRATION OFRANDOMACCESS ANDMULTIPLEACCESSTRANSMISSION.354.2 EXTREMELYSIMPLEGRANT-FREETRANSMISSIONSCHEME.384.3 DECENTRALIZEDGRANT-FREETRANSMISSIONSCHEME.404.4 CONCATENATEDCODESCHEME FORMASSIVERANDOMACCESS.424.5 BRIEFSUMMARY.455.CONCLUSIONS AND DEVELOPMENT TRENDS.47REFERENCES.49致谢致谢.51ACKNOWLEDGEMENTS.5251.Introduction1.1 BackgroundThedevelopmentofmobilecommunicationhasexperiencedtheprocessof1G/2G/3G/4G,and has entered the 5G commercial stage.From the perspective of the law oftime evolution,it basically follows the law of ten-year generation.The improvement of userdemand and the development of communication technology are the driving force for thedevelopment of mobile communications.The large-scale commercial use of 3G and 4G hasdriven the vigorous development of the mobile Internet market and applications,creating abrilliant mobile Internet era.5G and 6G will truly realize the digitization of industry andsociety,and greatly improve the efficiency of social operations.With the global deploymentof 5G commercialization,the ITU launched the research on the trend and vision of IMTtechnology for 2030 and the future in February 2020,which is expected to be completed in2022.China established the IMT-2030 Promotion Group in 2019 to carry out 6G-relatedresearch work.The current research work for 6G 1 is mainly to introduce new resources and newcapabilities,such as THz technology,reconfigurable intelligent surface technology andintegrated sensing and communication technology.In the development process from 1G to 5G,the research work is mainly focused on the improvement of communication efficiency and thereduction of transmission delay.Therefore,while introducing new resources and newcapabilities for 6G,it is also necessary to enhance the air interface technology such as toincreasethenumberofsupportedterminals,reducepowerconsumption,improvecommunication efficiency,and reduce transmission delay.Random access and data transmission are the basic processes of mobile communicationsystem.Its design idea is to make a compromise between implementation complexity andtransmission efficiency.In 5G,random access technology evolves from 4-step RACH processto 2-step RACH process,and data transmission technology centers on OFDMA and MIMO.In 6G,the industry generally believes that random access and multiple access transmissiontechnology scheme will be optimized and evolved.Therefore,random access and multipleaccess transmission technology may become one of the key technologies of 6G.1.2 NewApplication ScenariosThe new application scenarios of random access and multiple access transissiontechnology in 6G include:mMTC scenario with huge number of terminals,critical MTCscenarios such as the V2X,industrial IoT,etc.with stringent latency requirement,space-air-ground integration scenario with large interaction time delay,and eMBB scenariorequiring balance between spectral efficiency and fairness.This section will analyze theaccess and transmission characteristics of terminals in these application scenarios,including6the requirements of access and transmission success rate,dynamic range of packet size,dynamic range of uplink and downlink bandwidths,regular transmission and bursttransmission,and delay requirement,etc.To meet the new requirements of these applicationscenarios,it is necessary to design and implement new random access and multiple accesstransmission technologies in the 6G network.(1)mMTC scenario6G will be a network that can cover the communication demands of human to human,human to thing,thing to thing,and intelligence and intelligence,and fully empowereverything.The demand for the number of connections for such a brand-new communicationsystem may be unimaginable at present.The forecast data of the 6G connectivity indicatorsmentioned in the current mainstream white papers on future networks will be 10-100 timeshigher than that of 5G,and3D coverage is also considered.How to efficiently(includinghigh spectral efficiency,high energy efficiency,low latency/high time efficiency,lowcomplexity/high efficiency of hardware implementation)support the future huge connectionrequires breakthroughs in new technologies,and the research of potential enablingtechnologies is imminent.The research of enabling technology needs to be based on thecharacteristics of application scenarios.The main features of some foreseeable massiveterminals scenarios are:The number of machine-type terminals is extremely huge,for example,it will be10-100 times larger than that in 5G,and 3D coverage will be considered.The traffic type of the terminals is mainly information reporting.Although in mostcases,each of these individual terminals will only transmit small data portions verysporadically,it was observed that for some applications the group behavior followsone of a botnet 2,i.e.a large amount of devices will access the networksimultaneously,which is obviously different from human-oriented interactivecommunication.Some services have low requirement for transmission delay.The cost and power consumption of the terminals are required to be very low.Current wireless systems cannot meet these requirements of extremely large number ofterminals(10-100 times larger than that in 5G).The transmission of the current wirelesssystems is scheduling-based,and complicated interaction process is required before the datatransmission.If the above scenarios have to be carried on the scheduling-based system,it isbound to cause that the number of access terminals is far from meeting the requirements,thesignaling overhead is unacceptable,the terminal cost remains high,and the powerconsumption,especially the terminal power consumption,cannot be reduced by an order ofmagnitude.Therefore,it is necessary to design a new multiple access technology to meet theabove requirements.(2)Critical MTC scenario7The critical MTC scenario can be said to be a compromise between mMTC and uRLLC.The number of terminals and the requirements for transmission delay and reliability arebetween mMTC and uRLLC.It should be noted that when critical MTC terminals are widelydeployed,although the number of terminals is not comparable to that in the mMTC scenario,it will be much greater than that in uRLLC scenario.Although uRLLC can already support asmall number of ultra-high-reliability and low-latency terminals,how to efficiently supportcritical MTC scenarios with a much larger number of terminals requires further research.Specifically,for periodical critical MTC services,the traditional orthogonal access methodbased on semi-persistent periodical resource reservation(SPS or configured grant)can beconsidered as an efficient dynamic scheduling free and low-latency access method.Periodicresources reservations that matches the service generation periods,not only ensure efficientresource utilization,but also guarantee the performance because each user can exclusivelyoccupy resources without inter-user interference.However,for event-triggered critical MTCservices,the traditional periodically reserved orthogonal resources are often not fully utilized.The longer the period of the reserved orthogonal resources,the higher the utilization rate ofthe reserved resources.However,the average waiting time of terminal access is longer,thus itis more difficult to guarantee ultra-low latency requirement.In contrast,the shorter the periodof the reserved orthogonal resources,the easier to guarantee ultra-low latency requirement,but the lower the resource utilization.To improve the utilization of periodic reservedresources,a natural method is to allow multiple users to use shared resources.In this way,thevacancy rate of resources will be greatly reduced,but multiple users will collide,which is asevere challenge to reliability.Therefore,it is necessary to design a new multiple accesstechnology to meet the above requirements.(3)V2X scenarioEnhanced Vehicle to Everything(eV2X),as a key technology of the future intelligenttraffic and transportation system,provides technical support for opening up the informationbarrier between vehicles and vehicles,and between vehicles and base stations.Ref.3proposed 25 application scenarios for this technology.To adapt to the changing applicationscenarios,eV2X considers supporting low latency,high reliability,high data rate,anddynamically changing transmission range.The technical routes for eV2X to meet multiplerequirements include LTE-based eV2X and NR-based eV2X.V2X can be C-V2X based oncellular system and D2D-V2V based on direct connection between vehicles.Both C-V2X andD2D-V2V have low latency and high reliability transmission services.Further,with thepopularization of the V2X technology,the number of V2X terminals will be much larger inthe future.Combining these two features,the future V2X can be considered as a specificscenario belonging to critical MTC,but its own features are relatively obvious and listedseparately.C-V2X has the following features:Doppler effect is serious because vehicles move fast and the transmission channelschange quickly.Fast moving of vehicle nodes causes frequent handovers.8The vehicle moves fast,and the reliability of the terminals at the edge of the cell ispoor due to the interference from neighbor cells.The traffic loads of different cells are uneven.For example,the density of vehicleson busy road sections is much higher than that on non-busy road sections,and thefrequency of emergency events is much higher.D2D-V2X has the following features:The D2D connection topology changes rapidly due to the fast moving of the vehiclenodes.The density of vehicles on busy road sections is very high,and there are manyemergency events.Therefore,collisions caused by contention-based resourceselection during D2D transmissions may be serious.Doppler effect is serious because vehicles move fast and the transmission channelschange quickly.Vehicles are much larger than common communication terminals.The information of closer vehicles is more important.The two features of high density V2X are massive and burst information transmissions,and the information transmission of V2X also requires low-latency and high-reliability.It is abig challenge to meet the low-latency and high-reliability requirements in massive and burstinformation transmission scenarios.Furthermore,as the vehicle nodes move fast,the networktopology of the V2X changes rapidly,making it more difficult to meet the requirements ofmassive and burst transmissions,low latency,and high reliability at the same time.Therefore,it is necessary to design a new multiple access technology to meet the above requirements.(4)Space-air-ground integrated network(SAGIN)The industry has begun to discuss the vision and concept of 6G networks,and the keyrole of satellite communications in 6G networks has become a consensus.Satellitecommunications in 6G network will include GEO,MEO,and LEO satellites.6G network willbe more complex than 5G network and any previous generation of mobile communicationsystems.The future air-space-ground integration network(SAGIN)is actually a specificscenario of mMTC,but its own features are obvious,and listed separately as follows:Large coverage area,for example,the coverage diameter of a low-orbit satellitebeam can reach 400 km.With such a large coverage area,the number of terminalsis very large.Long transmission distance,long transmission delay,and large interaction delay.If there is no timing advance(TA)process,or TAprocess is inaccurate,the timingerror will be very large duo to the long transmission distance.9The frequency deviation/Doppler effect is large because the satellite moves at ahigh speed,and the transmission frequency is high.Ensuring the reliable access of massive connections,improving the utilization oflimited spectrum resources and supporting various technical indicators of 6Gnetwork are important issues faced by 6G satellite communication network.If the above scenarios are carried on the scheduling-based system,multiple signalinginteractions are required between the ground terminals and the air satellites,causingintolerable delay,and normal communication cannot be implemented.Therefore,it isnecessary to design a new multiple access technology to meet the above requirements.(5)eMBB scenarioFrom 1G to 5G,to meet the increasing demand for capacity,multiple access technologieshave been greatly improved,and FDMA,TDMA,CDMA and OFDMA were born.The firstthree technologies use the characteristics of frequency domain,time domain,and codedomain to distinguish users respectively,while OFDMA uses both time domain and frequencydomain.New services in 6G,such as XR and holographic communication,will furtherincrease the demand for capacity.With the continuous improvement of orthogonal multipleaccess technology,the improvement space of spectral efficiency is getting smaller and smaller.Many research results show that NOMA technology can effectively improve the spectralefficiency under the premise of ensuring the fairness of near-far users.That is to say,in thescenario where more emphasis is placed on the throughput and spectrum efficiency at the celledge,the NOMA technology has greater potential.Furthermore,as the complexity and gain ofthe massive MIMO technology reach a compromise,further increasing the number ofantennas and adopting more complex MIMO detection algorithms cannot obtain the gainefficiently.At this time,NOMA can become the main source of gain.NOMAs priority in 6Gdesign will be improved.1.3Advances in Fundamental TheoryMRA(Massive Random-Access)is the novel theoretical research proposed byPolyanskiy in MIT in 2017 4,which can support the access and transmission of massiveterminals without network coordination.The MRA technology includes three ideas of design:(1)all terminals share a common codebook,the payload packets of which are transmitted onthe shared channel after coding,and all terminals have the same transmission mode;(2)takePUPE(error probability per user)as the optimization goal,instead of system throughput,tosupport a large number of terminals;(3)The information length of all terminals is fixed,whilethe codeword length is finite,which introduces the finite block-length coding effect.The finalobservation on the performance of MRA from this theoretical research is that when thenumber of users is less than a certain threshold,the energy per bit required to meet the PUPEstandard barely changes with the increase of the active number of users,indicating that theuser interference can be canceled ideally.The performance bound of MRA is very close to the Shannon limit.When the number of10users is small,the required Eb/N0 is almost constant,but with the rapid increase of activeusers,it gradually approaches the Shannon limit,indicating the advantage of MRAtechnology in supporting a massive number of users.The performance bound of MRA isslightly better than that of OMA.Meanwhile,its important to note that,unlike MRA,OMArequires network coordination.There is an obvious gap between the performance oftraditional NOMA technology and that of MRA,especially when the number of users goeshigher.MRA is mainly oriented to the massive machine-type communication scenarios requiringlarge-scale,short packet,and low delay transmission 5.Compared with traditionalcoordinated multiple access schemes,the most prominent feature of the MRA scenario is theremoval of the coordinating center and the coordination process,as shown in Figure 1.Figure 1.Comparison between coordinated and uncoordinated multiple accessIn coordinated multiple access systems(including OMA and coordinated NOMA),thecoordinating center is responsible for managing the behavior of the user payloads in thechannel and identifying the users through certain information.Therefore,at the receiving end,the users collision mode is perceptible.Resource allocation of orthogonal multiple access,together with the power,codebook,and pattern allocation of non-orthogonal multiple access,all belong to the category of coordinating information.However,MRA does not require thecoordination centers presence.The collision of user payloads in the channel is randomlydistributed,with no transmission of coordinating information,so there is no extra cost ofresources,and processing delay is lower.At the same time,compared with the design methodof coordinated systems aiming at capacity domain optimization,the optimization goal of the11MRA system is changed to Per User Probability of Error(PUPE),namely,the average packetloss rate of the system.Polyanskiy theoretically proved that when the number of active usersin a coordinated multiple access system increases dramatically,the sum capacity can continueto grow;conversely,the capacity of subchannels per user tends to zero.Thus,taking PUPE asthe optimization criteria ensures the reliability of each users sub-channel preferentially,whichis more advantageous in the scenario of massive access.For OMA,the cost of network coordination signaling and interaction delay limits theincrease of supportable number of terminals.The analysis shows that the number of terminalswithout network coordination under MRA is greater than that of orthogonal multiple access.Thus,MRAis a competitive technology for 6G.1.4 Evolution RoutesThe random access and multiple access transmission technology in 6G will be optimizedand evolved.We believe that random access technology and multiple access transmissiontechnology will evolve respectively,and they will also be integrated to meet the requirementsof new scenarios in 6G.Figure 2 shows the evolution route of random access and multipleaccess transmission technology in 6G.Figure 2.Evolution route of multiple access technologyFor the initial access process,4G mainly adopts four-step random access technology(4-step RACH).After completing the initial access process,the terminal user initiates the datatransmission process through network scheduling.In order to reduce the access delay andmeet the requirements of uRLLC services,5G introduces two-step random access technology(2-step RACH).In 6G,random access technology will continue to evolve(Evolved RACH),including the optimization of both the establishment process and maintenance process of theconnection state,the simplification of procedures and the reduction of signaling overhead.Regarding the data transmission process,4G/5G uses OFDMA technology as its basic12scheme for multiple access transmission.In the process of research and standardization,attempts have been made to introduce NOMA technology,but no agreement has been reached.In 6G,NOMAtechnology may be adopted to improve spectrum efficiency and support a largenumber of terminals.Multiple access transmission technology will continue to evolve(Evolved MA),including the enhancement of multiple access discrimination and thereduction of signaling overhead.The initial access process and data transmission process can be integrated into oneprocess and jointly optimized to form a new random access and multiple access transmissiontechnology,which is expected to increase the number of users,increase the success rate ofaccess and transmission,and reduce communication delay.Based on the idea ofuncoordinated random access and transmission technology,the convergence technology ofrandom access and multiple access transmission will evolve to an integrated technology.With this integrated technology,when there is no or only a small amount of networkcoordination signaling,all terminals adopt the same transmission processing method,and thebase station can correctly decode the data sent by each terminal.132.Evolved Random Access TechnologiesThe cellular wireless network implements the initial access through random accessprocess,allowing users to choose a suitable network to reside on,receive necessary systeminformation,register user identity information,achieve uplink time synchronization,andrequest transmission resources to complete data transmission.2.1 Limitation of Existing RandomAccess TechnologiesInitial access is one of key building blocks for the cellular access networking to establishconnection state and facilitate network management and control.From 1G to 5G,the initialaccess procedure is designed to enable connection establishment for a relatively low numberof accessing devices.Additionally,each device has moderate to high data-rate requirementssuch that the establishing overhead for connection state is relatively small.Both assumptions,the low number of devices as well as moderate to high data rates are in contradiction toscalable cMTC(critical MTC)needs arising from massive machine-type terminals in IIoT.By 2030,our societies will become digitalized and data-driven,enabled through the keyverticals like connected industries,intelligent transport systems and smart cities.In thecoming decade,owing to emerging industrial use cases and the verticalization of the serviceprovision,several specialized subclasses will develop,hence demanding multi-dimensionaloptimization and scalable designs.So 6G will need to serve highly diverse applicationsrangingfromdata-ratehungryholographicimagesandconnected360XR(augmented/virtual/mixed reality)to massive access for various types of IoT devices.One ofMTC service classes for 6G is proposed to be classified as Scalable cMTC,which refers tosupporting massive connectivity with high reliability and low latency,e.g.critical medicalmonitoring and factory automation.Scalability and flexibility will continue to be paramountmeasures for 6G performance.6G is expected to support as high connection density as 10million devices/km26 7.Scalable cMTC generally possesses two distinctive features:Massive access:The overall system needs to support massive connectivitythe numberof devices connected to a cellular base-station(BS)may be in the order of 104to 107.Themacro BS prefers to provide a unified massive access for various types of IoT devices,offering a low-cost solution for supporting massive connectivity with high reliability andlow latency.Sporadicity of traffic pattern:At any given time only a small fraction of potentialdevices are active.Typically,the machine-type devices connect asynchronously andsporadically to a network to send small data payloads.The sporadicity is due to theinherent burstiness of event-driven IoT communications in controlled and/or sensingenvironments.Most of devices make random requests independently,less periodicity canbe tracked and utilized.As such,it is impossible for the network to predict when andwhich device will deliver packet in advance.14The 6G calls for a more radical redesign of the access,the random access procedure andthe definition of connection state should be further refined.The conventional random accessis unscalable for massive and critical MTC owing to the following limitations:Lack of preambles for PRACH(Physical Random Access Channel):The PRACHmechanism is an variant of the classic ALOHA,which imposes a limit on the number ofactive devices that are granted to access the network.On the other hand,a certaincoherence block just can support a fewer number of orthogonal preamble sequencesrelative to the massive number of machine-type devices.Massive devices performrandom access by independently picking one sequence out of the same small set,inevitably causing serious collision and incurring intolerable access delay.Ironically,therepeating cycles of transmission-collision-retransmission lead to an endless cascade ofsignaling exchange between device and BS,which is very higher relative to a smallpacket a machine-type device intends to send.Highcostformaintainingconnectionstate:Maintainingconnectionsatesimultaneously for the massive devices having potential service requests sustainsrelentless exchanges of periodic signaling,coming at the unacceptable waste in powerand spectrum.It will become unsustainable,as connection density increases to 10 milliondevices/km2.Alongside,maintaining continuous connection state is energy inefficient fora IoT device itself,the device is usually expected to have long battery life more than 10years.If a BS has the capability of fast connection-state establishment in an on-demandfashion,the maintaining cost can be avoided.Moreover,the conventional initial random access introduces an extra access latency of20 ms for establishing connection state for UEs 8.The 5G just accomplishes 1ms latencyon the user plane where the connection state is assumed to be held,whereas the latency on thecontrol plane lacks enough emphasis and optimization.At a certain transmission cycle or timeframe,the user equipment being active forms a random active subset to be identified out ofthe population of potential user equipment.Accordingly,user activity detection is necessaryfor connection state establishment.For massive access,an eligible user activity detectionmust be accurate,fast,high efficiency and scalable.The amount of the required measurementcost should be comparable to the average number of active devices,not being scaled upsuper-linearly by the population of massive devices.2.2 Fast Connection-like State EstablishmentIn contrast to the existing connection state establishment,the delay-sensitive massiveaccess should define a connection-like state that has simple network function,lowmaintenance cost,and can be quickly established on demand.The 6G should further simplifythe initial access procedure.It expects to adopt new method of user signature design,allocation and acquisition,allowing for fast connection-like state establishment,avoidingmaintenance for massive number of connection state,enlarging user capacity,and ensuringlow access delay on the control plane.The connection-like state is quickly established ondemand on the control plane,which is conducive to rapid delivery for massive short and15independent packets on the user plane.It is also helpful for network to fulfill the fast datatransmission with QoS guarantee and further reduce latency on the user plane.The keyenabling technologies for fast connection-like state establishment comprise:(1)New massive user signature design,allocation and acquisition that is developed toremove the extremely under-determined bottleneck of orthogonal signal space and carry outlow-dimensional representation for high-dimensional users.(2)New asynchronous random access that is designed to enable fast and reliable accessagainst the lack of timing information.(3)New random access for low-rate data transmission.Figure 3.Fundamental problem arising from initial accessUser activity detection continues an indispensable technical component for fastconnection-likestateestablishment,whichentailsawholeprocessfromsequencetransmission to reception.Analogous to random access,it help a network to identify therandom subset of active devices,solving the fundamental problem arising from initial access,As depicted in Figure 3,once the network knowns which device has actual demand of datadelivery at the beginning of one transmission cycle,a prompt response can be carried out.TheBS assigns the active user with the corresponding control channel and pilot sequence(forchannel estimation)such that necessary preparation and coordination between them can beimmediately made for a successful communication.The existing PRACH scheme permits anactive user to pick up a preamble randomly from the same sequence set and then send its IDin terms of an explicit message.This process executes successfully just under the case withoutsequence collision.For avoiding the collision,in contrast,the user activity detection schemeallows each device to be preassigned with a unique pilot sequence throughout all time slots.There is one-to-one correspondence between the pilot sequences and the user IDs.At eachtime slot,the BS first detects the active devices by detecting which pilot sequences arepresent.In this way,the activity information can be represented and acquired without sendingthe explicit information of user ID.Due to massive number of potential devices but thelimited size of coherence time-frequency block in the fading environment,the signaturesequences assigned for all users cannot be mutually orthogonal definitely.Non-orthogonalsignature sequences superimpose and cause dramatic multi-user interference,e.g.,when asimple matched filtering or correlation operation is applied at the BS for user activitydetection.For these reasons,the user activity detection becomes much challenging and needsa holistic design of sequence transmission and reception.Remarkably,it is also the sporadicity of traffic pattern itself that provides a promisingopportunity for tackling this challenge.As only a small subset of devices is active at each time16slot,user activity detection amounts to a sparse signal-recovery problem.To create anaccurate,fast and scalable user activity detection,6G should take advantage of the signalcharacteristics caused by sporadic traffic,make joint transmission and reception optimization,and resolve the conflicts between low dimensionality of useable signal space and highdimensionality of user population.Hopefully,the device transmits a minimum-lengthsequence with appropriate precoding strategy while the BS employs fast detection algorithmwith finite steps.At the heart of the holistic design,a nice design in sequence transmission iscrucial not only to construct a high-efficient measurement matrix,but also to promote thedetection efficiency and performance.A desirable user activity detection scheme must meetthe following design criteria:Uniqueness of massive user signatures:To avoid the sequence collision,the sequenceset should be sufficient large to discriminate the massive users over the limited coherenceblock under the fading environment.Meanwhile,constant modulus is desirable foreasing RF transmission for low-cost devices.Minimum-length sequence for perfect identification:To save spectrum resource,therequired sequence length for perfect active-user identification should be as small as theaverage number of active UEs,no matter how the total number of devices scales up.Such a scalability is convenient for network expansion.Moreover,a desirable sequencedesign is expected to remove the search penalty suffered by the existing CS-basedmethods 910 1112.Minimum-entropyandnon-negativesparserepresentationfor useractivityinformation:The activity information is desired to be represented in terms of anonnegative sparse vector,instead of a general sparse vector,which can ease thedetection process and improve the detection efficiency.For the purpose,an appropriateprecoding strategy should be introduced for sequence transmission by exploiting theuplink/downlink channel reciprocity in TDD system.In addition,the entropy of desirablerepresentation is expected to be minimized as small as that of random subset of activeusers.Fast sparsity detection for user activity:The nonnegative representation of activityinformation introduces the sparsity in a natural way.Therefore,it is possible to develop afast detection algorithm without convergence concern that converges within a finitenumber of steps,instead of introducing a norm-based regularization term in the design ofthe sparsity detection algorithm.Robustness against uncertain timing:Uplink timing is uncertain and unknown for BSat the beginning of the initial access,also they are different across the different users.This leads to asynchronous sequence transmission,which may destroy the nice algebraicproperty of the original measurement matrix and degrade the detection efficiency.Adesirable user activity detection scheme should be robust against the uncertain timing.Fast transmission of short packet data with zero signaling:In IIoT application,aconsiderable number of short packets consist just of few bits,which is far less than thesignaling overhead required to establish a connection state.Therefore,6G should further17consider the uncoordinated access for small data transmission.2.3 Slotted ALOHAEnhanced Technologies5Gs PRACH is a random access scheme based on slotted ALOHA.Due to the lack ofchannel sensing,the slotted ALOHA protocol has a high probability of packet collision.Therefore,due to a large number of retransmissions,a very high delay will be caused,whichis impractical to use in an IoT environment under high load.In order to solve the problem ofthe high collision probability of the slotted ALOHA protocol under high channel load,theenhanced technology of slotted ALOHA has appeared,in which Contention ResolutionDiversity Slotted ALOHA(CRDSA)13、Irregular Repetition Slotted ALOHA(IRSA)14and Coded SlottedALOHA(CSA)have attracted much attention 15。On the basis of the slotted ALOHA,in order to further improve the throughputperformance,the preliminary DSA scheme is that each user selects multiple data packets tosend in different time slots.As long as one of the data packets does not collide,the userstransmission is successful.The throughput performance of the DSA protocol is unsatisfactoryunder high channel load.An improved protocol,the CRDSA scheme,adds the locationinformation of another copy to each data packet copy.When a collision occurs,the time slotinformation will be stored.When another data packet sent is successfully received,you canuse interference cancellation technology to eliminate its impact in the time slot.The IRSAscheme is a further improvement of the CRDSA scheme.It expands the situation that eachuser in CRDSA can only send two data packet copies to the situation that variable number ofdata packet copies sent by each user and the specific number is determined by a group ofgiven probability distribution.Unlike the CRDSA and IRSA schemes that simply repeat datapackets,the CSA scheme encodes before sending the data packets,so the energy efficiency ofCSAis better than that of CRDSAand IRSA.The advantage of the IRSA protocol is that the number of data packet copies can beincreased to use more spare time slots and increase throughput in low channel load and can bedecreased to reduce the collision probability as well as ensure sufficient channel utilization inhigh channel load.18（a）（b）Figure 4.Packet loss rate and throughput of three slottedALOHAenhancementschemesFigure 4(a)shows the relationship between the systems packet loss rate and the channelload under the three schemes of IRSA,CRDSA and slotted ALOHA.When the channel loadis lower than 0.938,the packet loss rate is less than 10-6.This is because IRSA duplicates thedata packet copy and uses interference cancellation technology at the receiver to eliminatecollisions,thereby reducing the packet loss rate compared with the slotted ALOHA protocol.When the channel load is higher than the threshold,the packet loss rate of the IRSA protocolrises rapidly.Also because it duplicates the data packet,the number of data packets in thechannel increases exponentially under high channel load.Almost every time slot has morethan one data packet.Therefore,the decoding performance is very poor,iterative interferencecancellation is almost impossible and the packet loss rate will be very high.According to the simulation value in Figure 4(b),after 200 iterations,the maximumthroughput of IRSA can be close to 0.8,while CRDSA does not exceed 0.55.For IRSA,whenthe channel load is lower than 0.8,the throughput increases almost linearly,and basically alldata packets sent can be recovered at the receiver.When the channel load is close to 1,thethroughput performance of slotted ALOHA is better than that of IRSA and CRDSA.This isbecause when the channel load is close to 1,the number of copies of data packets sent byIRSA is much larger than the number of packets sent by the slotted ALOHA.The iterativeinterference cancellation process cannot eliminate most of the data packet conflicts,so it willcause the packet loss rate to increase and throughput to decrease.As shown in Figure 5,in the IRSA scheme,the number of transmissions of data packetcopies(K)determines the performance of the scheme.19(a)(b)Figure 5.Packet loss rate and throughput of K-IRSAschemeAs shown in Figure 5(a),when the value of K is different,the four kinds of K-IRSAsystem packet loss rate simulation values vary with the channel load curve.In the absence ofmulti-packet receiving capability(when K=1),the lowest packet loss rate of K-IRSA isbetween 10-4 and 10-3.As the value of K increases,the packet loss rate of the K-IRSAsystem decreases,indicating that the multi-packet receiving capability is beneficial to reducethe systems packet loss rate and improve the systems normalized throughput performance.As shown in Figure 5(b),the throughput varies with the channel load when the packeterror rate PER is different.When the channel load is lower than the channel threshold,because the channel load is low,the throughput performance of the actual analysis value isbetter than that of the packet error rate.When the channel load is higher than the channelthreshold,the throughput performance of the actual analysis value is worse than that of thepacket error rate.This is because under a high channel load,the reduction of time slotcollisions caused by the loss of a data packet copy dominates,thereby improving throughputperformance.In addition,when the channel load reaches the threshold,the performance of allcurves begins to decrease.This is because as the normalized channel load increases,moredata packet collisions will be caused,resulting in a decrease in normalized throughput.In order to further enhance the performance of IRSA,we can introduce NOMAtechnology such as the NOMA technology in the power level domain between users to formthe IRSA-NOMAtechnology,the performance of which is shown in Figure 6.Figure 6(a)shows the change of the packet loss rate of the IRSA-NOMA system duringtheoretical analysis and actual simulation when the number of selectable power levels Lchanges.With the increase of L,the performance of IRSA-NOMA packet loss rate is alsoimproving.When the number of time slots approaches infinity and the number of iterativedecoding is large enough,the simulation value and the theoretical value will approachinfinitely.Figure 6(b)shows the change in the throughput of the IRSA-NOMA scheme duringtheoretical analysis and actual simulation when the number of selectable power levels Lchanges.When L is low,the throughput performance difference between IRSA-NOMA and20IRSA-C(competition)is small while when L is large,the throughput performance differencebetween IRSA-NOMA and IRSA-C is small under low channel load and the throughput ofIRSA-NOMAis significantly greater than that of IRSA-C under high channel load.(a)(b)Figure 6.Packet loss rate and throughput of the combination of IRSAand NOMA2.4 Brief SummaryFor 6G,the initial access might confront a radical redesign to accommodatedelay-sensitive massive access,where the user capacity and access delay on the control planeshould be simultaneously improved while ensuring sustainable usage of radio spectrum.Theevolution of random access will center on the fundamental problem of active-useridentification by considering and exploiting the distributed and random nature arising frommobile terminals and their specific applications.213.Evolved Multiple Access Transmission TechnologiesIn wireless communication system,multiple access transmission is an important way toimprove system spectrum efficiency.When multiple users access shared resources,each userneeds to be given a specific transmission characteristics to distinguish users,which is calledsignature.Traditional design philosophy is to make a compromise between implementationcomplexity and transmission efficiency.More complex transceiver,such as interferencecancellation can be used to improve transmission efficiency.With the advancement of IC,wireless communication gradually adopts technologies with high receiver complexity andhigh resource utilization.Non-orthogonal multiple access(NOMA)technology can be used for uplink grant-freetransmission.In RRC_CONNECTED state or RRC_INACTIVE state,the terminal can evenomit the scheduling request process and scheduling grant process.When adopting NOMAtechnology,the terminal simplifies the transmission process and therefore saves signalingoverhead,reduces power consumption and time delay and increases system capacity.This section will analyze the 3GPP NOMA technology discussed at the 5G stage,andpropose five promising NOMA technologies as candidate solutions for the evolution ofmultiple access transmission technology in 6G.3.1 The Problems and Evolution of 5G NOMA3GPP discussed and evaluated NOMA 21.We think that 5G didnt adopt NOMAtechnology based on the following three main reasons:(1)There is no mMTC project plan for 5G.NB-IoT technology can meet the connectiondata requirements in the mMTC,so it is unnecessary to use NOMAto support more users.In the 3GPP LTE R13 protocol,the NB-IoT core standard was completed.The R15protocol supports the coexistence deployment plan of NR and NB-IoT and the R16 protocolsupports NB-IoT access to the new 5G core network.In 2015,the International Telecommunication Union(ITU)officially released the 5Grequirements for 2020 including support for the connection of 1 million devices per squarekilometer in the mMTC.In 2017,3GPP decided to use NB-IoT as a 5G technology to meetthe technical requirements of the mMTC and regard it as an official candidate technology forIMT-2020.According to the evaluation report,in terms of connection density,using low-costantenna deployment in a large station spacing of 1700 meters,NB-IoT can reach theconnection capacity of 2.3 million devices per square kilometer.If the station spacing isfurther reduced to 500 meters,the connection capacity will be further improved.NB-IoT canmeet the requirements of IMT-2020 in the IoT scenario.From the operators perspective,in order to ensure the investment value of NB-IoT,it isnot hoped that 5G will standardize another set of solutions for the same scenario.At the22RAN#79 meeting in March 2018,it was determined not to study or standardize NR-basedsolutions for LPWA scenarios.The evolution of LPWA is still going on through the evolutionof NB-IoT.Therefore,mMTC in NOMA SI will no longer target the LPWA scene,and thecoverage enhancement is not considered.The NOMA technology was researched and discussed during the 3GPP NR R14 SI.TheSI concluded that at least eMBB supports uplink orthogonal multiple access.At least formMTC,in addition to the orthogonal scheme,the goal is to support uplink non-orthogonalmultiple access.However,NOMA is not included in the project content of 3GPP NR R15 WIdue to time constraint.In March 2017,the RAN plenary meeting passed the 3GPP NR R15 NOMA SI andcontinued to study the uplink non-orthogonal multiple access technology,but in order to giveother NR WI more time,NOMA SI did not start the discussion in RAN1 until February 2018.The project was also postponed to 3GPP NR R16,and NOMA SI was closed in December2018,and no consensus was reached.(2)Contributors had different opinions on the performance gains of NOMA.3GPP NRR16 NOMA SI conducted simulation evaluations on 13 NOMA schemes.As for the gainsbrought by the NOMA,the contributors cannot reach an agreement.In the end,thestandardization of the NOMAtechnology had not been started.When each UE had same time-frequency resource configuration,same receiver,andactual channel estimation,the NOMA scheme had little difference in terms ofresource utilization compared with the baseline LCRS scheme.Specifically,servingthe same number of users(100 per cell)on the same number of resources(6PRB),and assuming MMSE hard IC receivers and actual channel estimation,the systemsimulation results showed that the PAR of the LCRS scheme was slightly higherthan the PAR of the NOMAscheme PDR=1%.When assuming ideal interference covariance matrix in addition to the aboveconditions,the NOMA scheme had a 100%improvement in resource utilizationcompared with the baseline LCRS scheme.The main reason was the assumption ofideal interference covariance matrix.（3）The discussion of the NOMA scheme did not converge.There were 17 kinds ofNOMA schemes.3GPP had not given a screening method and there was no way to select orintegrate various schemes.The 17 solutions submitted to the 3GPP,as shown in Table 1,involved spreading,scrambling,interleaving,coding,and modulation.Among them,CATTs PDMA,ZTEsMUSA,and Huaweis SCMA all belonged to the mainstream symbol-level spread spectrumscheme.In the end,the standardization of the NOMA transmitter was not agreed because thecompanies could not agree on the gains brought by the NOMAtechnology.23Table 1.3GPP 5G NOMAsolution summaryNOMAFull NameContributorPDMAPattern division multiple accessCATTMUSAMulti-user shared accessZTESCMASparse code multiple accessHuaweiIGMAInterleave Grid MultipleAccessSamsungRSMAResource spread multiple accessQualcommNOCANon-orthogonal coded accessNokiaIDMAInterleave Division MultipleAccessNokia,InterdigitalLCRSLow code rate spreadingIntelFDS/SSMAFrequency domain spreading/short sequence based spreadingIntelWSMAWelch bound equality based spreading MAEricssonNCMANon-orthogonal coded multiple accessLGENOMANon-orthogonal multiple accessNTT DCMLSSALow code rate and signature based shared accessETRILDS-SVELow density spreading with signature vector extensionFujitsuRDMARepetition division multiple accessMediaTekGOCAGroup Orthogonal Coded AccessMediaTekACMAAsynchronous Coded MultipleAccessHughesThe problems identified in 5G NOMA need to be solved in 6G research work.Therefore,the evolution of 5G NOMAtechnology will mainly consider:(1)Identify the application scenarios that can reflect the gain of NOMA.(2)The evolution of NOMAtechnology in 6G scenarios.According to the requirements of 6G and the characteristics of NOMA,the applicationscenarios that can reflect the benefits of NOMA can be identified from multiple aspects,specifically,including:：Large number of terminal scenarios.In the conclusion of 3GPP,if the inter-cellinterference covariance matrix was an ideal estimated value,NOMA technologywould have obvious gains.For application scenarios with a large number ofterminals in a cell,by using statistical characteristics,the inter-cell interferencecovariance matrix can be estimated more accurately.Therefore,the scenario of alarge number of terminals can show the gain of NOMA;Scenario of single antenna configuration.NOMA is more suitable for use caseswhere the network and/or terminal configure a single or a small number of antennas.At this case,the gain of space diversity is small,and NOMA can become the mainsource of gain.In order to meet the needs of 6G,including supporting the transmission of a largenumber of terminals,improving transmission efficiency,reducing implementation complexity24and so on,it is necessary to evolve the NOMAtechnology,which mainly includes pilot designand NOMAsequence design.Pilot design,including activation detection(for terminal identification)pilotsequence design and DMRS pilot sequence design.In addition to the traditionalmethod of increasing orthogonal pilot resources,the pilot design should considerthe following three requirements:(a)reduce the demand for the number of pilotsequences,(b)reduce the probability of pilot sequence collisions,(c)reduce thecomplexity of pilot detection.Design more and longer NOMA sequences.In order to support the simultaneoustransmission of a large number of terminals,more NOMA sequences need to begenerated,which in turn makes the length of NOMA sequences longer.To producesmore and longer NOMA sequences,there are two questions:(a)define the criteriafor NOMAsequence search,(b)reduce the complexity of NOMAsequence search.3.2 Millimeter Wave MIMO-NOMATransmission TechnologyAs mentioned in the previous section,with the increase of antenna scale,the spatialdiversity gain of MIMO technology is used to the extreme.At this time,the introduction ofNOMA may have a significant performance gain.This section introduces a NOMAtransmission scheme design based on millimeter wave massive MIMO system 22.Due to the large bandwidth of the millimeter-wave frequency band,the introduction ofultra-large-scale antennas and hybrid beamforming,millimeter-wave MIMO technology canprovide high-speed broadband services and is one of the important technologies for current5G system and future 6G system.In order to further improve the spectrum efficiency and theperformance of the number of connections,it can be considered to further introduce NOMAtechnology in the millimeter wave MIMO system.However,when the existing power domainNOMA is directly combined with millimeter wave MIMO,as shown in Figure 7(a),the powerdomain NOMA service is used for paired users in(or near)a beam,that is,multiple NOMAusers are served by the same analog beam.There may be situations where the beam directioncannot be aligned with a paired user.On the one hand,it may reduce the probability of theuser pairing successfully,and on the other hand,beam misalignment will also reduce the datarate performance and user fairness.Therefore,this section proposes a millimeter wave MIMO-NOMA scheme based onbeam aggregation,which can reduce the requirements for beam alignment and improve userfairness.Firstly,it is proposed to aggregate multiple adjacent analog beams to generate awider equivalent virtual beam,as shown in Figure 7(b).NOMA user pairing in theaggregated equivalent virtual wide beam can improve the probability of successful userpairing.Then,a non-orthogonal multi-user precoding method is proposed to ensure userfairness by maximizing the minimum achievable data rate of paired users in the aggregatebeam.Considering the complexity of the optimization problem,the non-orthogonal multi-userprecoding can be optimized offline with the help of AI tools,as shown in Figure 8,and the25model trained offline can be used in the online transmission system.Figure 7.Comparison of different NOMAschemes in mmWave MIMO scenariosFigure 8.Offline training process of non-orthogonal precoding matrix for mmWaveMIMO-NOMAschemeFigure 9.SIC reception of the NOMAscheme based on beam aggregationFigure 9 shows the receiver design of the millimeter wave MIMO-NOMA scheme basedon beam aggregation.Since user 3 will be served by two adjacent beams at the same time,itwill be interfered by both user 1 and user 2 at the same time.However,due to the lowsignal-to-noise ratio of user 3,higher power will usually be allocated,so user 3 can directlydetect its own data without interference deletion.However,user 1 and user 2 will beinterfered by user 3,it is necessary to detect the data of user 3 first,and after interference isdeleted,the detection of their own signals can be performed.Figure 10 shows the numerical simulation results,considering the scenario of 3 users and2 beams,where user 1 and user 2 are located in two beams respectively,and the distance fromthe base station is randomly distributed within 0m,10m,user 3 is an edge user,the distance26from the base station is randomly distributed within 10m,d.It shows that the proposed newscheme can achieve a higher maximum-minimum rate compared with the pure power domainNOMAmethod.Figure 10.Max-min Rate performance results(dis the maximum distance between theedge user and the base station)3.3 Tandem Spreading MultipleAccess(TSMA)TechnologyTandem spreading multiple access(TSMA)is a novel multiple access technique formassive machine type communications(mMTC)23.In mMTC,a base station needs tosupport random access of a large number of various devices.Here,random access can becategorized into grant-based random access and grant-free random access.In the grant-basedrandom access system,the data transmissions are coordinated by control signaling so thatcollisions can be avoided,but the resource overhead for coordinating a large number of userdevices is enormous.In the grant-free random access system,the transmissions are notcoordinated,and the corresponding control signaling overhead can be saved.However,due tothe lack of coordination,this kind of transmission is anonymous so that user identification hasto be considered.Meanwhile,autonomous transmission in grant-free random access willcause the uncertain user data packet collisions.In response to the aforementioned challengesof anonymous transmission and collision uncertainty,TSMA was proposed by recent researchstudies.The basic idea of TSMA is to divide user data packet into segments,then performsegment coding to generate redundant segments,and employ multiple spreading sequences tospread the data symbols of the coded segments.Therefore,collisions only occur in specificdata segments and can be resolved by redundant segments.Thereby,high connectivity andhigh reliability can be achieved in TSMAby sacrificing the user data rate.The TSMA transmitter is shown in Figure 11.The data bits sent by active user device isfirst channel encoded.Then data segmentation and segment encoding are completedaccording to the system parameters.After modulation,tandem spreading is implementedbased on the user specific tandem spreading combination from the tandem spreadingcodebook.Compared with conventional transmitters,TSMA introduces segment coding andtandem spreading.The segment coding divides the user channel coded data bits into segments,27and encodes them in terms of segments to generate redundant segments.Tandem spreadingutilizes multiple spreading sequences to spread the modulation symbols on different codedsegments.In tandem spreading,TSMA proposes a tandem spreading codebook to specify theselection of spreading sequences on different segments.In the meantime,the spreadingsequences selected by the user on different segments constitutes a tandem spreadingcombination that characterizes the identity of the user.The TSMA receiver is shown in Figure 12.After receiving the superimposed signal fromthe active users,the receiver first divides the signal into segments.Then energy detection isimplemented on each segment to detect potential orthogonal spreading sequences.Thetandem spreading codebook in TSMA is designed so that each user has a unique tandemspreading combination.Thus user identification can be realized through detecting theorthogonal spread sequences on different segments.After user identification,the signal ofeach identified user is despread through the corresponding tandem spreading combination.The colliding segments of each identified user can be obtained according to the tandemspreading combination.Then in the segment decoding,the colliding segments can berecovered by the redundant segments.Figure 11.The block diagram of TSMAtransmitter28Figure 12.The block diagram of TSMAreceiverThe tradeoff relationship among TSMA connectivity,reliability and user data rate isshown in Figure 13.With the determinate collision resolution probability,when the segmentcode rate is low,the number of simultaneous transmissions is relatively high.At the sametime,the number of simultaneous transmissions decrease with the segment code rate.Theachievable connectivity at each reliability level has an upper limit at the low segment coderate,which is caused by the user identification.It can be concluded on the relationship amongTSMA connectivity,reliability,and user data rate is:TSMA can improve the connectivity andthe reliability by sacrificing the user data rate.Figure 13.Tradeoff among the connectivity,the reliability,and the user data rate ofTSMAIn the high speed scenario with massive connections,there are two challenges in thedesign of the new multiple access scheme.The first is how to achieve high reliabilitygrant-free random access for a large number of user devices under limited channel resources.The second is how to tackle with the Doppler frequency offset and the fast channel fading toachieve reliable transmissions.To this end,the combination of TSMA and OTFS isconsidered and OTFS-TSMA scheme is proposed.The basic idea of the OTFS-TSMAscheme is the resource mapping scheme,which needs to consider the channel characteristicsin the delay Doppler(DD)domain.Based on ideal transmit and receive waveforms in OTFS,the channel impulse response(CIR)in the DD domain show a two-dimensional cyclic shiftcharacteristic.Therefore,an interleaving scheme is designed so that the two-dimensionalcyclic shift of the DD domain resource elements is converted into a recoverable cyclic shift inuser data.In particular,since the TSMA scheme is based on orthogonal spreading,a discreteFourier transform sequence with cyclic orthogonality can be selected.Based on this,thetwo-dimensional cyclic shift in the DD domain can be transformed into the cyclic shifts ofDoppler resource blocks,data segments,symbols and chips.In addition,the OTFS-TSMAsolution can take advantage of the multipath diversity brought by Doppler resource block shiftto improve the reliability of data reception.293.4 Lattice Partition MultipleAccess(LPMA)TechnologyThe conventional NOMA multiplexes users in the power domain to the transmitter side.The far-end users are allocated more transmission power,while the near-end users areallocated less transmission power.Therefore,the far-end user decodes the desired signal bytreating the other users signal as noise.On the other hand,the near-end user uses thesuccessive interference cancellation(SIC)to extract the desired signal.That is,the near-enduser first decodes the signal of the far-end user,and then subtract the corresponding signalbefore decoding itself.However,in the case of similar user channel conditions,theperformance gain of NOMArelative to orthogonal multiple access(OMA)is reduced.Lattice partition multiple access(LPMA)is proposed to cope with the aforementionedpower domain NOMA deficiencies.It is a downlink non-orthogonal multi-user superimposedtransmission scheme that can ensure the fairness of users in a symmetric broadcast channel24.In this design,the base station allocates different lattice code levels for multiple users.The lattice code levels of different users correspond to a different prime number,and userswith poor channel conditions correspond to smaller prime numbers,while users with betterchannel conditions correspond to larger prime numbers.The data of one user is weighted bythe coefficient from the product of the prime numbers corresponding to all other users.In thisway,the user with poor channel condition can be compensated by a large prime number.LPMA uses the common prime number structure of lattice codes to balance out theinterference of other users.Each user extracts the required signal by performing modulolattice operations and SIC/PIC decoding.When users have similar channel conditions,LPMAcan realize user multiplexing by allocating different homogeneous lattice points in the samefinite field.30Figure 14.The block diagram of LPMAFigure 14 shows the schematic diagram of LPMA.Similar to the power allocationprinciple of NOMA,LPMA guarantees user fairness through structured power allocation.Inother words,the codewords of the users with poor channel conditions are weighted by a largerproduct,and the users with better channel conditions are weighted by a smaller product.Decoding is performed by modulo lattice operation with SIC(MLO-SIC).The decodersubtracts all reconstructed interference and then performs modulo lattice operation.Theperformance of LPMA is shown in Figure 15.Compared with OMA and the power domainNOMA,LPMAincreases the user throughput by 63.7%and 28.3%respectively.Figure 15.User throughput comparison among LPMA,OMAand power domain NOMA3.5 MCFTN-NOMATechnologyThe multi-carrier faster-than-Nyquist(MCFTN)introduces inter symbol interference andinter carrier interference to increase the code rate of single user 24.Essentially theincrement is achieved by utilizing the time and frequency resources non-orthogonallybetween each transmitting modulated symbol,just share the similar idea as the conventionalnon-orthogonal multiple access(NOMA)techniques.Comparing to the conventional NOMAtechniques which utilize heuristic transmitting code book design,the non-orthogonality of thetime-frequency resource in MCFTN based NOMA(MCFTN-NOMA)can be preciselydescribed by the mathematical equations.Therefore,the level of non-orthogonality can beadjusted by a few scalar parameters,which avoids the overhead of signaling the code bookand mapping patterns,which often in matrix form.In MCFTN-NOMA,since the modulating and demodulating of the MCFTN signalrequires the accurate synchronization of symbol streams from different user,it is more proneto be adopted in the downlink.The modulated symbols of different users are partiallysuperimposed in the time frequency resources after the MCFTN signal processing.Each ofthe output time domain samples of the MCFTN modulator contains the information of severaladjacent modulated symbols in time and frequency resource grid.31Figure 16 illustrates the based-band processing of the MFTN-NOMA.In the figure,uijare the modulated symbols belonging to different users and the modulated symbols of eachuser can be subjected to some pre-operations such as power allocation.The modulatedsymbols of different users can be grouped as the input of the interleaving module,whichoutput the interleaved modulated symbol stream for FTN signal generation.The FTNmapping module first up-samples the modulated symbols and then using proper pulse train tocarry those symbols,i.e.,the pulse shaping.The output from the FTN mapping module are first serial-parallel converted and thenmapped to consecutive multiple subcarriers.If the subcarriers of multi-carrier modulation areorthogonal,this can be achieved directly by using the OFDM.Furthermore,to better exploitthe benefit of non-orthogonal subcarrier mapping,i.e.,the bandwidth compression in thefrequency domain,the principle of SEFDM signals generation can be applied 2526.In summary,for an MCFTN signal,partial reuse of time domain resources is achieved byunder-sampling in the time domain,and partial reuse of frequency domain resources isachieved by non-orthogonal subcarrier mapping in the frequency domain,thus achieving thesame effect of partial reuse of time and frequency resources as pursued by NOMA.Therefore,by exploiting the nature of non-orthogonal superposition of waveforms of different users inthe time and frequency domains,the MCFTN-NOMA is a potential candidate of the nextgeneration non-orthogonal multiple access technology.The advantage is that the signalingoverhead is small and allows for highly flexible configurations.At the same time,the receiveralgorithm can follow the mature algorithm of MCFTN technology,avoiding the errortransmission drawback of MMSE-SIC in the traditional NOMAscheme.Figure 16.MCFTN-NOMAbaseband processing3.6 Evolved SCMATechnologyThe basic idea of SCMA technology is that the transmitter sends multiple signals fromone or more users through the code domain spread spectrum and non-orthogonal overlay in32the same time-frequency resource unit,and the receiver separates multiple data layers in thesame time-frequency resource unit through linear despreading and interference elimination.On the basis of the time domain and the frequency domain,it increases the multiplexing ofthe code domain with fully considering the joint design of the time-frequency domainresources,the code domain resources and the power domain resources,and improves thespectral efficiency and the system capacity.The entire process of SCMA can be described as an encoding process from a binarydomain to a high-dimensional complex number domain.The codewords of different users arenon-orthogonal overlay in a spread spectrum mode on the same time-frequency resources.The encoding process of SCMA is first using FEC(Forward Error Correction)for a set of bitsto give the receiver the ability to correct errors.The data is then sent to the SCMAencoder formodulation,the essence of this step is to modulate the bit data into the SCMA codewords.Finally,the sparse mapping is performed,and each layer,as defined by the mapping matrix,issent on the block of resources.Figure 17 is a factor diagram of the map matrix F and the SCMA layer nodes and theresource nodes.The map matrix Vj B42of each layer defines which resource blocks thatthe layer occupies.Combining the map matrix of each layer can generate a matrix F,and afactor diagram of the SCMA layer nodes and the resource nodes can be drawn according to F.And the column of F matrix from left to right respectively represents the mapping matrix ofeach layer.Figure 17.Mapping matrix F and factor figure.In the complex domain constellation design of SCMA,a constellation map with an idealEuclidean distance is first chosen as the benchmark constellation,rotating the benchmarkconstellation to control the dimensional correlation and the difference in power,withoutchanging the original Euclidean distance,and then alternately producing each dimension inthe new N-dimensional constellation map.Compared with the traditional SCMA codebook design,the new regular SCMAcodebook design method based on the Star-QAM constellation map and the irregular SCMAcodebook design based on the rotation angle and external information transfer(EXIT)graphhave evolved the methods.As shown in Figure 18,we consider a four-loop Star-QAM constellation designed with a2-dimensional SQ-SCMA codebook of size M=4,where the radius is R1,R2,R1 and R2,respectively,and R1=R1,R2=R2,R2=R1,R1=12(2 2 22 1),where is a33predetermined constant determined by the mean energy,(1, )is the design parameter,and assume that each user codebook energy is Es=1.The designed master codebook MC,AMCis shown in the figure below.Then the numerical search algorithm is applied accordingto the codebook design criteria to obtain the optimal and values.Figure 18.The four-loop Star-QAM constellation with the four radii are R1,R2,R1,R2.The future wireless communication network is a highly integrated heterogeneousnetwork,which will be composed of ultra-high density arranged macro cell base stations,small zone base stations,as well as multiple types of access points(WiFi hotspots,home basestations,etc.).The way to achieve the rapid growth of network capacity in a highly integratedwireless environment is becoming an important problem.Network intensification has beenconfirmed to be one of the most significant means to improve the network capacity,and theultra-dense wireless fusion network will be one of the important scenarios of the futurewireless network,where non-orthogonal multiple access technology will be widely used.Therefore,BPM(Bounded Pathloss Model),which can accurately characterize the near-fieldpropagation properties of signals in ultra-dense wireless networks,and further verify theaccuracy of it through experimental measurements in a microwave dark chamber environment,is used in the scene.Then we use stochastic geometry theory and BPM to analyze the cellularuplink network throughput of SCMA and OFDMA technologies,and its asymptotic trendassociated with base station density.Figure 19 shows a progressive trend of network space throughput using SCMA andOFDMA technologies in the case that the negative dimension and the path loss factor is set to=4,NS=20 in the system and the number of available SCMA codebooks is NC=30.Ascan be seen from Figure 19,when the base station density in the uplink cellular networkexceeds a critical density(capacity distortion density),a further increase in the base stationdensity will lead to a rapid decrease in the network capacity.At the same time,the spatialnetwork throughput of SCMA network is always greater than that of OFDMA network,andabout twice as larger as OFDMA.From the above results,it can be seen that SCMAtechnology has great advantages over OFDMA technology in its spatial throughput inultra-dense scenarios,and it will be widely used in future wireless networks.34Figure 19.The progressive trend of network space throughput with the degree ofdensification under OFDMAand SCMA.3.7 Brief SummaryThis section discussed evolved multiple access transmission technology and analyzes thereasons why 3GPP did not adopt NOMA technology in the 5G stage.These technical reasonswill be further analyzed and resolved in 6G.This section also proposed five kinds ofpromising NOMA technologies,including millimeter wave MIMO-NOMA,TSMA,LPMA,MCFTN-NOMAand evolved SCMA.The millimeter wave MIMO-NOMA,which superimposed the NOMA technology in awide beam range,can improve user fairness.The TSMA,which segmented user data andencoded it in units of segments and used multiple spreading sequences to spread spectrum ofsymbols on the different coding section,can achieve a compromise between transmissionefficiency and reliability.The LPMA guaranteed user fairness through structured powerallocation.The users with poor channel conditions had larger factor weighting and the userswith better channel conditions were weighted by a smaller factor,which can improve userthroughput.The MCFTN-NOMA,by time domain under-sampling to achieve partial reuse oftime domain resources,or by non-orthogonal sub-carrier mapping to achieve partial reuse offrequency domain resources,can reach the effect of NOMA.The evolved SCMA technology,by regular SCMA codebook design based on improved codebook design,can improve userthroughput.In 6G research work,it is necessary to conduct more comprehensive and in-depthresearch on the above-mentioned technologies to meet the requirements of the newapplication scenarios in 6G.354.Integration of Random Access and MultipleAccess Transmission4.1 Necessity of Integration of Random Access and MultipleAccess TransmissionThe first two generations of cellular systems are mainly for voice services of low andconstant rate.There,the circuit switch is used and the radio resource scheduling is limited toradio resource control(RRC)level,e.g.,semi-statically configured.Hence,the controloverhead of physical layer is very small.The drawback of semi-static configuration is the lowresource utilization rate and its inability of fast link adaptation.In 3G,4G,and 5G networks,data services become dominant.In the OFDMA based systems started from 4G,linkadaptation and resource scheduling have become more mature,which can flexibly adapt tosmall-scale fading channel,bursty data services,and inter-user interference,etc.,so that thespectrum efficiency and data throughput of the system can be greatly improved.One important prerequisite to achieve high spectral efficiency via dynamic scheduling isthat the packet size should be large enough,e.g.,much larger than that of physical controlsignaling.This prerequisite is applicable to big data services,such as file transfer andhigh-fidelity video/audio services.In addition,because the total capacity of the system isfixed,the amount of data transmitted by a single connection multiplied by the number ofconnections cannot exceed the total capacity of the system.Therefore,for big data services,there are not many connections transmitted simultaneously,which makes dynamic schedulingeasier to implement.The 6G communication system will be a network that can include the communicationdemands of human to human,human to thing,thing to thing,and intelligence andintelligence,and fully empower everything.In the massive connection scenario of 6G,onemain feature of services is that the number of connections is very huge.Since the amount ofdata transmitted by a single connection multiplied by the number of connections cannotexceed the total capacity of the system,the size of single transmission packet in massiveconnection scenario will not be too large.These features require that the connection nodeshave low power consumption and cost,and that auxiliary procedures other than packettransmission cannot be too complicated.For example,to minimize the system load andterminal power consumption,the terminals can stay in the deep sleep state(also called RRCIdle)for most of the time.The terminal can send the message as soon as data packet arrives,and go to deep sleep immediately after the transmission.However,in traditional dynamicscheduling transmission,the terminal needs to enter the connection state first,and then applyfor transmission resources/grant to transmit data.However,the terminal has to performrandom access process from the sleep state to the connection state,which involves multiple36rounds of interaction with the base station:sending preamble,random access response(RAR),sending L2/L3 control information,and sending Message 4.This procedure will undoubtedlysignificantly increase the signaling overhead,power consumption,and delay.In addition,ahuge number of nodes means that collision or congestion will occur with high probability inthe procedures of random access and transmission resource request,as shown in Figure 20,which will inevitably worsen the transmission efficiency.As a result,the number of accessnodes is far from requirement,the signaling overhead is unacceptable,the node cost remainshigh,and the power consumption,especially node power consumption,cannot be reduced byorders of magnitude.Figure 20.Problems of massive nodes from deep sleep state to data transmissionIn 4G LTE,there is a scheduling mechanism called semi-persistent scheduling(SPS).Itsaim is to reduce the physical control overhead for small data packets,especially for periodictraffics such as voice-over-IP(VoIP).During a talk spurt,the data rate of VoIP remainsconstant,e.g.,a voice packet generated every 20 ms.The average duration of a talk spurt isabout 12 seconds,consisting of about 50100 voice packets.The small-scale fading can becompensated by closed-loop power control,so that the SNR at the receiver can be kept moreor less constant.SPS allows MCS to be unchanged over some time,and the resourceallocation can either remain unchanged or hop according to a fixed rule.Therefore,dynamicsignaling is saved.SPS can be considered as an enhanced semi-static configuration,mainlysuitable for periodical and small packets with fixed packet sizes.SPS normally operates in RRC connected state,that is,the terminal has completed theinitial access procedure.Although the scheduling frequency is far lower than the packetarrival rate,in most cases,the transmissions are non-contention based,e.g.,no resourcecollisions between different users.In 5G system,the evolved SPS can be used for URLLCscenario to ensure high reliability and reduce the latency in user plane.At this time,SPS iscalled configured grant.It is very similar to SPS in LTE,with some new features added,forinstance,no need for physical layer control signaling to activate or de-activate SPS and itsconfiguration.Configured Grant can be considered as a special case of grant-free,in the sensethat the dynamic scheduling request for each data transmission can be saved.Essentially,Configured Grant is dynamic-scheduling free.It should be pointed out that SPS-baseddynamic-scheduling free operates in a non-contention manner,because resources of differentusers are essentially pre-configured by the base station,instead of being randomly selected incontention-based manner by each user.Furthermore,for non-contention grant-free,datademodulation related spreading sequence/scrambling sequence/interleaver,and referencesignal are pre-configured by the base station to avoid collision.For example,thepre-configuration of the base station can ensure that the spreading sequence/scrambling37sequence/interleaver,and the reference signals,etc.of users transmitted on the sametime-frequency resource can be different or low-correlated.Although pre-configured grant-free such as SPS or Configured Grant can reduce theoverheadofphysicalcontrolsignaling,ifthetraditionalpre-configuredgrant-freetransmission is used to implement massive connections,the efficiency is still very low.Thereasons are as follows:The pre-configured transmission resources are usually regular and periodical.If a node ispre-configured with periodic resources within a period of time,but the node has no serviceswithin this period of time,then the resources are wasted,and the efficiency is very low.Tosave the resources,long-period resources can be reserved for the node.However,this willincrease the average time that the node needs to wait for transmission,that is,increase thetransmission delay.For a small number of connections,a certain spectrum efficiency and timeefficiency can still be exchanged for simplification of the transmission process withoutdynamic scheduling.But in the massive connection scenario,it will be difficult for the systemto pre-allocate periodical transmission resources for all nodes.The services of a huge number of nodes are diverse,some nodes have more frequentservice packages,and some are sparser.To improve efficiency,it is necessary to pre-configuretransmission resources with different periods for different type of services.In the massiveconnection scenario,a large number of nodes with different requirements make it difficult tomanage pre-configured resources efficiently.If a node requests periodical transmission resources for a period of time in a cell,butcellular handover occurs during this process,the node needs to request a new pre-configuredperiodical transmission resources from the new cell,and informs the original cell to release itspre-configured resources.In addition,for the scenario with huge number of connections andburst small packets,to improve efficiency,the interval for pre-configuring resources isusually long.This means that the negative impact of handover will be increased significantly,reducing system efficiency and increasing system complexity.Even for fixed nodes,if theenvironment around the terminal changes,handover may also occur,especially for the nodeslocated at the edge of the cell.Terminals need to request new resources when handover,which is a complicated process,with time delay and power consumption problems.In themassive connection scenario,the number of non-central nodes is also huge,so the problemscaused by handover will be more serious.In SPS or Configured Grant,if the initial transmission is not successfully decoded,it isnot easy to retransmit the packet.For instance,when the initial transmission of a packet by auser experiences deep fade and its reference signal is not detected,the base station would notknow whether the user has tried to access.Then there is no way for the base station to indicateNACK and the resources for retransmission to the user.The user can assumes that thetransmission failed when the user does not received the ACK.The user has to wait for thenext pre-configured time-frequency resources for retransmission,which undoubtedly increasethe delay.If the user has a new packet for the next pre-configured time-frequency resources,the transmission occasions of retransmitted packets and new packets need to be further38coordinated,which also increase the delay.For the scenarios with unfixed packet sizes,SPS or Configured Grant also has someproblems.As long as the size of a packet exceeds the transport block size of onepre-configured resource,block segmentation would be needed.When a packet is divided intomultiple smaller packets,the latency of the transmission would be increased.4.2 Extremely Simple Grant-free Transmission SchemeAs can be seen from the above analysis,although SPS/Configured Grant can achievegrant-free transmission,it is not applicable to massive connection scenarios in 6G.Tominimize the system load and power consumption of the terminal,it is better for the terminalto stay in the deep sleep state(also called RRC Idle)for most of the time.When data arrives,the terminal can transmit data autonomously without the control signaling of the base station,that is,without the need to establish a connection in advance,and go to deep sleepimmediately after the transmission.In this way,extremely simple transmission can beachieved,and the system spectral efficiency and terminal power consumption can be achievedto the extreme,as shown in Figure 21.Figure 21.Extremely simple grant-free transmission for massive connectionsHowever,the shortcomings of extremely simple grant-free transmission withoutconnection establishment in advance are also very prominent.The base station may receivedata packets from different terminals on the same time-frequency resources.The transmissionof these packets is contention-based,so resource collision exists.Especially for the massiveconnection scenario,the user load is very high,and the collision may be very serious,therefore,it is a great challenge for base station to separate these packets.In the massive connection scenario,if contention-based grant-free transmission uses onlyone-dimensional power domain,that is,the near-far effect,it is very difficult to implementefficient transmission.Further,code domain and space domain need to be used to supportmulti-user transmission.In code-domain contention-based grant-free transmission,the transmitters randomlyselect spreading codes,and the receiver decodes all users.The property of the spreadingsequence will directly affect the performance and receiver complexity of the code-domaincontention-based NOMA scheme,which is the key to the code domain spreading scheme.If along pseudo random sequence(PN sequence)like the traditional DS-CDMA(for example,IS-95 standard)is used,it is easy to guarantee the low correlation between sequences.Inaddition,it can provide a soft capacity for the system,that is,the number of users(that is,the39number of sequences)that are allowed to access the system at the same time is greater thanthe sequence length.At this time,the system is in an overloaded state.Although the long PNsequence can provide a certain soft capacity,that is,a certain overload rate,SIC process willbe very complicated and lengthy if a long PN sequence is used in the scenario with highoverload rate,e.g.,mMTC scenario.Furthermore,the information is spread over too manytime-frequency resources in the transmitter side,which also increases the complexity of theterminal.If a short sequence can reach the high overload rate close to that of the longsequence by optimizing the design,it is more appropriate to use such a short sequence inconsideration of the complexity of receiving/transmitting.However,the supported useroverload rate drops quickly if the length of traditional PN sequence is shortened.The reasonis that the traditional PN sequence is a binary real sequence,the low correlation of therandomly generated sequence set cannot be guaranteed after the sequence is shortened.Existing NOMA technical solutions such as MUSA,PDMA,SCMA,etc.,improve theperformance of code-domain contention-based NOMA by optimizing the symbol-levelspreading sequence.Among them,MUSA uses the M-ary(M=2)complex code sequence asthe spreading sequence 13.The correlation of this type of sequences with short length(e.g.,the length is 4 or 8)is also relatively low.For example,the real and imaginary part ofcomplex number in one type of MUSA complex spreading sequence is from the simple binaryset-1,1 or 3-ary set-1,0,1,and this type of MUSA sequence can also achieve excellentperformance.The corresponding positions of the elements in the two sets in the constellationdiagram are shown in Figure 22.Due to the excellent MUSA complex M-ary codes andhigh-efficiency SIC receiver,MUSA can support a considerable number of concurrent accessusers on the same time-frequency resources.It is worth mentioning that these access users canselect spreading sequences randomly,and then spread their modulation symbols to the sametime-frequency resources.MUSA allows a large number of shared access users to transmit asthey want,or go to deep sleep if they dont transmit.It does not require each access user to gothroughcomplicatedcontrolproceduressuchasresourcerequest,scheduling,andconfirmation before they can transmit.Figure 22.(a)Elements of binary complex spreading code(b)Elements of 3-arycomplex spreading codeIn the traditional MU-MIMO system,the base station can estimate the space domainchannels of different users through the orthogonal reference signals of the users,and then40formdifferentreceivingbeamstoreceiveandseparatetheusers.However,incontention-based grant-free transmission,users autonomously select their reference signals,therefore,different users may select the same reference signal,called reference signalcollision.The probability of reference signal collision is very high in high-loading scenario.Once multiple users reference signals collide,the base station cannot separate these usersthrough reference signals.To alleviate the pilot collision and contamination,and estimate thewireless and time/frequency offsets accurately,the number of reference signals needs to beincreased several times,that is,the length of reference signal needs to be increased severaltimes.As a result,the pilot overhead is increased several times and the detection complexityincreases squarely.In recent works,Data-only scheme 17 18 and independent multi-pilotscheme 19 were proposed to fully exploit the spatial capability and alleviate or eveneliminate the pilot collision.Data-only scheme,not relying on pilot,can achieve efficientblind detection by utilizing the statistical and geometric features of the data symbols.Inindependent multi-pilot scheme,user transmits multiple independent pilots once time withoutincreasing the pilot overhead.Under the same pilot overhead,the probability of pilot collisionin independent multi-pilot scheme will be much lower than that of traditional single pilotscheme.A simple receiver based on iterative multi-user detection and interferencecancellation can be adopted at the base station.In each detection round,uncollided users canbe correctly detected,and the reconstructed pilot and data symbols of correctly detected usersare removed from the received superimposed signals,and then next round of multi-userdetection is performed until all the users are detected.By fully exploiting the spatial degree offreedom and alleviate or even eliminate the pilot collision,Data-only scheme and independentmulti-pilot scheme can be used in high-loading scenario.In summary,by efficiently solving the problem of resource collision and fully exploitingthe multi-user multiplexing capability in spatial/code/power domain,competition-basedgrant-free transmission scheme combined with random access and non-orthogonal multipleaccess can support considerable number of access users on the same time-frequency resource.It should be noted that these considerable number of access users can transmit as they want,or go to deep sleep if they dont transmit.It does not require each access user to go throughcomplicated control procedures such as resource request,scheduling,and confirmation beforethey can transmit.4.3 Decentralized Grant-free Transmission SchemeContention-based grant-free transmission can be applied not only to cellular-based IoTscenarios,but also to wider scenarios such as V2V communication.The two features ofhigh-density V2X are massive and burst information transmissions with the requirements oflow-latency and high-reliability.It is a big challenge to meet the low-latency andhigh-reliability requirements in massive and burst information transmission scenarios.Furthermore,as the vehicle nodes move fast,the network topology of the V2X changesrapidly,making it more difficult to meet the requirements of massive and burst transmissions,low delay,and high reliability at the same time.41In the case of high-density vehicle distribution,the main reason for the performancedeterioration of eV2X is the conflict caused by contention-based resource selection during thelarge-scale vehicle information exchange.Both LTE-V and IEEE802.11p/IEEEl609 areconfronted with this problem.Without the need to establish a connection in advance,decentralized grant-free transmission,combined with high-efficiency full-duplex technology,is expected to solve this problem.Specifically,in the scenario where a large number of vehicles exchange informationfrequently,multiple vehicles may broadcast their own messages at the same time.Since thereis no cooperation between vehicles,the transmission signal generation and the selection oftransmission resource are determined by each vehicle autonomously.It will appear thatdifferent vehicles choose the same transmission resource and the same transmission signalgeneration method,that is,collisions occur,which makes it difficult for other vehicles toreceive these broadcast messages.Decentralized contention-based grant-free high-loadingtransmission technology makes it easier to demodulate the superimposed multi-user signals inV2V scenario.The traditional V2V scheme,whether it is LTE-V or IEEE802.11p/l609,is essentiallybased on a half-duplex transmission mechanism,that is,if one vehicle transmits data on acertain channel,it will not receive data on this channel,which will lead to missing reception.That is,if more than one vehicle are broadcasting their own data packets on a channel,thesevehicles will not receive the other partys data packets,which is very unfavorable forreliability.Although the existing methods can reduce the missing reception throughimproved sensing mechanisms combined with larger time-frequency resources,it is stilldifficult to ensure that there is no miss reception.The full duplex mechanism cancompletely avoid the problem of missing reception,that is,only one channel is required,andeach vehicle transmits and receives signals on this channel simultaneously.In this way,muchless spectrum is needed to solve missing reception problem.However,the traditionalfull-duplex communication will face very strong self-interference.It is still difficult toeliminate the interference even by using complicated self-interference elimination methods,which will seriously reduce the reliability of V2V transmission.Reference 20 proposes ahighly efficient V2V solution that makes full use of two features:(1)The vehicle size is muchlarger than that of a common communication terminal.(2)The information of the surroundingvehicles is more important.In this solution,the transmitting and receiving antennas are placedseparately as much as possible,as shown in Figure 23,so that even if the self-interferenceelimination is not used at all,the full-duplex self-interference will not be much larger than thetarget signal.Furthermore,high reliability full-duplex grant-free transmission is achieved byfully utilizing the advanced inference cancellation in space domain/code domain/powerdomain,and by treating full-duplex self-interference as target signal.No self-interferencecancellation module is needed,which further simplifies full-duplex V2V communication.42Figure 23.Decentralized grant-free transmission schemeBy combining the above-mentioned high-efficiency grant-free high-loading transmissiontechnology and high-efficiency full-duplex technology,it is possible to allow a large numberof vehicle terminals to directly exchange information without having to listen before sending,andthe problems of missed reception and hidden nodes of traditional technology can beavoided.Finally,the ultra-low delay and ultra-high reliability V2V direct communication isachieved.The simulation results in Figure 24 show that in the high-density vehicle networkscenario,this method can achieve 1/5 to 1/10 transmission delay of the traditional method,and improve the reliability by 1 to 3 orders of magnitude.Figure 24.Performance of decentralized grant-free transmission scheme4.4 Concatenated Code Scheme for Massive RandomAccessAt present,concatenated code scheme is a low-complexity per-slot implementation 27of massive access,which has good potential for industrial application.On the premise ofrealizing the conceptual goal of massive random access,the system energy efficiencyincreases linearly with the number of active users.In this scheme,multiple access codes andchannel codes are concatenated,and the superposition principle of linear codes is used torealize the two objectives:collision resolving and reliable information transmission againstchannel noise.In addition,hard and soft decision can be adopted at the receiver to ensure the43equivalence between binary domain superposition and the real GMAC(Gaussian MultipleAccess Channel).Figure 25 is the system block diagram of the concatenated code scheme:Figure 25.System structure of concatenated code schemeIn the concatenated code system,each user has the same codebook and the code word ismapped randomly and uniformly to a certain time slot.The transmitting and receiving endsonly encode and decode slot-by-slot respectively,to deal with at most T users superimposedon the same slot,i.e.the T-fold ALOHA problem.The cascade codes deal with collision andnoise problems separately in GMAC channels.First,the BCH multiple access code is used tosolve the multi-user superposition problem(under tolerance threshold T).Then an FEC(Forward Error Correction)is cascaded to counter noise.BCH multiple access code maps theinformation bits of each user to one of the columns in the parity-check matrix(as a syndromecomponent),so when the encoded packets of each user are stacked in the channel,the receivercan treat the received signal as a syndrome encoded by the FEC according to its linear nature.The decoder at the receiving end is mirrored to the encoder at the transmitting end.FECdecoding is performed first,and the result is the syndrome of BCH multiple access code,thatis,the modulo-2 plus result of the syndrome component mapped by each user.The decodingalgorithm of BCH multiple access code is based on remapping the syndrome back to thecolumn space of BCH check matrix.Then the information payload of each user is determinedaccording to the column index.This process is the same as the decoding process of BCH codeas FEC,which can be accomplished by the classic Berlekamp-Massey algorithm 28.Afterdecoding,the data packets superimposed on each time slot are recovered as the informationpayloads of each user.Due to the unsourced nature of massive random access,the receiverdoes not care about the belonging of each packet,but only the correctness of the decoding ofthe packets.Therefore,the performance index of the concatenated code system is exactly thePUPE proposed in Section 1.2.3,that is,only the overall average packet loss rate underspecific system configurations is counted.According to the above coding structure,the characteristics of the cascaded code schemecan be summarized as follows:The overload factor is unknown and variable.Due to theuncoordinated feature of massive random access,the superposition behavior of users in thetime slot is random,so the overload factor varies randomly from 0 to T.The concatenatedcode scheme uses BCH multiple access code to realize blind detection of superimposed userinformation without pre-allocated codebook.The overload factor of the classical NOMAscheme is fixed and relatively small,generally at the level of 1.53.While in a typical44massive access scenario(50 users,30 time slots,PUPE0.05),the overload factor can be upto 10.Therefore,the concatenated code system conforms to the transmission characteristics ofmassive access and random collision.As a typical realization of the T-Fold ALOHAparadigm,the concatenated code adopt a slot-by-slot encoding and decoding structure.Afterthe time slot division and coding parameters configuration,according to the active user scaleand the number of orthogonal resources,the encoding gain of concatenated code is utilized totackle with the user superposition interference and channel noise in GMAC.The system onlyneeds to deal with the encoding problem of at most T users superimposed on one time slot.Moreover,the purpose of the joint optimization of the above coding parameters is to find acompromise balance between the single-slot coding gain introduced by the code length rateadjustment and the overload factor reduction caused by the time slot number adjustment.Decoding can be carried out slot-by-slot,so the processing delay is low,to meet the demandsof delay-sensitive services.(3)Low implementation complexity.BCH multiple access codeexploit the algebraic structure of BCH code,and the complexity of the coding mapping anddecoding process is greatly reduced compared with the traditional NOMA scheme based onthe iterative structure on the factor graph.The cascaded short FEC can directly apply lowcomplexity and high performance linear codes such as Polar and LDPC.Its low complexity isalso in accordance with th