Yifei Yuan and Xiaowu Zhao

（ZTE Corporation，Shenzhen 518057，China）

5G:Vision，Scenarios and Enabling Technologies

Yifei Yuan and Xiaowu Zhao

（ZTE Corporation，Shenzhen 518057，China）

This paper presents the authors' vision for 5G wireless systems，which are expected to be standardized around 2020（IMT-2020）. In the future，ubiquitous service will be the key requirement from an end-user's prospective，and 5G networks will need to support a vast mesh of human-to-human，human-to-machine，and machine-to-machine connections.Moreover，5G will need to support these connections in an energy-efficient manner.Various 5G enabling technologies have been extensively discussed.These technologies aim to increase radio link efficiency，expand operating bandwidths，and increase cell density.With these technologies，5G systems can accommodate a massive volume of traffic and a massive number of connections，which is fundamental to providing ubiquitous services.Another aspect of 5G technology is the transition to an intelligent cloud that coordinates network access and enables flatter architecture.

5G；IMT-2020；ultra-dense networks；massive MIMO；service ubiquity

1 Introduction

Cellular communications have gone through four decades of development，from 1G analog systems to 2G GSM and IS-95；3G CDMA2000，UMTS and HSPA；and finally，4G LTE.Worldwide，penetration of mobile phones is now more than 60%，even when counting under-developed countries where basic living conditions are still not guaranteed.The deployment of 3G and 4G has facilitated the proliferation of smart devices，which enable much easier access to electronic information and encourage interaction with remote computing systems，regardless of whether the user is stationary or on the move.This trend will continue with future 5G systems［1］.Human beings will rely more and more on cellular networks to acquire，disseminate，exchange，and manage information.5G cellular services will be oriented towards user experience and satisfaction and will be supported by high-performance systems with capacity three orders of magnitude greater than 4G.This rich，vivid content will be instantly available anytime and anywhere.

User experience and satisfaction is the driving force of 5G. Wireless researchers and operators will come up with more innovative ways of converging devices，networks and services. As well as individuals，businesses，organizations and governments will also benefit greatly from 5G networks，which will be versatile，intelligent，and able to support a myriad of applications.5G networks combine the advantages of cellular systems and wireless LANs.These two families of wireless technology have evolved along quite different paths since 3G，each being used for particular scenarios and unable to replace the other. With converged technologies，5G networks will be comprehensive and able to penetrate more aspects of human life.

The so-called“mobile ICT era”implies ubiquitous mobility. Innovations in 5G will significantly increase efficiency in fields as diverse as education，healthcare，manufacturing，government，transportation，and finance.The boundary between the physical and digital worlds will be further blurred with 5G.

The rest of this paper is organized as follows.In section 2，we discuss 5G in terms of service ubiquity，vast interconnectedness，and energy efficiency.In section 3，we discuss some technical issues related to 5G，emphasizing the massive volume of data traffic and transition to intelligent cloud.In section 4，we draw some conclusions.

2 5G Vision

The story of high-end smart phones reveals that the need to provide better user experience is the main impetus for increasing network capacity.In 2007，there were no“killer applications”in the US.As a result，3G cellular systems in the US were loaded to less than 20%of capacity，and operators began to question why they had outlaid such huge amounts of capital for 3G.However，in June 2007，Apple debuted its high-end phones，and this completely changed the situation［2］.The resulting jump in demand for wireless data throttled networks and forced operators to deploy more 3G equipment.Operatorswere also pressed to accelerate 4G standardization.Increased system capacity fulfills the needs of users；the resultant appetite for fast data pushes the development of technology；and this further increases capacity.This cycle will continue with 5G，which will support a wide range of services and applications that will be vastly more interconnected and have greater effect on business and social life.

2.1 Service Ubiquity

The previous four generations of wireless systems were all designed to improve the peak rate and average throughput of a cell.The peak rate is the maximum data rate that can be achieved for a single user in the best propagation environment. Key performance indicators（KPIs）related to a technology give a good indication of the full potential of the technology but are often only significant for marketing.In reality，the user rarely fully experiences what the technology is touted to provide，even if the user is close to base stations in a lightly loaded network.In addition，the peak rate is usually only obtainable using the highest category of mobile terminal for the particular technology.Terminals that are not top-of-the-range will not be capable of delivering data at the peak rate advertised.Average cell throughput more accurately reflects what a user can typically expect from their services（Fig.1）.However，this is also not guaranteed for the majority of users.In many cases，a user at the cell edge experiences low data rate and a high rate of call dropping.This is a cause of customer complaints or even lawsuits.

5G can significantly increase network capacity，peak data rate，number of connections，and traffic density within an area. It can also significantly reduce latency and provide highly accurate indoor positioning.Service ubiquity is a high priority for a system designed to be user-centric and tailored to different applications（Table 1）.Ubiquity can be better measured by taking into account the resource usage patterns and traffic characteristics of wireless services to be provided.5G will provide diverse services in areas such as office，social networking，and e-commerce，and online financial services.Peak data rate and average cell throughput are certainly not sufficient indicators of service ubiquity—KPIs such as data rate of worst 5%of users are more revealing in this respect.User experience depends on deployment scenario.In an office or dense residential area，the user is usually in a hotspot and can expect a much better experience than if they were in a wide-coverage scenario.The rate desired by someone participating in a big outdoor event in a regional area may not be possible in that area.5G systems need to create an equally good experience regardless of location or scenario.This is what truly ubiquitous service implies.

▲Figure 1.Increase in data rate，generation by generation.

▼Table 1.User experience in different scenarios

2.2 A Mesh of Connections

Cell phone penetration in developed countries is already 100%.As more and more people are expected to replace their old phones with powerful new terminals，such as tablets，the sheer number of cell phones is not expected to increase significantly.Nowadays，a smartphone is so omnipotent that you can talk，message，watch videos，listen to music，play games，chat，and surf the web all from one terminal.Many phones now support multiple standards for a single radio access，which make roaming much easier in many parts of the world.If we consider that cellphones are only used for human-to-human communication，cell phone penetration will remain flat or increase only slightly.However，5G goes beyond traditional cellular services for personal use.A large chunk of traffic will derive from human-to-machine and machine-to-machine communication.The total number of devices that will need to be wirelessly connected in a few industries，e.g.，retail，healthcare，manufacturing，transportation and agriculture，will be much higher than the human population.The total number of machine-to-machine connections will easily be counted in the hundreds of billions［3］.The requirements for the machine-to-machine communication in each industry will be drastically different:some industries will require a very high data rate；some require very short latency；some will require extremely high reli-ability.This creates great challenges for 5G networks.How can we serve vast meshes of human-to-human，human-to-machine，and machine-to-machine connections that far exceed those of 4G networks？Researchers need to come up with intelligent designs the ensure networks are robust and operate smoothly with massive numbers of connections.

Vast meshes of connections calls for 5G to be a conceptual change.In addition to interactive services such as conversation，gaming，video-conferencing and web surfing，5G will have to provide many more automated services for machine-to-machine communication.Network design principles need to be rethought for 5G so that M2M communication is efficient and cost -effective.Compared with other wireless techniques and systems that used to be vertically integrated in each industry，5G is better in terms of required performance and total cost of development and deployment.This would unleash the potential of economy-of-scale，which is often observed in the cellular industry，and would increase efficiency in many parts of society.

2.3 More Energy-Efficient Future

With the thousand-fold increase in capacity engendered by 5G networks，energy efficiency becomes a top priority.Capacity should not just be increased on average；throughput also needs to be significantly increased at the cell edge so that a user is guaranteed a superior wireless experience wherever they are.For service ubiquity and to support a massive number of connections，a 5G network infrastructure has to be very densely deployed.If efficiency remains at current levels，energy consumption will shoot up.

The idea of“going green”has taken root in many industries worldwide.The cellular industry is a major contributor of global CO2emission［4］.In the future，it will not be socially acceptable to chase ultra-fast network speed and excellent user experience at the expense of the environment.Therefore，researchers need to devise smarter ways to present energy information to users，reduce harmful interference caused by aimless transmission，and conform to Moore's Law by further shrinking circuits and reducing power consumption.Increasing energy efficiency involves more cost-effective site planning，construction，and maintenance.These traditionally account for a large proportion of energy consumption within a cellular system.

5G terminals also need to be energy-efficient，and this requires cooperation between system designers and device manufacturers.Moore's Law also applies to terminals as well as networks—the size of circuits in a terminal will continue to reduce，and more complicated data processing will occur with less power and smaller die area.

3 Enabling Technologies

From 1G to 4G，each generation is defined by a standout technology that represents the most important advancement in that generation.Traditionally，each cellular generation is distinguished by its unique multi-access method.For example，the distinguishing feature of 1G was frequency-division multiple access（FDMA）and the reliance on analog-domain signal processing and communication.The distinguishing feature of 2G was time-division multiple access（TDMA），and GSM is a good example of where digital signal processing and communications were first widely used in the cellular world. The distinguishing feature of 3G is code-division multiple access（CDMA），which enables radio resources to be shared between multiple users.The introduction of capacity-approaching Turbo codes significantly increases spectral efficiency.The distinguishing feature of 4G is orthogonal frequency-division multiple access（OFDMA）which，when combined with spatial multiplexing schemes such as MIMO，drastically increase system capacity without unreasonably increasing complexity at receivers.

5G will have diverse requirements and applications and is unlikely to be defined by a single radio-access scheme or dominant technology.However，there are a few noticeable technological trends that could underscore 5G networks.These technological trends will intersect with each other to construct the 5G system as a whole and fulfill the vision outlined earlier.

As with the previous four generations of cellular networks，capacity will continue to be very important technical goal of 5G.Capacity is a“hard”performance criteria and the basis of service ubiquity.There are three issues related to traffic volume that need to be considered in order to achieve a thousandfold increase in capacity（Fig.2）.These aspects issues are:improving radio link efficiency，expanding and managing spectrum，and making cell sites denser.

Another important trend in future wireless networking is the transition to intelligent cloud.Cloud-like concepts are seen in various branches of information technology，and 5G network architecture is no exception.Many kinds of network intelligence will be re-formulated and integrated into the cloud structure. With the cloud，human-to-human，human-to-machine，and ma-chine-to-machine communication will all be seamless.The huge digital ecosystem created by 5G will allow any person or machine to be a content generator or digital library that can potentially accelerate the growth of data traffic in future networks.

▲Figure 2.Three aspects of technology breakthrough for 5G.

3.1 Improved Efficiency of Radio Links

Since the advent of 3G，more technologies related to multiple antennas have been adopted to increase system capacity. In a 3G system，multiple transmitting antennas are primarily used for beamforming，which can improve the signal-to-noise ratio（SNR）of a link.Orthogonal-frequency-division multiplex（OFDM）greatly reduces receiver complexity in a 4G MIMO system，especially in a multipath propagation environment. OFDM has given rise to a wave of spatial multiplexing schemes that have now been standardized.With spatial multiplexing，a radio link can support multilayer transmission，which helps double or even triple the peak data rate of a 4G system compared to a 3G system.The advancement of MIMO in LTE［5］allows more flexible multiuser MIMO，a technology that increases the sum capacity of users who are simultaneously scheduled.

With 5G，there will be order of magnitude increase in the number of antennas at the base station.Dozens or even hundreds of antennas are possible and will form a so-called massive MIMO system［6］.The freedom offered by massive MIMO in the spatial domain is so huge that the effect of thermal noise will become negligible，and system performance will，in theory，only be limited by the pilot pollution.This huge degree of freedom can support high-order multiuser communication，where a large number of users share the same time and frequency resources and do not significantly interfere with each other.Improved SNR and high-order multiuser transmission can provide a several-fold increase in spectral efficiency in the system，although the gain may not be exactly proportional to the number of antennas.

With massive MIMO，radiated energy is directed more towards intended users rather than being radiated in all directions.Link budget can also be dramatically improved.In other words，a transmitter uses less power while providing higher system throughput and wider coverage.In addition，transmit power can be distributed over many antennas in massive MIMO，and the power amplifier of each antenna operates at linear region.This facilitates high-order modulation and coding without using the need for expensive power amplifiers and pre-distortion algorithms.Therefore，massive MIMO will be critical in reducing energy consumption of 5G systems and achieving the goal of green communication.

Turbo codes in 3G and 4G have already pushed the spectral efficiency of a single-antenna channel very close to the Shannon limit；however，there is still some room for more research on new coding and modulation schemes.Channel coding in 3G and 4G is mainly aimed at approaching the Shannon limit for traffic channels on the condition that code blocks are enough long and the channel is only affected by additive white Gaussian noise（AWGN）.In 5G，more flavors will be added to the design of the channel codes and modulation schemes.For vast meshes of connections，the number of links that access to the network is huge，yet each link carries only small amount of data.Because short blocks of data prevail，the channel-coding community is searching for powerful codes that can approach the channel capacity even when the code block is short，e.g.，less than 100 bits［7］.Cellular communication is often corrupted by fast fading，either at frequency domain or time domain. Link adaptation is effective for handling fast fading and helps transmitters choose modulation/coding schemes to match the current channel condition.The link adaptation schemes in 3G and 4G can work but often fail to keep up with the channel variation when users are moving fast.New link-adaptation methods in 5G enable the transmission scheme to be quickly adjusted to suit the channel characteristics.This reduces resource waste and improves link efficiency in fading environment.

Faster than Nyquist（FTN）is another link-level coding scheme that has recently gained attention in 5G.Instead of using traditional QAM to map coded bits to complex symbols，coded bits in FTN are shifted and overlapped in time.This superimposition forms a real-valued convolutional encoder.Relying on the sequence detection at the receiver，coded bits can be detected with low bit-error rate.The amplitude distribution of FTN signals is closer to that of Gaussian signals than that of QAM signals.In an environment with high signal-to-noise-ratio（SNR），F(xiàn)TN more closely approaches the Shannon limit.An FTN signal has lower peak-to-average power ratio（PAPR）than a QAM signal，which allows power amplifiers to operate more efficiently.

Multiuser communication has been around since 3G.Standards specifications define some physical control signaling and reference signal structure（format）to better support multiuser features；however，many multiuser schemes are implementation -oriented.One example of this is linear superposition of signals of different users and then using receiver-side interference cancellation to pick up each user's signal（data）.Interference cancellation at receivers is usually implementation-specific.In 5G，more sophisticated superposition is expected for multiuser support.Such superposition includes code-domain superposition，which more closely approaches the sum-rate of the broadcast channel/multi-access channel or reduces the need for resource scheduling and is less dependent on interference cancellation at receivers.These superposition codes form new family of new coding and modulation techniques that includes bit-division multiplexing［9］.

At the system level，network coding should increase the total throughput of a system when it has multihop transmissions. Relay node deployment is an example of multi-hop systems，especially when the backhaul link（base station to relay node）and access link（relay node to user terminal）share the samefrequency.So far，network coding has never gone beyond academia.This situation will change with 5G，and we may see the widespread adoption of network coding in various standards.

With an increase in processing power at the receivers，interference cancellation can be brought to a higher level so that transmission and reception can occur at the same time in the same frequency［10］with little spatial isolation in-between. With analog-and digital-domain interference cancellation，full duplex communication is possible in 5G.In theory，link efficiency will be doubled if all co-existence issues can be solved.

In applications such as machine-to-machine communication for manufacturing or vehicle-to-vehicle communication for intelligent transport systems，short latency and high reliability are critical requirements.3G and 4G systems were designed primarily for human-to-human communication，and their physical-layer structures and higher-layer protocols do not meet these stringent latency and reliability requirements.Physical structures with shorter transmission time interval（TTI）are needed in 5G so that the number of residual errors is made extremely low within a very short timeframe.New coding and modulation schemes will help achieve this goal as well.

OFDM，the widely used waveform in 4G，enables simple receiver implementation in MIMO systems.However，OFDM signals tend to have significant out-of-band emissions，and this requires precise synchronization and orthogonal resources.Filter-bank multi-carrier（FBMC）is a promising waveform technology that reduces out-of-band emissions and lowers the requirement for synchronization.The basic idea of FBMC is to replace the rectangular window in OFDM with a bank of filters.In MIMO systems，F(xiàn)BMC requires some extra signal processing in channel estimation and filter algorithms in MIMO systems，F(xiàn)BMC is very suitable for dynamic spectrum allocation scenarios where the low out-of-band emission makes the systems more compatible with various band combinations.

Radio link efficiency can also be improved by using a software-defined air interface so that the system can support multiple radio access technologies（multi-RAT）in a multi-spectrum deployment.Such operational flexibility can improve the utilization of radio resources and result in better 5G performance.

3.2 Spectrum Expansion and Management

Data rate is the product of spectral efficiency and occupied bandwidth.As cellular communication evolves from 1G to 4G，system bandwidth continuously increases，as does the bandwidth each user can occupy.Over the past few decades，link-level spectral efficiency has increased significantly；however，it is wideband operation that really provides order-of-magnitude gain in system capacity and peak rate.Each user now has access to wide bandwidth because of the fast development of radio frequency components and digital communication processors.Nowadays，terminal chipsets are so powerful that they support bandwidth well beyond 20 MHz with 64 QAM and multilayer MIMO transmission.User experience is greatly improved by this.

Cellular bands are traditionally allocated below 3 GHz，where scattering，reflection and diffraction help the electromagnetic（EM）wave uniformly shine on the targeted terminals. This can fill coverage holes when shadow fading is present.Below 3 GHz，an EM wave also usually has less penetration loss when passing through walls and windows.So EM signals can penetrate deeper into buildings and reach indoor users.Such superb link budget in this lower spectrum enables an operator to deploy fewer base stations to cover an area and reduce capex and opex.This is why in many countries 700 MHz，which was formerly used for TV broadcasting，is now being re-farmed for cellular use.The propagation characteristics at 700 MHz are very favorable for coverage.

Wireless spectrum is a scarce resource and is regulated for satellite，military，scientific，and radio and TV uses，not just for cellular communication.Over the past four generations of cellular development，the spectrum below 3 GHz regulated for cellular services has been used for terrestrial communication. Unless 2G is phased out soon and its occupied bands are refarmed，there is little room for spectrum expansion below 3 GHz.This problem is compounded by the fact that spectrum allocations are very fragmented in lower bands.It is extremely difficult to find a wide and contiguous band at low frequencies，which means a high data rate is very unlikely.Propagation characteristics between 3 GHz and 6 GHz are slightly different from those below 3 GHz.Between 3 GHz and 6 GHz，there is less scattering and path loss tends to be a little higher，although the fundamental propagation mechanism remains the same as that in the lower bands.A lot of un-licensed spectrum is regulated in the 3-6 GHz range，e.g.，Wi-Fi which provides cost-effective solutions for local wireless access.5G will not only operate in the licensed spectrum；it will also be compatible with technologies that are tailored to unlicensed spectrum.In this way，spectrum can be shared，and network potential can be further unleashed.3GPP［11］has started to look into this opportunity in their recently proposed study item［11］.Fig.3 shows more choices for spectrum sharing.

Spectrum shortage can be largely solved by using high bands，e.g.，above 6 GHz or even millimeter wavelength.Traditionally，high-frequency communication has been limited to point-to-point communication，e.g.，between macro base stations.There are several reasons for this.First，base station antennas are usually well above the building's roofline，which ensures line of sight（LOS）and compensates for the heavier pathloss of free-space propagation at higher bands.Second，the transmitter and receiver are all stationary so that antennas with sharp directivity can be used to further improve the channel SNR.Third，the cost of implementing high-frequency RF and chipsets is not a major concern given the small number of wireless backhauls needed and the cost of devices compared to the total cost of a macro base station.

▲Figure 3.Choices for spectrum sharing.

There is a strong motivation to use high-frequency bands for cellular access in 5G.The relatively abundant spectrum above 6 GHz makes it possible to allocate a contiguous band with huge bandwidth，e.g.，500 MHz，so that wireless subscribers have a super-fast experience.Indoor hotspots，where access points are deployed indoors，are especially suitable for high-frequency communication because short distance does not require high-powered transmission.Greater penetration loss helps isolate each access point；therefore，less inter-accesspoint interference is expected.

High-frequency bands might also be used for wide-area communication.There are many challenges:heavy path loss needs to be addressed so that users deep within the building or far from the base station are guaranteed coverage.In a wide-area scenario，the user is usually moving.Because of the short wavelength，moderate traveling speed can result in very high Doppler Effect.Mobility management therefore becomes more difficult.

One benefit of using a shorter wavelength is the increased effective aperture of both the transmitter and receiver antennas when the antenna size is kept the same.Massive MIMO operating in high frequency bands can be very compact and highly integrated for lower cost.This significantly increases the opportunities to deploy massive MIMO，which may be deployed for access points.

For high-frequency communication，the most challenging issue may be devices.Despite their low cost and ease of integration，traditional silicon-based chips may not be capable of providing the required processing speed，noise level，or energy efficiency.New materials such as Gallium-silicon are being studied with the prospect of delivering good performance for lower cost.

Spectrum management is not only a technological issue；it is also a political issue.Regulating spectrum involves balancing the interests of various parties，some of which may have historical rivalries.That is why the ITU is cautious about discussing and allocating spectrum.Right now，the ITU is still working on WRC-15 and is focused on bands under 6 GHz.Discussion on spectrum allocation of bands above 6 GHz for cellular use will not start until 2018.Consequently，work on the full standardization of 5G high-frequency communication will not kick off until after 2018.

3.3 Cell Site Densification

For hotspot scenarios，such as offices and dense urban apartments，cell density needs to be drastically increased to accommodate huge volumes of traffic within a small geographical area.In the first three generations of cellular networks，cell layouts had rather homogenous topologies，e.g，system base stations had the same configurations for transmit power and antenna gain.Base stations were more or less equally distanced.In 4G，heterogeneous networks（HetNets）came into the picture.Low-power nodes（LPN），such as pico，femto and relay［12］nodes，combined with macro base stations to provide high throughput via cell-splitting gain（Fig.4）.In a HetNets scenario，cell splitting can be viewed as offloading traffic from macro base stations to low-power nodes，which help fill the coverage holes in a homogeneous layout so coverage approaches 100%.Users who are not in hotspots are still attached to macro base stations.The effect on mobility management is minimal because users in hotspots are not expected to be fast-moving.

In 5G，many more LPNs are deployed within a macro area，resulting in less than 20 meters inter-node distance.With such closeness，even the building penetration may not be enough to reduce interference from neighboring LPNs.With an outdoor gathering，more inter-site interference is expected.Advanced interference coordination and cooperative transmission schemes are crucial to ensure reasonably good spectral efficiency per LPN and good user experience，regardless of whether the user is at the center or edge of the cell.

As the cell site density is increased，it is more difficult to choose sites that have wired backhaul connected to the networks.In cities with historical sites，preserving old streets and structures is a high priority，and it is almost impossible to break new ground to lay down telecommunications cables.For systems with a large number of LPNs，the cost of installing and maintaining the cable backhaul is daunting.Therefore，there is strong motivation for wireless backhaul.Conventional wireless backhaul uses proprietary technologies and often operates in non-cellular bands at high frequency.In 5G，wireless backhaul will be based on standard air interface and work for cellular bands that may not necessarily be high-frequency.Standardization can help achieve economy of scale，and low-to-medium frequency bands enable wireless backhaul to work in diverse en-vironments，including non-line-of-sight（NLOS）environments. Consequently，capex and opex can be minimized，even with super-dense cell deployment.

▲Figure 4.Evolution of cell topology.

Device-to-device（D2D）communication is a special case of network densification［13］.A D2D-capable device can act as a low-power node for unicasting，multicasting or broadcasting traffic directly to the user without being routed through the network.D2D is especially useful for proximity services where users in the vicinity share and exchange local information.D2D communication helps increase the density of low-power nodes with wireless backhaul.In 5G networks，D2D is expected to increase system capacity，particularly in dense urban environments and for big outdoor events.

3.4 Cloud Coordination of Network Access

By the time 5G systems are deployed，many 3G and 4G networks（and even some 2G networks）will still be in use.Within each generation of network，the allocated spectra may be different，depending on operator，country，or year of deployment.Radio resources may also include lightly licensed and unlicensed spectrum where technologies of wireless LANs dominate.In this sense，5G networks will be a mix of new network components and existing systems and assets as well as radio access technologies（RATs）of a non-cellular kind.Cloud architecture will coordinate different types of radio and network resources and manage inter-RAT and inter-frequency radio access in a seamless and transparent manner.Multi-RAT convergence is possible in 5G with a unified cloud architecture（Fig.5）.

Cell densification worsens interference between cells.Potential solutions include advanced interference coordination or cooperative transmission.The integrated nature of cloud means that more dynamic interference coordination is necessary to significantly improve network performance at the cell edge.In addition，a significant amount of capex and opex can be saved by using cloud because a lot of the DSP can be centralized and implemented in common-purpose digital processors.Less effort in terms of customization and higher utilization of digital processors saves cost.Energy can also be saved because many processors can share the same air conditioner.

3.5 Flatter Architecture

To increase commercial viability and versatility，flatter architecture has been recommended for 5G.Virtualization and software-defined networking（SDN）are promising techniques for reducing the complexity of 5G networks and optimizing the system performance.In 5G networks，the control plane will be centralized and the user plane will be streamlined.With a flatter architecture，an operator can focus on value-added features or services.Flatter architecture is helps reduce latency；low latency is critical in machine-to-machine and vehicle-to-vehicle communication.With flat architecture，services and networks can be deeply converged.

SDN encourages innovation in network protocols and rapid deployment of networks by operators.With SDN，traffic flow can be controlled in more flexible way.Centralized control also enables more coordinated management of traffic flow and network resources.The OpenFlow protocol currently being discussed still has issues in mobile scenarios.

Figure 5.?Cloud coordination for network access.

An indispensable aspect of flat architecture is network function virtualization（NFV）.The key idea behind NFV is decoupling node functions from node hardware.Standard high-performance hardware replaces special equipment that has been cus-tomized for each node.This can simplify the design of the hardware platform and reduce network cost.Not only is the deployment more flexible，but NFV also encourages openness and innovation in equipment manufacturing.Nevertheless，issues such as software portability，interoperability，and stability need to be addressed before NFV can be widely used.

With flatter architecture，a content-delivery network can have advanced technology such as a smart content router，which redirects user requests to the nearest web cache.This significantly improves responsiveness and user experience，especially for data-heavy applications such as high-definition video.

4 Conclusions

Service ubiquity，a massive number of connections，and energy efficiency are the three key requirements for 5G networks. 5G technologies need to support massive traffic volume by helping increase radio link efficiency，expand spectrum to high -frequency bands，and densify cells.Networks should also be transitioned to cloud architecture，which is flatter and can coordinate radio resources of multi-RATs and intercell interference arising from dense network deployment.

Acknowledgement

The authors would like to thank Xinhui Wang and Longming Zhu at ZTE for the contribution to an earlier white paper which provides some of the foundations of this paper.

［1］ITU.（2012）.IMT for 2020 and beyond［Online］.Available:http://www.itu.int/en/ ITU-R/study-groups/rsg5/rwp5d/imt-2020

［2］F.Vogelstein，“The untold story:how the iPhone blew up the wireless industry，”Wired Magazine，issue 16.02，January 2008.

［3］P.Popovski，V.Braun，H.-P.Mayer，et al.，“Scenarios，requirements and KPIs for 5G mobile and wireless system，”METIS，ICT-317669-METIS/D1.1，April 2013.

［4］The Climate Group，“Smart 2020:Enabling the low-carbon economy in the information age，”2008.

［5］Y.Yuan，LTE/LTE-Advanced Key Technologies and System Performance，Beijing，China:Posts&Telecom Press，Jun.2013.

［6］T.L.Marzetta，“Noncooperative cellular wireless with unlimited numbers of base station antennas，”IEEE Transactions on Wireless Communications，vol.9，no.11，pp.3590-3600，Nov.2010.doi:10.1109/TWC.2010.092810.091092.

［7］M.Dohler，R.W.Heath，A.Lozano，et al.，“Is the PHY layer dead，”IEEE Communications Magazine，vol.49，no.4，pp.159-161，Apr.2011.doi:10.1109/ MCOM.2011.5741160.

［8］J.B.Anderson，F(xiàn).Rusek，and V.Owall，“Fast-than-Nyquist signaling，”Proceedings of the IEEE，vol.101，no.8，pp.1817-1830，Aug.2013.doi:10.1109/ JPROC.2012.2233451.

［9］H.Jin，K.Peng，and J.Song，“Bit division multiplexing for broadcasting，”IEEE Transactions on Broadcasting，vol.59，no.3，pp.539-547，Sept.2013.doi: 10.1109/TBC.2013.2254269.

［10］J.II Choi，S.Hong，M.Jain，et al.，“Beyond full duplex wireless，”in 2012 Conference Record of the Forty Sixth Asilomar Conference on Signals，Systems and Computers，Pacific Grove，USA，pp.40-44.doi:10.1109/ACSSC.2012. 6488954.

［11］Ericsson，“Study on Licensed-assisted access using LTE，”3GPP RP-141397 RAN#65，Sept.2014.

［12］Y.Yuan，LTE-Advanced Relay Technology and Standardization，New York City，USA:Springer，Jul.2012.

［13］N.Bhushan，J.Li，D.Malladi，R.Gilmore，D.Brenner，A.Damnjanovic，R.T. Sukhavasi，C.Patel，and S.Geirhofer，“Network densification:the dominant theme for wireless evolution in 5G，”IEEE Communications Magazine，pp.82-89，F(xiàn)eb.2014.doi:10.1109/MCOM.2014.6736747.

Manuscript received:2014-09-06

Biographiesphies

Yifei Yuan（yuan.yifei@zte.com.cn）received his BS and MS degrees from Tsinghua University，China.He received his PhD from Carnegie Mellon University，USA. From 2000 to 2008，he worked with Alcatel-Lucent on 3G and 4G key technologies. Since 2008，he has worked for ZTE researching 5G technologies and standards for LTE-Advanced physical layer.His research interests include MIMO，iterative codes，and resource scheduling.He was admitted to the Thousand Talent Plan Program of China in 2010.He has written a book on LTE-A relay and a book on LTEAdvanced key technologies and system performance.He has had more than 30 patents approved.

Xiaowu Zhao（zhao.xiaowu@zte.com.cn）received his PhD degree from the Institute of Software，Chinese Academy of Science，in 2001.He has participated in the Third Generation Partnership Project 2（3GPP2）since 2001.He was the 3GPP2 Technical Specification Group-Access Network Interface（TSG-A）chair from 2009 to 2012 and System Aspect and Core Network（TSG-SX）chair from 2013 to the present.In recent year，he has also participated in the 3GPP RAN plenary meeting.He has submitted hundreds of contributions that have been accepted by 3GPP2.His current research interests include beyond LTE-Advanced and 5G technology and taking charge of the whole industry standardization of ZTE Corporation.