Cellular Architecture and Key Technologies for 5G Wireless Networks (1) (1).pdf

WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 122 5G WIRELESS COMMUNICATION SYSTEMS: PROSPECTS AND CHALLENGES Cellular Architecture and Key Technologies for 5G Wireless Communication Networks Cheng-Xiang Wang, Heriot-Watt University and University...

WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 122 5G WIRELESS COMMUNICATION SYSTEMS: PROSPECTS AND CHALLENGES Cellular Architecture and Key Technologies for 5G Wireless Communication Networks Cheng-Xiang Wang, Heriot-Watt University and University of Tabuk Fourat Haider, Heriot-Watt University Xiqi Gao and Xiao-Hu You, Southeast University Yang Yang, ShanghaiTech University Dongfeng Yuan, Shandong University Hadi M. Aggoune, University of Tabuk Harald Haas, University of Edinburgh Simon Fletcher, NEC Telecom MODUS Ltd. Erol Hepsaydir, Hutchison 3G UK ABSTRACT The European Mobile Observatory (EMO) reported that the mobile communication sector The fourth generation wireless communica- had total revenue of €174 billion in 2010, there- tion systems have been deployed or are soon to by bypassing the aerospace and pharmaceutical be deployed in many countries. However, with sectors. The development of wireless tech- an explosion of wireless mobile devices and ser- nologies has greatly improved people’s ability to vices, there are still some challenges that cannot communicate and live in both business opera- be accommodated even by 4G, such as the spec- tions and social functions. trum crisis and high energy consumption. Wire- The phenomenal success of wireless mobile less system designers have been facing the communications is mirrored by a rapid pace of continuously increasing demand for high data technology innovation. From the second genera- rates and mobility required by new wireless tion (2G) mobile communication system debuted applications and therefore have started research in 1991 to the 3G system first launched in 2001, on fifth generation wireless systems that are the wireless mobile network has transformed expected to be deployed beyond 2020. In this from a pure telephony system to a network that article, we propose a potential cellular architec- can transport rich multimedia contents. The 4G ture that separates indoor and outdoor scenar- wireless systems were designed to fulfill the ios, and discuss various promising technologies requirements of International Mobile Telecom- for 5G wireless communication systems, such as munications-Advanced (IMT-A) using IP for all massive MIMO, energy-efficient communica- services. In 4G systems, an advanced radio tions, cognitive radio networks, and visible light interface is used with orthogonal frequency-divi- communications. Future challenges facing these sion multiplexing (OFDM), multiple-input multi- potential technologies are also discussed. ple-output (MIMO), and link adaptation technologies. 4G wireless networks can support INTRODUCTION data rates of up to 1 Gb/s for low mobility, such as nomadic/local wireless access, and up to 100 The innovative and effective use of information Mb/s for high mobility, such as mobile access. and communication technologies (ICT) is Long-Term Evolution (LTE) and its extension, becoming increasingly important to improve the LTE-Advanced systems, as practical 4G systems, economy of the world. Wireless communica- have recently been deployed or soon will be tion networks are perhaps the most critical ele- deployed around the globe. ment in the global ICT strategy, underpinning However, there is still a dramatic increase in many other industries. It is one of the fastest the number of users who subscribe to mobile growing and most dynamic sectors in the world. broadband systems every year. More and more 122 0163-6804/14/$25.00 © 2014 IEEE IEEE Communications Magazine February 2014 WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 123 people crave faster Internet access on the move, Networks described how the underlying radio trendier mobiles, and, in general, instant com- access technologies can be developed further to One of the key ideas munication with others or access to information. support up to 1000 times higher traffic volumes of designing the 5G More powerful smartphones and laptops are compared to 2010 travel levels over the next 10 becoming more popular nowadays, demanding years. Samsung demonstrated a wireless sys- cellular architecture is advanced multimedia capabilities. This has tem using millimeter (mm) wave technologies to separate outdoor resulted in an explosion of wireless mobile with data rates faster than 1 Gb/s over 2 km. devices and services. The EMO pointed out that What will the 5G network, which is expected and indoor scenarios there has been a 92 percent growth in mobile to be standardized around 2020, look like? It is so that penetration broadband per year since 2006. It has been now too early to define this with any certainty. loss through building predicted by the Wireless World Research However, it is widely agreed that compared to Forum (WWRF) that 7 trillion wireless devices the 4G network, the 5G network should achieve walls can be some- will serve 7 billion people by 2017; that is, the 1000 times the system capacity, 10 times the how avoided. This number of network-connected wireless devices spectral efficiency, energy efficiency and data will reach 1000 times the world’s population. rate (i.e., peak data rate of 10 Gb/s for low will be assisted by As more and more devices go wireless, many mobility and peak data rate of 1 Gb/s for high distributed antenna research challenges need to be addressed. mobility), and 25 times the average cell through- system (DAS) and One of the most crucial challenges is the put. The aim is to connect the entire world, and physical scarcity of radio frequency (RF) spectra achieve seamless and ubiquitous communica- massive MIMO allocated for cellular communications. Cellular tions between anybody (people to people), any- technology. frequencies use ultra-high-frequency bands for thing (people to machine, machine to machine), cellular phones, normally ranging from several wherever they are (anywhere), whenever they hundred megahertz to several gigahertz. These need (anytime), by whatever electronic frequency spectra have been used heavily, mak- devices/services/networks they wish (anyhow). ing it difficult for operators to acquire more. This means that 5G networks should be able to Another challenge is that the deployment of support communications for some special sce- advanced wireless technologies comes at the cost narios not supported by 4G networks (e.g., for of high energy consumption. The increase of high-speed train users). High-speed trains can energy consumption in wireless communication easily reach 350 up to 500 km/h, while 4G net- systems causes an increase of CO2 emission indi- works can only support communication scenarios rectly, which currently is considered as a major up to 250 km/h. In this article, we propose a threat for the environment. Moreover, it has potential 5G cellular architecture and discuss been reported by cellular operators that the some promising technologies that can be energy consumption of base stations (BSs) con- deployed to deliver the 5G requirements. tributes to over 70 percent of their electricity bill The remainder of this article is organized as. In fact, energy-efficient communication was follows. We propose a potential 5G cellular not one of the initial requirements in 4G wire- architecture. We describe some promising key less systems, but it came up as an issue at a later technologies that can be adopted in the 5G sys- stage. Other challenges are, for example, aver- tem. Future challenges are highlighted. Finally, age spectral efficiency, high data rate and high conclusions are drawn. mobility, seamless coverage, diverse quality of service (QoS) requirements, and fragmented user experience (incompatibility of different A POTENTIAL 5G WIRELESS wireless devices/interfaces and heterogeneous networks), to mention only a few. CELLULAR ARCHITECTURE All the above issues are putting more pres- To address the above challenges and meet the sure on cellular service providers, who are facing 5G system requirements, we need a dramatic continuously increasing demand for higher data change in the design of cellular architecture. We rates, larger network capacity, higher spectral know that wireless users stay indoors for about efficiency, higher energy efficiency, and higher 80 percent of time, while only stay ourdoors mobility required by new wireless applications. about 20 percent of the time. The current On the other hand, 4G networks have just about conventional cellular architecture normally uses reached the theoretical limit on the data rate an outdoor BS in the middle of a cell communi- with current technologies and therefore are not cating with mobile users, no matter whether they sufficient to accommodate the above challenges. stay indoors or outdoors. For indoor users com- In this sense, we need groundbreaking wireless municating with the outdoor BS, the signals have technologies to solve the above problems caused to go through building walls, and this causes very by trillions of wireless devices, and researchers high penetration loss, which significantly dam- have already started to investigate beyond 4G ages the data rate, spectral efficiency, and ener- (B4G) or 5G wireless techniques. The project gy efficiency of wireless transmissions. UK-China Science Bridges: (B)4G Wireless Mobile One of the key ideas of designing the 5G cel- Communications (http://www.ukchinab4g. ac.uk/) is lular architecture is to separate outdoor and perhaps one of the first projects in the world to indoor scenarios so that penetration loss through start B4G research, where some potential B4G building walls can somehow be avoided. This will technologies were identified. Europe and China be assisted by distributed antenna system (DAS) have also initiated some 5G projects, such as and massive MIMO technology , where geo- METIS 2020 (https://www.metis2020. com/) sup- graphically distributed antenna arrays with tens ported by EU and National 863 Key Project in or hundreds of antenna elements are deployed. 5G supported by the Ministry of Science and While most current MIMO systems utilize two Technology (MOST) in China. Nokia Siemens to four antennas, the goal of massive MIMO sys- IEEE Communications Magazine February 2014 123 WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 124 tems is to exploit the potentially large capacity rate services with reduced signaling overhead. The 5G cellular archi- gains that would arise in larger arrays of anten- The above proposed 5G heterogeneous cellular nas. Outdoor BSs will be equipped with large architecture is illustrated in Fig. 1. tecture should also antenna arrays with some antenna elements (also be a heterogeneous large antenna arrays) distributed around the cell one, with macrocells, and connected to the BS via optical fibers, bene- PROMISING KEY fiting from both DAS and massive MIMO tech- microcells, small cells, nologies. Outdoor mobile users are normally 5G WIRELESS TECHNOLOGIES and relays. To equipped with limited numbers of antenna ele- In this section, based on the above proposed ments, but they can collaborate with each other heterogeneous cellular architecture, we discuss accommodate high- to form a virtual large antenna array, which some promising key wireless technologies that mobility users such together with BS antenna arrays will construct can enable 5G wireless networks to fulfill perfor- as users in vehicles virtual massive MIMO links. Large antenna mance requirements. The purpose of developing arrays will also be installed outside of every these technologies is to enable a dramatic capac- and high-speed building to communicate with outdoor BSs or ity increase in the 5G network with efficient uti- trains, we have pro- distributed antenna elements of BSs, possibly lization of all possible resources. Based on the with line of sight (LoS) components. Large anten- well-known Shannon theory, the total system posed the mobile na arrays have cables connected to the wireless capacity Csum can be approximately expressed by femtocell concept, access points inside the building communicating which combines the with indoor users. This will certainly increase the ⎛ P ⎞ infrastructure cost in the short term while signifi- Csum ≈ ∑ ∑ Bi log 2 ⎜ 1 + i ⎟ concepts of mobile cantly improving the cell average throughput, HetNets Channels ⎝ Np ⎠ (1) relay and femtocell. spectral efficiency, energy efficiency, and data rate of the cellular system in the long run. where Bi is the bandwidth of the ith channel, Pi Using such a cellular architecture, as indoor is the signal power of the ith channel, and N p users only need to communicate with indoor denotes the noise power. From Eq. 1, it is clear wireless access points (not outdoor BSs) with that the total system capacity Csum is equivalent large antenna arrays installed outside build- to the sum capacity of all subchannels and het- ings, many technologies can be utilized that are erogeneous networks. To increase Csum, we can suitable for short-range communications with increase the network coverage (via heteroge- high data rates. Some examples include WiFi, neous networks with macrocells, microcells, femtocell, ultra wideband (UWB), mm-wave small cells, relays, MFemtocell , etc.), num- communications (3–300 GHz) , and visible ber of subchannels (via massive MIMO , spa- light communications (VLC) (400–490 THz) tial modulation [SM] , cooperative MIMO,. It is worth mentioning that mm-wave and DAS, interference management, etc.), bandwidth VLC technologies use higher frequencies not (via CR networks , mm-wave communica- traditionally used for cellular communications. tions, VLC , multi-standard systems, etc.), These high-frequency waves do not penetrate and power (energy-efficient or green communi- solid materials very well and can readily be cations). In the following, we focus on some of absorbed or scattered by gases, rain, and the key technologies. foliage. Therefore, it is hard to use these waves for outdoor and long distance applications. MASSIVE MIMO However, with large bandwidths available, mm- MIMO systems consist of multiple antennas at wave and VLC technologies can greatly both the transmitter and receiver. By adding increase the transmission data rate for indoor multiple antennas, a greater degree of freedom scenarios. To solve the spectrum scarcity prob- (in addition to time and frequency dimensions) lem, besides finding new spectrum not tradi- in wireless channels can be offered to accom- tionally used for wireless services (e.g., modate more information data. Hence, a signif- mm-wave communications and VLC), we can icant performance improvement can be also try to improve the spectrum utilization of obtained in terms of reliability, spectral effi- existing radio spectra, for example, via cogni- ciency, and energy efficiency. In massive MIMO tive radio (CR) networks. systems, the transmitter and/or receiver are The 5G cellular architecture should also be a equipped with a large number of antenna ele- heterogeneous one, with macrocells, microcells, ments (typically tens or even hundreds). Note small cells, and relays. To accommodate high- that the transmit antennas can be co-located or mobility users such as users in vehicles and high- distributed (i.e., a DAS system) in different speed trains, we have proposed the mobile applications. Also, the enormous number of femtocell (MFemtocell) concept , which receive antennas can be possessed by one device combines the concepts of mobile relay and fem- or distributed to many devices. Besides inherit- tocell. MFemtocells are located inside vehicles ing the benefits of conventional MIMO sys- to communicate with users within the vehicle, tems, a massive MIMO system can also while large antenna arrays are located outside significantly enhance both spectral efficiency the vehicle to communicate with outdoor BSs. and energy efficiency. Furthermore, in mas- An MFemtocell and its associated users are all sive MIMO systems, the effects of noise and viewed as a single unit to the BS. From the user fast fading vanish, and intracell interference point of view, an MFemtocell is seen as a regu- can be mitigated using simple linear precoding lar BS. This is very similar to the above idea of and detection methods. By properly using multi- separating indoor (inside the vehicle) and out- user MIMO (MU-MIMO) in massive MIMO door scenarios. It has been shown in that systems, the medium access control (MAC) users using MFemtocells can enjoy high-data- layer design can be simplified by avoiding com- 124 IEEE Communications Magazine February 2014 WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 125.......... The 5G CR network cha LO nel.......... Internet is an innovative soft- S n......... ware defined radio Large MIMO technique which has network Core network been considered as... one of the promising Mobile femtocell...... network technologies to..... improve the utiliza- tion of the congest- Internet ed RF spectrum. CR network Internet Adopting CR is moti- vated by the fact that a large portion Femtocell of the radio spec- VLC trum is underutilized WiFi 60 GHz most of the time. t Etherne Gigabit Figure 1. A proposed 5G heterogeneous wireless cellular architecture. plicated scheduling algorithms. With MU- as an example. The receiver can then employ MIMO, the BS can send separate signals to optimal maximum likelihood (ML) detection to individual users using the same time-frequency decode the received signal. resource, as first pro. Consequently, these main Spatial modulation can mitigate three major advantages enable the massive MIMO system problems in conventional MIMO systems: inter- to be a promising candidate for 5G wireless channel interference, inter-antenna synchroniza- communication networks. tion, and multiple RF chains. Moreover, low-complexity receivers in SM systems can be SPATIAL MODULATION designed and configured for any number of Spatial modulation, as first proposed by Haas et transmit and receive antennas, even for unbal- al., is a novel MIMO technique that has been anced MIMO systems. We have to point out that proposed for low-complexity implementation of the multiplexing gain in SM increases logarith- MIMO systems without degrading system per- mically with the increase in the number of trans- formance. Instead of simultaneously trans- mit antennas, while it increases linearly in mitting multiple data streams from the available conventional MIMO systems. Therefore, the low antennas, SM encodes part of the data to be implementation complexity comes at the expense transmitted onto the spatial position of each of sacrificing some degrees of freedom. Most transmit antenna in the antenna array. Thus, research on SM focuses on the case of a single the antenna array plays the role of a second (in receiver (i.e., single-user SM). Multi-user SM addition to the usual signal constellation dia- can be considered as a new research direction to gram) constellation diagram (the so-called spa- be considered in 5G wireless communication sys- tial constellation diagram), which can be used tems. to increase the data rate (spatial multiplexing) with respect to single-antenna wireless systems. COGNITIVE RADIO NETWORKS Only one transmit antenna is active at any time, The CR network is an innovative software while other antennas are idle. A block of infor- defined radio technique considered to be one of mation bits is split into two sub-blocks of the promising technologies to improve the uti- log2(NB) and log2(M) bits, where NB and M are lization of the congested RF spectrum. the number of transmit antennas and the size Adopting CR is motivated by the fact that a of the complex signal constellation diagram, large portion of the radio spectrum is underuti- respectively. The first sub-block identifies the lized most of the time. In CR networks, a sec- active antenna from a set of transmit antennas, ondary system can share spectrum bands with while the second sub-block selects the symbol the licensed primary system, either on an inter- from the signal constellation diagram that will ference-free basis or on an interference-tolerant be sent from that active antenna. Therefore, basis. The CR network should be aware of SM is a combination of space shift keying (SSK) the surrounding radio environment and regulate and amplitude/phase modulation. Figure 2 its transmission accordingly. In interference-free shows the SM constellation diagram with 4 CR networks, CR users are allowed to borrow transmit antennas (N B = 4) and quadrature spectrum resources only when licensed users do phase shift keying (QPSK) modulation (M = 4) not use them. A key to enabling interference- IEEE Communications Magazine February 2014 125 WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 126 MOBILE FEMTOCELL Im The MFemtocell is a new concept that has been proposed recently to be a potential candidate technology in next generation intelligent trans- portation systems. It combines the mobile relay concept (moving network) with femtocell technology. An MFemtocell is a small cell that can move around and dynamically change its 01(00) connection to an operator’s core network. It can be deployed on public transport buses, trains, and even private cars to enhance service 10(00) 00(00) quality to users within vehicles. Deployment of Im MFemtocells can potentially benefit cellular 11(00) networks. First, MFemtocells can improve the 00 Re. spectral efficiency of the entire network. To. Signal constellation for the demonstrate this fact, Fig. 4 compares the aver-. first transmit antenna. age spectral efficiency of the direct transmis-. sion scheme and an MFemtocell-enhanced 01 scheme with two resource partitioning schemes (i.e., orthogonal and non-orthogonal resource partitioning schemes) as a function of the per- 01(11) centage of users associated with MFemtocells. 10 Also, the comparison is done between maxi- 00(11) mum signal-to-noise ratio (MAX-SNR) and 10(11) proportional fairness (PF) scheduling algo- 11(11) rithms. We can see that increasing the percent- 11 Re age of users that communicate with the BS Signal constellation for the second transmit antenna through MFemtocells leads to an increase in Spatial constellation spectral efficiency, which is much better com- pared to the case where users communicate Figure 2. SM constellation diagram using four transmit antennas (NB = 4) directly with the BS (i.e., the direct transmis- and QPSK modulation. sion scheme). Second, MFemtocells can con- tribute to signaling overhead reduction of the network. For instance, an MFemtocell can per- form a handover on behalf of all its associated free CR networks is figuring out how to detect users, which can reduce the handover activities the spectrum holes (white space) that spread out for users within the MFemtocell. This makes in wideband frequency spectrums. CR receivers the deployment of MFemtocells suitable for should first monitor and allocate the unused high-mobility environments. In addition, the spectrums via spectrum sensing (or combining energy consumption of users inside an MFem- with geolocation databases) and feed this infor- tocell can be reduced due to relatively shorter mation back to the CR transmitter. A coordinat- communication range and low signaling over- ing mechanism is required in multiple CR head. networks that try to access the same spectrum to prevent users colliding when accessing the VISIBLE LIGHT COMMUNICATION matching spectrum holes. In interference-toler- Visible light communication uses off-the-shelf ant CR networks, CR users can share the spec- white light emitting diodes (LEDs) used for trum resource with a licensed system while solid-state lighting (SSL) as signal transmitters keeping the interference below a threshold. In and off-the-shelf p-intrinsic-n (PIN) photodi- comparison with interference-free CR networks, odes (PDs) or avalanche photo-diodes (APDs) interference-tolerant CR networks can achieve as signal receivers. This means that VLC enhanced spectrum utilization by opportunisti- enables systems that illuminate and at the same cally sharing the radio spectrum resources with time provide broadband wireless data connectiv- licensed users, as well as better spectral and ity. If illumination is not desired in the uplink, energy efficiency. However, it has been shown infrared (IR) LEDs or indeed RF would be that the performance of CR systems can be very viable solutions. In VLC, the information is car- sensitive to any slight change in user densities, ried by the intensity (power) of the light. As a interference threshold, and transmission behav- result, the information-carrying signal has to be iors of the licensed system. This fact is illustrated real valued and strictly positive. Traditional dig- in Fig. 3, where we notice that the spectral effi- ital modulation schemes for RF communication ciency decreases quickly with the increase in the use complex valued and bipolar signals. Modifi- number of primary receivers. However, the spec- cations are therefore necessary, and there is a tral efficiency can be improved by either relaxing rich body of knowledge on modified multi-carri- the interference threshold of the primary system er modulation techniques such as OFDM for or considering only the CR users who have short intensity modulation (IM) and direct detection distances to the secondary BS. In , hybrid (DD). Data rates of 3.5 Gb/s have been report- CR networks have been proposed for adoption ed from a single LED. It has to be noted that in cellular networks to explore additional bands VLC is not subject to fast fading effects as the and expand the capacity. wavelength is significantly smaller than the 126 IEEE Communications Magazine February 2014 WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 127 detector area. While the link-level demonstra- tions are important steps to prove that VLC is a 14 viable technique to help mitigate spectrum bot- Simulation results Analytical results tlenecks in RF communications, it is essential to 12 show that full-fledged optical wireless networks can be developed by using existing lighting infrastructures. This includes MU access tech- 10 Spectral efficiency (b/s/Hz) niques, interference coordination, and others. To this end, let us assume multiple light fixtures in a room. Each of these light fixtures is envis- 8 aged to function as a very small optical BS resulting in a network of very small cells called Q = 20 dB optical attocells. This is in analogy to femtocells 6 in RF communications and in recognition of the fact that a single room can be served by many of 4 these very small cells. An optical attocell covers Q = 10 dB an area of 1–10 m2 and distances of about 3 m. It is well known in cellular RF communications 2 that small cells have significantly contributed to Q = –10 dB recent improvements in network spectral effi- ciency. However, the main limiting factor is 0 0 100 200 300 400 500 interference. Optical attocells are less subject to Number of primary receivers interference since lightwaves do not propagate through walls. The ratio of the area spectral Figure 3. The average system spectral efficiency of a CR network as a function efficiency (ASE) in bits per second per Hertz of the number of primary receivers with different values of interference thresh- per square meter attained for the attocell net- olds Q (number of secondary receivers = 20). work and the ASE for the femtocell network is illustrated in Fig. 5 against a varying number of femtocells per floor. The number of optical access points per room varies from one to four. 6 Direct transmission, MAX SINR The gains diminish as the number of femtocells Single transmission, PF per floor is increased as expected, but the gain Non-orthogonal, MAX SINR is still above 100 for 20 femtocells per floor and 5.5 Non-orthogonal, PF Orthogonal, MAX SINR 4 optical attocells per room. The maximum gain Average spectral efficiency (b/s/Hz) Orthogonal, PF in ASE is close to 1000. As an example, let us 5 assume a typical ASE of 1.2 b/s/Hz/m 2 for the optical attocell network and a bandwidth of 10 4.5 MHz for LED and RF. This would mean that users can on average share a total of 300 Mb/s 4 in a 5 m × 5 m × 3 m room in the case of the optical attocell network. In the case of the RF 3.5 femtocell network and the best case of 20 fem- tocells per floor, the attainable capacity for the 3 same room is only about 3 Mb/s. GREEN COMMUNICATIONS 2.5 The design of 5G wireless systems should take 2 into account minimizing the energy consump- 0% 20% 40% 60% 80% 100% tion in order to achieve greener wireless com- Percentage of users within MFemtocells munication systems. Wireless system operators around the world should aim to Figure 4. Average spectral efficiency of system-level MFemtocells with multi- achieve such energy consumption reductions, user scheduling and resource partitioning schemes. which consequently contribute to the reduction of CO 2 emissions. The indoor communication technologies are promising deployment strate- FUTURE CHALLENGES IN gies to get better energy efficiency. This is because of the favorable channel conditions 5G WIRELESS COMMUNICATION they can offer between the transmitters and NETWORKS receivers. Moreover, by separating indoor traffic from outdoor traffic, the marcocell BS will have Although there have been some developments in less pressure in allocating radio resources and the above potential key 5G wireless technolo- can transmit with low power, resulting in a sig- gies, there are still many challenges ahead. Due nificant reduction in energy consumption. VLC to the limited space, in this section we only dis- and mm-wave technologies can also be consid- cuss some of these challenges. ered as energy efficient wireless communication solutions to be deployed in 5G wireless systems. OPTIMIZING PERFORMANCE METRICS For example, in VLC systems the consumed The evaluation of wireless communication net- energy in one bulb is much less than that in its works has been commonly characterized by con- RF-based equivalents for transmitting the same sidering only one or two performance metrics high-density data. while neglecting other metrics due to high com- IEEE Communications Magazine February 2014 127 WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 128 REDUCING SIGNAL PROCESSING COMPLEXITY 1000 FOR MASSIVE MIMO 1 AP 900 1 AP, FL = 17 dB 2 AP One technical challenge in developing massive 2 AP, FL = 17 dB MIMO systems is the signal processing complex- 800 4 AP 4 AP, FL = 17 dB ity. As transmit and receive signals are quite 700 lengthy, the search algorithms must be per- formed over many possible permutations of sym- ASE VLC/ASE RF 600 bols. In the current literature, massive MIMO research is often treated as a detection problem 500 based on a search motivated by the well-known ML criterion. The existing detection algorithms 400 assume that the channel has been perfectly esti- mated, which appears to be an unreasonable 300 assumption given the size of the channel matrix 200 and thus amount of channels to be tracked. A possible solution to this problem is to apply the 100 SM concept to massive MIMO systems. In this case, the spatial signature of each antenna needs 0 to be different from the point of view of the 4 6 8 10 12 14 16 18 20 Number of femtocells per floor receiver because data is encoded into the choice of transmit antenna active in the transmit array. It is therefore possible that channel estimation Figure 5. The ratio of ASE attained for the optical attocell network to ASE for does not need to be exact but rather be merely the femtocell network against varying numbers of femtocells per floor. sufficient to distinguish each transmit antenna. This may be a reasonable prospect, especially if the receive array is large, in which case each plexity. For a complete and fair assessment of transmit antenna would have a quite detailed 5G wireless systems, more performance metrics and thus distinct spatial signature. should be considered. These include spectral efficiency, energy efficiency, delay, reliability, INTERFERENCE MANAGEMENT FOR fairness of users, QoS, implementation complexi- CR NETWORKS ty, and so on. Thus, a general framework should be developed to evaluate the performance of 5G A major issue in interference-tolerant CR net- wireless systems, taking into account as many works in 5G is how to reliably and practically performance metrics as possible from different manage the mutual interference of CR and pri- perspectives. There should be a trade-off among mary systems. Regulating the transmit power is all performance metrics. This requires high-com- essential for the CR system to coexist with other plexity joint optimization algorithms and long licensed systems. An interference temperature simulation times. model is introduced for this purpose to charac- terize the interference from the CR to the REALISTIC CHANNEL MODELS FOR licensed networks. Interference cancellation 5G WIRELESS SYSTEMS techniques should also be applied to mitigate the interference at CR receivers. Another issue in Realistic channel models with proper accura- interference-tolerant CR networks is that a feed- cy-complexity trade-off are indispensable for back mechanism is important to periodically some typical 5G scenarios, such as massive inform the CR network about the current inter- MIMO channels and high-mobility channels ference status at the licensed system. A practical (e.g., high-speed train channels and vehicle-to- solution is that the interference state informa- vehicle channels). Conventional MIMO chan- tion can be sent from licensed systems and col- nel models cannot be directly applied to lected by a central unit (or a third party system). massive MIMO channels in which different Any CR network should first register to the cen- antennas may observe different sets of clus- tral unit in order to be updated regarding the ters. Massive MIMO channel models should allowed spectrum and interference. Alternative- take into account specific characteristics that ly, the CR transmitters can listen to beacon sig- make them different from those in convention- nals transmitted from the primary receivers and al MIMO channels, such as the spherical wave- rely on the channel reciprocity to estimate the front assumption and non-stationary channel coefficient. In this case, the CR trans- properties. Also, 3D massive MIMO models, mitters can cooperate among themselves to reg- which jointly consider azimuth and elevation ulate the transmit power and prevent the angles, are more practical but more complicat- interference at the primary receivers being above ed. Some existing massive MIMO channel the threshold. models are briefly summarized and classified in Table 1. Compared to conventional low-mobility wire- CONCLUSIONS less channels, high-mobility channels have In this article, the performance requirements of greater dynamics and possibly more severe fad- 5G wireless communication systems have been ing, and are essentially non-stationary. How to defined in terms of capacity, spectral efficiency, characterize non-stationary high-mobility chan- energy efficiency, data rate, and cell average nels is also very challenging. throughput. A new heterogeneous 5G cellular 128 IEEE Communications Magazine February 2014 WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 129 architecture has been proposed with separated indoor and outdoor applications using DAS and Ready for Channel model Complexity Description massive massive MIMO technology. Some short-range MIMO? communication technologies, such as WiFi, fem- tocell, VLC, and mm-wave communication tech- nologies, can be seen as promising candidates to Narrowband i.i.d. Low Uncorrelated model No Rayleigh provide high-quality and high-data-rate services to indoor users while at the same time reducing the pressure on outdoor BSs. We have also dis- Narrowband CBSM Medium Jointly correlated model No cussed some potential key technologies that can (Weichselberger) be deployed in 5G wireless systems to satisfy the expected performance requirements, such as CR Narrowband CBSM Medium Classic correlated model No networks, SM, MFemtocells, VLC, and green (Kronecker) communications, along with some technical chal- lenges. Wideband elliptical Massive MIMO High Yes GBSM properties considered ACKNOWLEDGMENT The authors gratefully acknowledge the support Table 1. Recent advances in massive MIMO channel models. of this work from SNCS Research Center, the University of Tabuk, the Ministry of Higher nication networks. He has served or is serving as an Editor Education in Saudi Arabia, the Opening Project or Guest Editor for 11 international journals, including IEEE of the Key Laboratory of Cognitive Radio and Transactions on Vehicular Technology (2011–), IEEE Trans- Information Processing (Guilin University of actions on Wireless Communications (2007–2009), and IEEE Journal on Selected Areas in Communications. He has edit- Electronic Technology), Ministry of Education ed one book, and published one book chapter and about (No. 2013KF01), Hutchison 3G UK, Natural Sci- 190 papers in journals and conferences. ence Foundation of China (No. 61231009), the STCSM (No. 11JC1412300), the International F OURAT H AIDER ([email protected]) received his Bachelor’s degree in electrical and electronic engineering/communica- Science & Technology Cooperation Program. tion engineering from the University of Technology, Iraq, in 2004, and his M.Sc. degree with Distinction at Brunel Uni- REFERENCES versity, United Kingdom, in 2009. Since December 2010, he has been a Ph.D. student at Heriot-Watt University and the Commission of the European Communities, Staff Work- University of Edinburgh, United Kingdom. His main research ing Document, “Exploiting the Employment Potential of interests include spectral-energy efficiency trade-off, wire- ICTs,” Apr. 2012. less channel capacity analysis, femtocell and mobile femto- Euro. Mobile Industry Observatory, GSMA, Nov. 2011. cell, and MIMO systems. He received the Best Paper Award A. Hashimoto, H. Yorshino, and H. Atarashi, “Roadmap from IEEE ICCT 2011. of IMT-Advanced Development,” IEEE Microwave Mag., vol. 9, no. 4, Aug. 2008, pp. 80–88. X IQI G AO [SM’07] ([email protected]) received his Ph.D. WWRF, L. Sorensen and K. E. Skouby, User Scenarios degree in electrical engineering from Southeast University, 2020, report, July 2009; http://www.wireless-world- Nanjing, China, in 1997. He joined the Department of research.org. Radio Engineering, Southeast University, in April 1992. C. Han et al., “Green Radio: Radio Techniques to Enable Since May 2001, he has been a professor of information Energy Efficient Wireless Networks,” IEEE Commun. systems and communications. From September 1999 to Mag., vol. 49, no. 6, June 2011, pp. 46–54. August 2000, he was a visiting scholar at Massachusetts Nokia Siemens Networks, “2020: Beyond 4G: Radio Evo- Institute of Technology and Boston University. From August lution for the Gigabit Experience,” white paper, 2011. 2007 to July 2008, he visited the Darmstadt University of A. Bleicher, “Millimeter Waves May Be the Future of 5G Technology as a Humboldt scholar. His current research Phones,” IEEE Spectrum, Aug. 2013. interests include broadband multicarrier communications, V. Chandrasekhar, J. G. Andrews, and A. Gatherer, MIMO wireless communications, and signal processing for “Femtocell Networks: A Survey,” IEEE Commun. Mag., wireless communications. vol. 46, no. 9, Sept. 2008, pp. 59–67. F. Rusek et al., “Scaling Up MIMO: Opportunities and XIAOHU YOU [SM’11, F’12] ([email protected]) received his Challenges with Very Large Arrays,” IEEE Sig. Proc. Master’s and Ph.D. degrees from Southeast University, Mag., vol. 30, no. 1, Jan. 2013, pp. 40–60. Nanjing, China, in electrical engineering in 1985 and 1988, H. Haas, “Wireless Data from Every Light Bulb,” TED respectively. Since 1990, he has been working with Nation- website, Aug. 2011; http://bit.ly/tedvlc al Mobile Communications Research Laboratory at South- X. Hong et al., “Secondary Spectrum Access Networks: east University, where he held the rank of professor. His Recent Developments on the Spatial Models,” IEEE research interests include mobile communication systems, Vehic. Tech. Mag., vol. 4, no. 2, June 2009, pp. 36–43. and signal processing and its applications. Now he is a F. Haider et al., “Spectral Efficiency Analysis of Mobile member of the China 863 Expert Committee responsible Femtocell Based Cellular Systems,” Proc. IEEE ICCT ’11, for telecommunications. Jinan, China, Sept. 2011, pp. 347–51. M. D. Renzo et al., “Spatial Modulation for Generalized Y ANG Y ANG [S’99, M’02, SM’10] ([email protected]) MIMO: Challenges, Opportunities, and Implementation,” received his Ph.D. degree in information engineering from Proc. IEEE, vol. 102, no. 1, Jan. 2014, pp. 56–103. the Chinese University of Hong Kong in 2002. He is cur- C.-X. Wang and S. Wu, “Massive MIMO Channel Mea- rently the director of the Shanghai Research Center for surements and Modeling: Advances and Challenges” Wireless Communications (WiCO) and a professor with the IEEE Wireless Commun.., submitted for publication. School of Information Science and Technology, Shang- X. Hong et al., “Capacity Analysis of Hybrid Cognitive haiTech University. His general research interests include Radio Networks with Distributed VAAs,” IEEE Trans. wireless ad hoc and sensor networks, wireless mesh net- Vehic. Tech., vol. 59, no. 7, Sept. 2010, pp. 3510–23. works, next generation mobile cellular systems, intelligent transport systems, and wireless testbed development and BIOGRAPHIES practical experiments. C H E N G -X I A N G W A N G [S’01, M’05, SM’08] (cheng- DONGFENG YUAN [SM’01] ([email protected]) received his [email protected]) received his Ph.D. degree from Aal- M.Sc. degree from the Department of Electrical Engineer- borg University, Denmark, in 2004. He joined Heriot-Watt ing, Shandong University, China, in 1988, and his Ph.D. University, United Kingdom, as a lecturer in 2005 and degree from the Department of Electrical Engineering, became a professor in August 2011. His research interests Tsinghua University, China, in January 2000. Currently he is include wireless channel modeling and 5G wireless commu- a full professor in the School of Information Science and IEEE Communications Magazine February 2014 129 WANG_LAYOUT.qxp_Layout 1/30/14 1:29 PM Page 130 Engineering, Shandong University, China. His current wireless networks, spatial modulation, and optical wireless research interests include cognitive radio systems, coopera- communication. He holds more than 15 patents. He has tive (relay) communications, and 5G wireless communica- published more than 50 journal papers, including a Science tions. article, and more than 140 peer-reviewed conference papers. HADI M. AGGOUNE [M’00, SM’09] (haggoune.sncs@ ut.edu.sa) received his Ph.D. degree in electrical engi- SIMON FLETCHER ([email protected]) has respon- neering from the University of Washington (UW). He is a sibility for Products Strategy and Innovation for the emerg- Professional Engineer registered in the State of Wash- ing communications infrastructure platforms of NEC’s ington, winner of the IEEE Professor of the Year Award, Global Market product portfolio. He has core interests in UW Branch, and listed as an inventor in a patent sustainability, 3GPP radio access technologies, and the assigned to Boeing. Currently he is a professor and management of innovation processes. As a director of director of the Sensor Networks and Cellular Systems mobile VCE, a UK based research consortium, he acts as Research Center, University of Tabuk, Saudi Arabia. His CEO and Chair of the Advisory Board. He also represents research interests include modeling and simulation of the community as a director in the ICT-KTN. large-scale networks, sensors, visualization, and control and energy. EROL HEPSAYDIR ([email protected]) has been work- ing in the cellular mobile industry for 22 years. He has HARALD HAAS [S’98, AM’00, M’03] ([email protected]) holds launched several mobile networks in Australia, Singapore, the Chair of Mobile Communications in the Institute for Korea, and the United Kingdom. Since 2001, he has been Digital Communications at the University of Edinburgh. His responsible for network technology. He is now head of main research interests are in the areas of wireless system Network Technology for Hutchison 3G UK. He is currently design and analysis as well as digital signal processing, working on LTE and small cell deployment planning in the with a particular focus on interference coordination in United Kingdom. 130 IEEE Communications Magazine February 2014