Radio Network Planning Towards 5G Mmwave Standalone Small-Cell Architectures

electronics

Article Radio Network Planning towards 5G mmWave Standalone Small-Cell Architectures

Georgia E. Athanasiadou 1,*, Panagiotis Fytampanis 1 , Dimitra A. Zarbouti 1 , George V. Tsoulos 1, Panagiotis K. Gkonis 2 and Dimitra I. Kaklamani 3

1 Wireless & Mobile Communications Lab, Department of Informatics and Telecommunications, University of Peloponnese, 22131 Tripolis, Greece; [email protected] (P.F.); [email protected] (D.A.Z.); [email protected] (G.V.T.) 2 General Department, National and Kapodistrian University of Athens, Sterea Ellada, 34400 Dirﬁes Messapies, Greece; [email protected] 3 Intelligent Communications and Broadband Networks Laboratory, School of Electrical and Computer Engineering, National Technical University of Athens, 9 Heroon Polytechneiou str, Zografou, 15780 Athens, Greece; [email protected] * Correspondence: [email protected]

Received: 30 December 2019; Accepted: 13 February 2020; Published: 16 February 2020

Abstract: The 5G radio networks have introduced major changes in terms of service requirements and bandwidth allocation compared to cellular networks to date and hence, they have made the fundamental radio planning problem even more complex. In this work, the focus is on providing a generic analysis for this problem with the help of a proper multi-objective optimization algorithm that considers the main constraints of coverage, capacity and cost for high-capacity scenarios that range from dense to ultra-dense mmWave 5G standalone small-cell network deployments. The results produced based on the above analysis demonstrate that the denser the small-cell deployment, the higher the area throughput, and that a sectored microcell conﬁguration can double the throughput for ultra-dense networks compared to dense networks. Furthermore, dense 5G networks can actually have cell radii below 400 m and down to 120 m for the ultra-dense sectored network that also reached spectral eﬃciency 9.5 bps/Hz/Km2 with no MIMO or beamforming.

Keywords: radio network planning; 5G wireless communications systems; mmWave small cells

1. Introduction Radio planning is an essential task for wireless networks that mainly refers to calculating the number, location and configuration of the radio network nodes. In the early days, since no prior network infrastructure existed, this task considered only the estimation of the number of base stations (BSs) and their locations, i.e., the BS location problem [1–4]. In addition, for early cellular networks, radio network planning was split into two separate tasks: coverage and capacity. Nevertheless, as it was shown in several publications (e.g., [5–7]), capacity and coverage planning are not un-correlated. On the contrary, they are inter-related in 3G, 4G and now 5G wireless networks and hence, they must be treated together. The 5G networks introduce really different elements from the previous generations, mainly due to virtualization and service-based architecture. Among other things, they are designed for considerably higher data rates, very large numbers of connected Internet of Things (IoT) devices and low latency while providing adaptive means for network scalability and flexibility. The number of the 5G radio frequency bands is targeted to be higher than in previous generations of cellular networks, more specifically multiple mmWave bands [8]. Also, massive Multiple Input Multiple Output (MIMO) and

Electronics 2020, 9, 339; doi:10.3390/electronics9020339 www.mdpi.com/journal/electronics Electronics 2020, 9, 339 2 of 10 hybrid beamforming are core techniques to achieve the targeted high data rates and the large number of devices [9]. Typically the cell deployment architectures can be classified into standalone and overlay architectures. In the context of 5G, the first refers to a network deployment that consists of mere mmWave small cells, while the latter refers to the deployment of mmWave small cells on top of the existing macro-relay networks in the form of hierarchical or mixed cell structures. In the overlay architecture, the existing (pre5G) macrocell layer is mainly for coverage as well as mobility and signaling problems originating from the mmWave small-cell layer, which exists for capacity boosts. The much wider bandwidth as well as the beamforming/MIMO capabilities, together with the reduced access-link distances, give the mmWave small cells the capability to substantially increase the system capacity. It has to be mentioned here that another advantage of the overlay network architecture is the separation of control and message transmissions. All the control signaling is supported via the existing macrocells and high data rate transmissions go through the mmWave small-cell network. 5G radio planning is now of utmost importance since not only does it require cost-optimized deployments capable of handling a variety of demand constraints, but also since it then affects the optimal placement of the core network elements, e.g., for achieving low latency values. In order to tackle these issues, this article presents a generic analysis that allows for evolutionary radio network planning towards 5G mmWave standalone small-cell architectures. Due to the complex nature of the problem, a multi-objective optimization algorithm is appropriately modified in order to provide solutions in the context of the multiple constraints and specifics of a 5G mmWave network with particular emphasis in the 28 GHz band. The three main constraints of coverage, capacity and cost are studied for a range of cases that reflect scenarios from dense to ultra-dense network deployments in order to achieve high capacity. The optimization algorithm used in this paper is the evolution of the algorithm previously used for the analysis of Long-Term Evolution (LTE) systems [7]. In this study, the analysis is focused on network dimensioning for the mmWave band at 28 GHz. Hence, the propagation module is modified accordingly and a path loss model based on results of field trials at the 28 GHz band is applied. For this analysis only microcells are considered with tri-sector directional antennas, as well as omnidirectional antenna patterns. Moreover, the system module is upgraded to encompass 5G rate calculations, while the core analysis is focused on the impact of high-density microcell deployments on the spectral efficiency and throughput efficiency of these systems. The remainder of the paper is organized as follows: Section2 starts with the formulation of the radio network planning optimization problem and then explains the developed simulation methodology in terms of the employed propagation model, the necessary 5G system characteristics and the planning optimization method. Section3 presents the results for di fferent 5G standalone small-cell network deployments, namely dense, very dense and ultra-dense. Finally, concluding remarks are provided in Section4.

2. Problem Formulation and Simulation Methodology For the study of dense networks at high frequencies, two types of serving radio network nodes were considered in the planning process of our analysis: omnidirectional and tri-sectored microcells. When sectorized scenarios were considered, at each site three gNBs were co-located (reuse pattern 1 3 1). During the planning process, all three sectors or only a subset of them could be activated, × × according to the required signal strength in an area and/or the need to limit the overall interference of the system. Although the algorithm was developed to also examine tri-sectored macrocells with relay nodes, they were not considered here since the focus was to analyze the capacity capabilities of ultra-dense standalone small-cell architectures. Note that each cell can also have MIMO capabilities. Control Nodes (CNs) were used in our study in order to represent throughput requirements as well as coverage constraints. Speciﬁcally, K CNs were distributed in the area under study and each CN Electronics 2020, 9, 339 3 of 10

was associated with a rate requirement, i.e., RCN,min. Introduced by system link budget calculations, the received power at each CN should always exceed a minimum power threshold. Furthermore,Electronics 2020, the9, x FOR optimization PEER REVIEW process was employed on discrete locations3 rather of 11 than on continuous x-yFurthermore, space, anthe approachoptimization already process appliedwas employed and on tested discre forte locations previous rather generation than on network deploymentscontinuous [2,5,6 ].x-y This space, way an theapproach complexity already ofapplied the NP-hard and tested problem for previous of continuous generation network space calculations was avoideddeployments [10], but [2,5,6]. also, This this way was the closer complexity to the of actualthe NP-hard planning problem process of continuous employed space by network operatorscalculations where existing was avoided sites [10], are mostbut also, certainly this was reusedcloser to inthe anyactual future planning network process layout.employed by network operators where existing sites are most certainly reused in any future network layout. The outlineThe outline of the of radio the radio network network planning planning proc processess is shown is shown in Figure in Figure 1. A more1. Adetailed more detailed descriptiondescription is provided is provided in the in nextthe next sections. sections.

FigureFigure 1. 1.Radio Radio network planning planning outline. outline.

2.1. Scenario:2.1. Scenario: Network Network Deployment Deployment Requirements Requirements Herein, our planning process was employed in a rectangular area with dimensions 1 Km × 1 Km. Herein, our planning process was employed in a rectangular area with dimensions 1 Km 1 Km. Due to the area characteristics, the requirements/constraints of the planning process were defined × Due to theagainst area a characteristics,set of 25 equi-spaced the CNs requirements (separation distance/constraints of 200 m). ofThis the grouping planning of CNs process represented were defined against athe set high-capacity of 25 equi-spaced requirements CNs (separationin a small area, distance i.e., a demanding of 200 m). network Thisgrouping planning operational of CNs represented the high-capacityscenario that requirements requires a dense in anetwork small area,deployment. i.e., a demanding network planning operational scenario Since the trade-off between performance and network density was our main concern, a clear that requirespreexisting a dense network network layout deployment. was considered and new transmitting sites were explored, i.e., microcells Sincewith the various trade-o densityff between deployments performance were examined. and Scen networkarios with density 30 (5 × 6 was grid), our 100 main(10 × 10 concern, grid) a clear preexistingand network 400 (20 × layout20 grid) wascandidate considered microcells and uniformly new transmitting distributed in sitesthe area were were explored, studied in i.e.,the microcells with variousanalysis density below deployments in order to investigate were examined. the capabilities Scenarios of a dense with network. 30 (5 However,6 grid), 100it should (10 be10 grid) and pointed out that the locations of the candidate network sites were an exte×rnal parameter to ×our 400 (20 20 grid) candidate microcells uniformly distributed in the area were studied in the analysis ×platform, thus could be defined by any interested stakeholder. below in order to investigate the capabilities of a dense network. However, it should be pointed out that the locations2.2. The Planning of the Method candidate network sites were an external parameter to our platform, thus could be definedThe planning by any analysis interested consisted stakeholder. of three pillars: 1. Propagation analysis 2.2. The Planning2. System Method analysis 3. Planning optimization The planning analysis consisted of three pillars:

1. Propagation analysis 2. System analysis 3. Planning optimization Electronics 2020, 9, 339 4 of 10

Electronics 2020, 9, x FOR PEER REVIEW 4 of 11 2.2.1. Propagation Analysis 2.2.1. Propagation Analysis Various studies have been conducted in projects like METIS, MiWEBA, ITU-R M, mmMAGIC etc. Various studies have been conducted in projects like METIS, MiWEBA, ITU-R M, mmMAGIC in order to deploy channel models in higher frequency bands, such as millimeter waves, to support etc. in order to deploy channel models in higher frequency bands, such as millimeter waves, to 5G use cases [8,11–13]. The path loss model that was adopted here was the Alpha-Beta-Gama (ABG) support 5G use cases [8,11–13]. The path loss model that was adopted here was the Alpha-Beta-Gama empirical path loss model: (ABG) empirical path loss model:

PL𝑃𝐿(dB(𝑑𝐵) =) =10 10a log𝑎 𝑙𝑜𝑔10(d)(𝑑+)β+𝛽+10𝛾𝑙𝑜𝑔+ 10γlog10( fGHz(𝑓) ) (1)(1)

TheThe modelmodel hashas threethree parametersparameters toto describedescribe meanmean pathpath lossloss overover frequencyfrequency andand distance,distance, whichwhich derivederive fromfrom the the statistical statistical process process of measurements.of measuremen Ints. ourIn our analysis, analysis, at urban at urban environments environments at 28 GHz,at 28 theseGHz, parametersthese parameters took the took following the following values values [13]: [13]: 𝑎 = 3.5, 𝛽 = 24.4, 𝛾 =1.9 (2) a = 3.5, β = 24.4, γ = 1.9 (2) The propagation losses of this model vs. distance for f = 28 GHz are depicted in Figure 2 along withThe the propagationlosses of the WINNER losses of this model model for vs.microcells distance at for thef 2= GHz28 GHz band are of depicted LTE [14]. in Figure2 along with the losses of the WINNER model for microcells at the 2 GHz band of LTE [14].

FigureFigure 2.2. PathPath lossloss vs.vs. distancedistance atat 2828 GHzGHz andand 22 GHzGHz bands.bands.

TableTable1 1summarizes summarizes typical typical values values used used in in the the performed performed simulations simulations for for generic generic mmWave mmWave operationaloperational scenarios.scenarios. We considered the downlink of a 5G network since it is usually the bottleneck for such a system. The Signal to Interference plus NoiseTable Ratio 1. (SINR) Simulation at the parameters.kth CN is:

Simulation ParameterPb AbA b Gch Value tx 0 φ k b,k SINRCentralk = P Frequency → 28 GHz (3) Pi Ai Ai Gch + P Transmitter EIRP andi UserB b Antennatx 0 φ kGaini,k N 60 dBm ∈{ \ } → Sector Antenna Gain (3GPP pattern for 120° sector) 18 dBi b k Pb Ab where is the serving gNB forOmnidirectional the th CN, tx Antennaand 0 are Gain the transmitted power 4 dBi and maximum antenna gain for the bth gNB, respectively, while Ab is the normalized antenna gain of the bth gNB towards Total cableφ k losses 2 dB → k Gch Ai the th CN and b,k is the channelChannel gain between Bandwidth them. Similarly, for the interference 100 MHz calculation, φ k → is the normalized antennaAggregated gain of the Bandwidthith interfering (16 Carriers) gNB towards the served 1600 MHz CN. Lastly, PN is the noise power at the receiver. TheResource micro gNBs Blocks employ (RBs) either omnidirectional 132 antennas (Ab = 1) or b = ( ( )2 ) directional 120◦ sector antennasMinimum [15] Received with A0 = Power18 dBi 𝑃 (see, Table 1) and A −5 dBmmin 12 φ/φ3dB , 25 , o − where φ3dB = 70 . It has to be noted here that the Eﬀective Isotropic Radiated Power (EIRP) of the transmittingWe considered BS antennas the downlink remains theof a same 5G network with sector since and it is omnidirectional usually the bottleneck antennas, for i.e.,such 60 a dBm.system. The Signal to Interference plus Noise Ratio (SINR) at the kth CN is:

Electronics 2020, 9, 339 5 of 10 Electronics 2020, 9, x FOR PEER REVIEW 5 of 11

Table 1. Simulation parameters. 𝑃𝐴𝐴→𝐺, 𝑆𝐼𝑁𝑅 = (3) Simulation Parameter Value ∑∈{\} 𝑃𝐴 𝐴→𝐺, +𝑃

Central Frequency 28 GHz where b is the serving gNB for the kth CN, 𝑃 and 𝐴 are the transmitted power and maximum Transmitter EIRP and User Antenna Gain 60 dBm antenna gain for the bth gNB, respectively, while 𝐴⟶ is the normalized antenna gain of the bth gNB towards theSector kth CN Antenna and 𝐺 Gain, is (3GPP the channel patternfor gain 120 between◦ sector) them. Similarly, 18 dBi for the interference calculation, 𝐴⟶ is the normalizedOmnidirectional antenna Antenna gain Gain of the ith interfering 4gNB dBi towards the served CN. Lastly, PN is the noise power at theTotal receiver. cable losses The micro gNBs employ either 2 dB omnidirectional antennas 𝐴 =1 0 𝐴 = ( ) or directional 120° Channelsector Bandwidthantennas [15] with A = 18 100dBi MHz (see Table 1) and −min(12(𝜙𝜙⁄) , 25), where 𝜙 =70 . It has to be noted here that the Effective Isotropic Aggregated Bandwidth (16 Carriers) 1600 MHz Radiated Power (EIRP) of the transmitting BS antennas remains the same with sector and omnidirectional antennas, i.e.,Resource 60 dBm. Blocks (RBs) 132 Minimum Received Power P 5 dBm rx, min − 2.2.2. 5G System Module 2.3. 5GThe System rate Modulerequirements set at each CN are to be met by 5G system options [16–18], hence, a generic 5G Therate ratecalculation requirements module set was at each used CN in are order to be to met associate by 5G systemSINR calculations options [16– 18at ],each hence, CN a genericwith the 5Gmaximum rate calculation throughput module offered was by used the insystem. order As to prev associateiously SINR mentioned, calculations the system at each module CN with thatthe was maximumoriginally throughputdesigned upon offered the byLTE-A the system. system’s As spec previouslyifications mentioned, [7] is now theextended system to module consider that 5G wassystem originally options designed in FR2 bands, upon thei.e., LTE-An257 in system’s the mmWave specifications range around [7] is the now 28 extended GHz. The to system consider module 5G systemconsiders options a wideband in FR2 bands, channel i.e., of n257 100 in MHz the mmWave with subcarrier range around spacing the of 28 60 GHz. KHz, The a combination system module that considersleads to 132 a wideband resource channel blocks [19]. of 100 The MHz carrier with aggr subcarrieregation spacing technique of 60was KHz, also a considered combination with that 16 leads New toRadio 132 resource (NR) carriers blocks [20], [19]. the The maximum carrier aggregation allowable techniqueby the 5G NR was Radio also considered Access Technology with 16 New (RAT). Radio The (NR)total carriers bandwidth [20], theincreased maximum to 1.6 allowable GHz, while by the the 5G total NR Radionumber Access of available Technology RBs (RAT).was 2112 The (16 total NR bandwidthcarriers × 132 increased RBs per to carrier). 1.6 GHz, The while module the total used number the measured of available SINR RBs at wasthe CN 2112 side (16 NRto estimate carriers the × 132required RBs per number carrier). of TheRBs moduleto offer the used capacity the measured requirement. SINR at the CN side to estimate the required numberFigure of RBs 3 toprovides offer the an capacity insight requirement.on the rate offered from a 5G system as configured herein, when comparedFigure3 to provides a typical an LTE-A insight system on the with rate 20 o MHzffered of frombandwidth a 5G system and five as component configured carriers herein, leading when comparedto 100 MHz to a typicalof aggregated LTE-A system bandwidth. with 20 MHzBoth ofsystems bandwidth were and considered five component without carriers MIMO leading spatial- to 100multiplexing MHz of aggregated schemes bandwidth. since an eight-layer Both systems transmission were considered can be supported without MIMO by both. spatial-multiplexing schemes since an eight-layer transmission can be supported by both.

FigureFigure 3. 3.Rate Rate calculationscalculations forfor anan LTE-ALTE-A system system with with 20 20 MHz MHz bandwidth bandwidth and and ﬁve five aggregated aggregated componentcomponent carriers carriers compared compared to to the the 5G 5G NR NR system system employed employed in in this this work. work. 2.3.1. The Planning Optimization Algorithm 2.2.3. The Planning Optimization Algorithm Our goal herein was to propose a network layout of optimally located microcells that keep the Our goal herein was to propose a network layout of optimally located microcells that keep the deployment cost to a minimum while at the same time speciﬁc coverage and capacity criteria are met. deployment cost to a minimum while at the same time specific coverage and capacity criteria are met. Since in this analysis we only considered microcells, the deployment cost of each node was the same,

Electronics 2020, 9, 339 6 of 10

Since in this analysis we only considered microcells, the deployment cost of each node was the same, C, and hence, the total cost of a network deployment was directly proportional to the number of nodes deployed. Equations (4)–(7) are formulating this optimization problem.    X  OPT argmin   Bm  Cb (4) ← B m B   b B 0 ∈ ∈ m0 Subject to: Pk P , k K (5) rx ≥ rx,min ∀ ∈ X N RBmax, g G (6) RB,k ≤ ∀ ∈ 0 k Kg ∈ R R , k K (7) k ≥ CN,min ∀ ∈ where Bm0 is the set of candidate microsites that are under examination, Cb is the cost of the b microsite k (herein, Cb = C), K is the fixed set of CNs defined in Section 2.1, Prx is the received power of the k CN, G is the set gNBs associated with the B sites (as previously mentioned, when sectors are considered, 0 m0 each site comprises three gNBs), Kg is the set of CNs currently served by g gNB, RBmax is the maximum allowed resource blocks of a gNB, Rk is the rate offered by the system to the k CN and finally RCN,min is the capacity requirement. Note that the different candidate solutions that were examined (B) were derived in the context of the Genetic Algorithm (GA) methodology as it is described in the following. OPT The solution to the optimization problem defined earlier was the Bm set of microsites that entails the minimum total cost [7] of network deployment while at the same time meets the constraints set by Equations (5)–(7). Herein, the optimization problem was tackled with a multi-objective evolutionary algorithm [10] and specifically with the NSGA-II (Non-sorting Dominated Genetic Algorithm-II). For this, the problem was analyzed and translated into a three-dimensional fitness function, F = F(1), F(2), F(3) . Each dimension corresponded to an objective that needed to be minimized. Equations (8)–(10) define the fitness function to be minimized.

F(1) = Bm0 (8)

n o F(2) = K0 , K0 = k K N 0 (9) ∈ RB,k →   X  X    F(3) =  NRB,k RBmax (10)  −  g G k Kg  ∈ 0 ∈ The ﬁrst objective, F(1), corresponds to the weighted network cost, the second, F(2), is the total number of CNs that do not meet their requirements, either capacity or coverage, while F(3) is the excess RBs needed at a network level. The NSGA-II algorithm imitates evolutionary mechanisms and employs biological operations found in nature for generating optimal solutions. It starts with a random selection of an initial population (sets of microsites) and by iteratively using operators such as selection, evaluation and crossover or mutation evolves into optimal solutions. When NSGA-II reaches an optimal pareto front, i.e., it cannot be further improved (average pareto spread and distance criteria were used [6]), then a series of solutions corresponding to that pareto front is derived.

3. Radio Network Planning Results Uniformly distributed CNs in the service area represented the area coverage and capacity criterion. An initial throughput requirement was allocated at each CN, which gradually increased. At relatively low capacity requirements, microcells with omnidirectional antenna patterns were uniformly placed Electronics 2020, 9, 339 7 of 10 Electronics 2020, 9, x FOR PEER REVIEW 7 of 11

Electronicstoby the the relatively optimization 2020, 9, x FORhigh PEER algorithm, frequency, REVIEW as the seen SINR inFigure did not4a, fall in orderas rapidly for each duemicrocell to the equivalent to serve allreduction neighbor7 of of11 theCNs interference (shown in Figure from 4othera with cells. black stars). The SINR distribution (Figure4b) also followed this regular topattern. the relatively Note that high although frequency, the receivedthe SINR power did not fell fall sharply as rapidly with due distance to the due equivalent to the relatively reduction high of thefrequency, interference the SINR from didother not cells. fall as rapidly due to the equivalent reduction of the interference from other cells.

(a) (b)

Figure 4. (a) Received power (dBm) and (b) SINR distribution in the service area for a dense (a) (b) omnidirectional micro gNB deployment. Figure 4. (a(a) )Received Received power power (dBm) (dBm) and and ( (bb)) SINR SINR distribution distribution in in the the service service area area for for a a dense omnidirectionalMicrocells with micro omnidirectional gNB deployment. patterns were then replaced by tri-sectored microcells, with each sector being a separate gNB in order to offer increased spatially selective throughput. Note that the EIRPMicrocells of the transmitting withwith omnidirectionalomnidirectional BS antennas patterns patterns remained were were thenthe th sameen replaced replaced for sector by tri-sectoredby an tri-sectoredd omnidirectional microcells, microcells, withantennas with each each(60sector dBm), sector being while being a separate a a sectorseparate gNB could gNB in orderbe in switched order to oﬀ toer offeroff, increased if increased it was spatially not spatially needed, selective selective in order throughput. throughput. to keep Note interference Note that that the thelevelsEIRP EIRP oflow. the of Hence,the transmitting transmitting in the BSresults antennasBS antennas depicted remained remained in Figure the the same5, there same for was for sector sectora microcell and an omnidirectionald omnidirectional with only one antennas operating antennas (60 (60sector,dBm), dBm), whileand while another a sector a sector microcell could could be with switched be twoswitched transmitting oﬀ, ifoff, it wasif it se notwasctors. needed, not The needed, rest in orderof inthe order to microcells keep to interferencekeep were interference placed levels as levelsfarlow. apart Hence, low. as Hence,possible in the in results in the order results depicted for depictedall inthree Figure sectorsin Figure5, there to be5, was theretransmitting a microcellwas a microcell and with covering only with one aonly wider operating one area. operating sector, sector,and another and another microcell microcell with two with transmitting two transmitting sectors. se Thectors. rest The of the rest microcells of the microcells were placed were as placed far apart as faras possibleapart as inpossible order forin order all three for sectorsall three to sectors be transmitting to be transmitting and covering and acovering wider area. a wider area.

(a) (b)

Figure 5. (a) Received power (dBm) and (b) SINR distribution in the service area for a dense tri-sector Figure 5. (a) Received(a) power (dBm) and (b) SINR distribution in the service(b area) for a dense tri-sector micro gNB deployment. Figure 5. (a) Received power (dBm) and (b) SINR distribution in the service area for a dense tri-sector microAs the gNB throughput deployment. requirements increased, denser grids of microcells were deployed. As can be seen in Figure Figure 66a,a, forfor high-capacity requirementsrequirements withwith thethe ultra-denseultra-dense networks,networks, onlyonly microcellsmicrocells closeclose to the As CNs the werewerethroughput transmitting,transmitting, requirements with with only only increased, one one sector sector denser pointing pointing grids directly directly of microcells towards towards were the the corresponding deployed. corresponding As CN.can CN. Asbe seenAsa consequence, a inconsequence, Figure 6a, in for Figure in high-capacity Figure6b, the 6b, positionsthe requirements positions not close not wi cl tooseth a the sectorto ultra-densea sector antenna antenna pointingnetworks pointing directly, only directly microcells towards towards themclose tothemhad the very hadCNs low very were SINR low transmitting, dueSINR to increaseddue with to increased only interference one interfersector from pointingence the from dense directly the transmitting dense towards transmitting antennathe corresponding grid. antenna However, grid. CN. AsHowever, a consequence, due to the in optimizationFigure 6b, the algorithm, positions thenot anclosetenna to aposition sector antenna was such pointing that there directly was no towards CN in them had very low SINR due to increased interference from the dense transmitting antenna grid. However, due to the optimization algorithm, the antenna position was such that there was no CN in

ElectronicsElectronics 20202020,, 99,, x 339x FORFOR PEERPEER REVIEWREVIEW 88 8 ofof of 1111 10 aa positionposition ofof lowlow SINR.SINR. WithWith suchsuch optimaloptimal placemplacement,ent, thethe maximummaximum throughputthroughput forfor thethe 2525 CNsCNs inin 2 thethedue areaarea to the reachedreached optimization asas highhigh algorithm, asas almostalmost the1616 Gbps/KmGbps/Km antenna position2.. was such that there was no CN in a position of low SINR. With such optimal placement, the maximum throughput for the 25 CNs in the area reached as high as almost 16 Gbps/Km2.

((aa)) ((bb))

FigureFigure 6. 6. ((aa)) ReceivedReceived powerpower (dBm)(dBm) andand (( bb)) SINRSINR distributiondistribution inin thethe serviceservice areaarea for for an an ultra-dense ultra-dense tri-sectortri-sector micromicro gNB gNB deployment. deployment.

Overall,Overall, comparingcomparing thethethe performance performanceperformance of ofof the thethe various variouvariou networkss networknetwork conﬁgurations configurationsconfigurations that thatthat resulted resultedresulted from fromfrom the thetheoptimization optimizationoptimization algorithm, algorithm,algorithm, it was itit waswas evident evidentevident that thatthat the denserthethe denserdenser the microcellthethe microcellmicrocell deployment, deployment,deployment, the higher thethe higherhigher the area thethe areaareathroughput throughputthroughput (see Figure(see(see FigureFigure7). Moreover, 7).7). Moreover,Moreover, with tri-sector withwith trtri-sectori-sector microcells, microcells,microcells, the throughput thethe throughputthroughput almost almostalmost doubled doubleddoubled at the atatultra-dense thethe ultra-denseultra-dense networks networksnetworks with respect withwith respectrespect to that toto of thatthat dense ofof dense networks,dense networks,networks, either either witheither omnidirectional withwith omnidirectionalomnidirectional or sector oror sectorsectorantennas. antennas.antennas. At these AtAt scenarios,thesethese scenarios,scenarios, as seen asas in seenseen Figure inin FiguFigu6, mostrere 6,6, of mostmost the timesofof thethe each timestimes microcell eacheach microcellmicrocell had only hadhad one onlyonly sector oneone sectorsectorantenna antennaantenna operating operatingoperating and hence, andand hence, thehence, advantage thethe advantagadvantag with respectee withwith respect torespect the omnidirectional toto thethe omnidirectionalomnidirectional deployment deploymentdeployment was due waswasto interference duedue toto interferenceinterference reduction reductionreduction to adjacent toto adjacent cells.adjacent cells.cells.

Figure 7. Throughput per Km2 for different network densities with omnidirectional and tri-sector Figure 7. Throughput per Km 22 forfor different diﬀerent network densities with omnidirectional and tri-sector micro gNBs. micro gNBs.

FigureFigure 8 88provides providesprovides an insightanan insightinsight into howintointo dense howhow 5Gdensedense networks 5G5G withnetworksnetworks standalone withwith small-cellstandalonestandalone architectures small-cellsmall-cell architecturesarchitecturescan actually get. cancan actuallyactually Even for get.get. the EvenEven omnidirectional forfor thethe omnidireomnidire referencectionalctional scenario, referencereference the scenario,scenario, cell radius thethe cell didcell radius notradius get diddid higher notnot getgetthan higherhigher 400 m, thanthan while 400400 for m,m, the whilewhile ultra-dense forfor thethe ultra-denseultra-dense network the networknetwork cell radius thethe furthercellcell radiusradius decreased furtherfurther down decreaseddecreased to 120 downdown m. The toto 2 120120spectral m.m. TheThe eﬃ spectralciencyspectral achieved efficiencyefficiency now achievedachieved was 9.5 nownow bps /waswasHz/ Km 9.59.5 2bps/Hz/Kmbps/Hz/Km, approximately2,, approximatelyapproximately three times lowerthreethree timestimes than the lowerlower 5G thanthantarget thethe of 305G5G bps targettarget/Hz, ofof although 3030 bps/Hz,bps/Hz, no MIMO althoughalthough/beamforming nono MIMO/beamformingMIMO/beamforming was considered was inwas our consideredconsidered generic analysis. inin ourour Ifgenericgeneric this is analysis.analysis.also considered, IfIf thisthis isis then alsoalso it considered,considered, becomes obvious thenthen itit that becomesbecomes the spectral obviousobvious eﬃ thatthatciency thethe goal spectralspectral will be efficiencyefficiency easily exceeded. goalgoal willwill bebe easilyeasily exceeded.exceeded.

Electronics 2020, 9, 339 9 of 10 Electronics 2020, 9, x FOR PEER REVIEW 9 of 11

FigureFigure 8. 8. ThroughputThroughput (and (and spectral spectral efficiency efficiency per per Km Km2)2 )associated associated with with the the cell cell radius radius for for the the omnidirectionalomnidirectional and and sectorized sectorized scenarios scenarios and and the the de densense/very-dense/ultra-dense/very-dense/ultra-dense scenarios scenarios (from (from left left toto right). right). 4. Conclusions 4. Conclusions The ever-increasing demand for new wireless services, together with the evolution of cellular The ever-increasing demand for new wireless services, together with the evolution of cellular networks, has led to the 5G era. The new network generation introduces major changes and, as a networks, has led to the 5G era. The new network generation introduces major changes and, as a consequence, significant challenges in terms of radio planning. This work focused on providing a consequence, significant challenges in terms of radio planning. This work focused on providing a generic analysis for this problem, emphasizing the need for dense standalone small-cell networks in generic analysis for this problem, emphasizing the need for dense standalone small-cell networks in the mmWave band of 28 GHz, in order to achieve the capacity goals of 5G. The complex nature of the mmWave band of 28 GHz, in order to achieve the capacity goals of 5G. The complex nature of the the problem led us to employ a properly modified multi-objective optimization algorithm in order to problem led us to employ a properly modified multi-objective optimization algorithm in order to provide solutions in the context of the multiple constraints and specifics of a 5G mmWave network. The provide solutions in the context of the multiple constraints and specifics of a 5G mmWave network. three main constraints of coverage, capacity and cost were studied for a range of cases that reflected The three main constraints of coverage, capacity and cost were studied for a range of cases that scenarios from dense to ultra-dense network deployments in order to achieve high capacity. reflected scenarios from dense to ultra-dense network deployments in order to achieve high capacity. The produced results showed that the denser the microcell deployment, the higher the area The produced results showed that the denser the microcell deployment, the higher the area throughput and also that sectored microcells double the throughput for ultra-dense networks compared throughput and also that sectored microcells double the throughput for ultra-dense networks to dense networks since better interference control is achieved through the operation of one, instead compared to dense networks since better interference control is achieved through the operation of of two or three, sector antennas. In terms of how dense 5G networks with standalone small-cell one, instead of two or three, sector antennas. In terms of how dense 5G networks with standalone architectures can actually get, the analysis showed cell radii below 400 m that reached 120 m for small-cell architectures can actually get, the analysis showed cell radii below 400 m that reached 120 the ultra-dense sectored network. Finally, the spectral efficiency reached 9.5 bps/Hz/Km2 with no m for the ultra-dense sectored network. Finally, the spectral efficiency reached 9.5 bps/Hz/Km2 with MIMO/beamforming, which leads to the conclusion that the 5G spectral efficiency goal can be easily no MIMO/beamforming, which leads to the conclusion that the 5G spectral efficiency goal can be exceeded if these techniques are also exploited. easily exceeded if these techniques are also exploited.

AuthorAuthor Contributions: Contributions: Conceptualization, G.E.A.,G.E.A., D.A.Z.,D.A.Z., G.V.T., G.V.T., P.K.G. P.K.G. and and D.I.K.; D.I.K.; methodology, methodology, G.E.A., G.E.A., D.A.Z. D.A.Z.and P.F.; and validation, P.F.; validation, G.E.A., G.E.A., P.F. and P.F. D.A.Z.; and D.A.Z.; formal formal analysis, analysis, P.F.; investigation, P.F.; investigation, G.E.A., G.E.A., D.A.Z. D.A.Z. and G.V.T.; and writing—original draft preparation, G.E.A, D.A.Z. and G.V.T.; writing—review and editing, G.E.A, D.A.Z., G.V.T., G.V.T.; writing—original draft preparation, G.E.A, D.A.Z. and G.V.T.; writing—review and editing, G.E.A, P.F., P.K.G. and D.I.K. All authors have read and agreed to the published version of the manuscript. D.A.Z., G.V.T., P.F., P.K.G. and D.I.K. All authors have read and agreed to the published version of the manuscript.Funding: This research received no external funding. Conﬂicts of Interest: The authors declare no conﬂict of interest. Funding: This research received no external funding.

ConflictsReferences of Interest: The authors declare no conflict of interest.

1. Hurley, S. Planning eﬀective cellular mobile radio networks. IEEE Trans. Veh. Technol. 2002, 51, 243–253. [CrossRef] 2. Molina, A.; Athanasiadou, G.; Nix, A. The automatic location of base-stations for optimised cellular coverage: A new combinatorial approach. Proceedings of the Vehicular Technology Conference. In 1999 IEEE 49th; IEEE: Houston, TX, USA, 1999; pp. 606–610. 3. Amaldi, E.; Capone, A.; Malucelli, F. Radio planning and coverage optimization of 3G cellular networks. Wirel. Netw. 2007, 14, 435–447. [CrossRef]

Electronics 2020, 9, 339 10 of 10

4. Lakshminarasimman, N.; Baskar, S.; Alphones, A.; Willjuice Iruthayarajan, M. Evolutionary multiobjective optimization of cellular base station locations using modified NSGA-II. Wirel. Netw. 2011, 17, 597–609. [CrossRef] 5. Athanasiadou, G.E.; Zarbouti, D.; Tsoulos, G.V. Automatic location of base-stations for optimum coverage and capacity planning of LTE systems. In The 8th European Conference on Antennas and Propagation (EuCAP 2014); IEEE: Houston, TX, USA, 2014; pp. 2077–2081. 6. Valavanis, I.K.; Athanasiadou, G.E.; Zarbouti, D.A.; Tsoulos, G.V. Multi-Objective Optimization for Base-Station Location in Mixed-Cell LTE Networks. In Proceedings of the 10th European Conference on Antennas and Propagation (EuCAP), Davos, Switzerland, 10–15 April 2016. 7. Athanasiadou, G.; Tsoulos, G.; Zarbouti, D.; Valavanis, I. Optimizing Radio Network Planning Evolution Towards Microcellular Systems. Wirel. Pers. Commun. 2019, 106, 521–534. [CrossRef] 8. Rappaport, T.S.; Xing, Y.; MacCartney, G.R.; Molisch, A.F.; Mellios, E.; Zhang, J. Overview of Millimeter Wave Communications for Fifth-Generation (5G) Wireless Networks—With a Focus on Propagation Models. IEEE Trans. Antennas Propag. 2017, 65, 6213–6230. [CrossRef] 9. Tsoulos, G.V. Smart antennas for mobile communication systems: Benefits and challenges. Electron. Commun. Eng. J. 1999, 11, 84–94. [CrossRef] 10. Deb, K.; Pratap, A.; Agarwal, S.; Meyarivan, T. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Trans. Evol. Comput. 2002, 6, 182–197. [CrossRef] 11. 3GPP TR 38.900 v14.2.0, LTE; 5G; Study on Channel Model for Frequency Spectrum Above 6 GHz; ETSI: 650 Route des Lucioles, F-06921 Sophia Antipolis CEDEX, France, 2017. 12. Piersanti, S.; Annoni, L.; Cassioli, D. Millimeter waves channel measurements and path loss models. In Proceedings of the 2012 IEEE International Conference on Communications (ICC), Ottawa, ON, Canada, 10–15 June 2012; pp. 4552–4556. 13. Sun, S.; Rappaport, T.S.; Rangan, S.; Thomas, T.A.; Ghosh, A.; Kovacs, I.Z.; Rodriguez, I.; Koymen, O.; Partyka, A.; Jarvelainen, J. Propagation Path Loss Models for 5G Urban Micro- and Macro-Cellular Scenarios. In Proceedings of the IEEE 83rd Vehicular Technology Conference–Spring, Nanjing, China, 7 May 2016. 14. Pekka, K.; Juha, M.; Hentilä, L. WINNER II - D1.1.2 - Channel Models. IST-4-027756 WINNER II. 2007. Available online: http://www.ero.dk/6799B797-53CC-417E-B6FA-059A8E6AF350?frames=no& (accessed on 1 February 2020). 15. 3GPP TR 36.814; Technical Specification Group Radio Access Network; Evolved Universal Terrestrial Radio Access (E-UTRA);Further Advancements for E-UTRA Physical Layer Aspects (Release 9); 3GPP: 650 Route des Lucioles, Sophia Antipolis Valbonne, France, 2010. 16. Salem, A.A.; El-Rabaie, S.; Shokair, M. A Proposed Efficient Hybrid Precoding Algorithm for Millimeter Wave Massive MIMO 5G Networks. Wirel. Pers. Commun. 2020, 1–16. [CrossRef] 17. Sharma, V.; You, I.; Leu, F.; Atiquzzaman, M. Secure and efficient protocol for fast handover in 5G mobile Xhaul networks. J. Netw. Comput. Appl. 2018, 102, 38–57. [CrossRef] 18. Liu, L.; Zhou, Y.; Vasilakos, A.V.; Tian, L.; Shi, J. Time-domain ICIC and optimized designs for 5G and beyond: A survey. Sci. China Inf. Sci. 2019, 62, 1–28. [CrossRef][PubMed] 19. 3GPP TS 38.104, “5G; NR; Base Station (BS) Radio Transmission and Reception”; ETSI: 650 Route des Lucioles, F-06921 Sophia Antipolis CEDEX, France, 2019. 20. 3GPP TR 21.915, “Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; 5G; Release description; Release 15; ETSI: 650 Route des Lucioles, F-06921 Sophia Antipolis CEDEX, France, 2019.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).