On Topology Optimization and Routing in Integrated Access and Backhaul Networks: A Genetic Algorithm-based Approach
Charitha Madapatha, Behrooz Makki, Ajmal Muhammad, Erik Dahlman, Mohamed-Slim Alouini, Tommy Svensson
>> XXXXX < On Topology Optimization and Routing in Integrated Access andBackhaul Networks: A Genetic Algorithm-based Approach
Charitha Madapatha, Behrooz Makki,
Senior Member, IEEE,
Ajmal Muhammad,
Senior Member, IEEE,
Erik Dahlman,Mohamed-Slim Alouini,
Fellow, IEEE and Tommy Svensson,
Senior Member, IEEE.
In this paper, we study the problem of topology optimization and routing in integrated access and backhaul (IAB) networks, asone of the promising techniques for evolving 5G networks. We study the problem from different perspectives. We develop efficientgenetic algorithm-based schemes for both IAB node placement and non-IAB backhaul link distribution, and evaluate the effectof routing on bypassing temporal blockages. Here, concentrating on millimeter wave-based communications, we study the servicecoverage probability, defined as the probability of the event that the user equipments’ (UEs) minimum rate requirements aresatisfied. Moreover, we study the effect of different parameters such as the antenna gain, blockage, and tree foliage on the systemperformance. Finally, we summarize the recent Rel-16 as well as the upcoming Rel-17 3GPP discussions on routing in IAB networks,and discuss the main challenges for enabling mesh-based IAB networks. As we show, with a proper network topology, IAB is anattractive approach to enable the network densification required by 5G and beyond.
Index Terms —Integrated access and backhaul, IAB, Genetic algorithm, Node selection, Topology optimization, Densification,Millimeter wave, (mmWave) communications, 3GPP, Stochastic geometry, Poisson point process, Coverage probability, Germ-grainmodel, Wireless backhaul, 5G NR, Blockage, Relay, Routing, Tree foliage, Machine learning
I. INTRODUCTIONSeveral reports have shown an exponential growth of de-mand on wireless communications, the trend which is expectedto continue in the future [1]. To cope with such demands, 5Gand beyond networks propose various methods for capacityand spectral efficiency improvement. Here, one of the promis-ing techniques is network densification, i.e., the deploymentof many base stations (BSs) of different types such that thereare more resource blocks per unit area [2]–[5].The BSs need to be connected to the operators’ core networkvia a transport network, the problem which becomes challeng-ing as the number of BSs increases. Such a transport networkmay be provided via wireless or wired connections. Wired(fiber) connections are typically used for transport closer to thecore network and in the core network, where we need to handleaggregated traffic from multiple BSs. Wireless connections, onthe other hand, are used for backhaul transport in the radioaccess network (RAN) closer to the BSs.As reported in [2], the backhaul technology has largeregional variations. However, on a global scale, wirelessmicrowave technology has been a dominating media for the
Manuscript received xxxx; accepted November xxxx. Date of publicationxxxx; date of current version xxxx.This work was supported by the ChaseOnproject of Dept. of Electrical Engineering, Chalmers University of Technology.The work of C. Madapatha in this publication is part of his research work atChalmers University of Technology.C. Madapatha and T. Svensson are with the Department of ElectricalEngineering, Chalmers University of Technology, 412 96 Gothenburg, Sweden(e-mail: [email protected]; [email protected]).B. Makki is with Ericsson research, Lindholmen, Sweden, 417 56 G¨oteborg,Sweden (e-mail: [email protected]).A. Muhammad is with Ericsson research, Kista, Sweden, P.O. Box 16440(e-mail: [email protected]).E. Dahlman is with Ericsson research, Kista, Sweden, P.O. Box 16440 (e-mail: [email protected]).M.-S. Alouini is with the Computer, Electrical and Mathemati-cal Science and Engineering, King Abdullah University of Scienceand Technology (KAUST), Thuwal 23955-6900, Saudi Arabia (e-mail:[email protected]). last few decades. Recently, there is an increase in fiberdeployments attributed to geopolitical decisions and majorgovernmental investments. Thus, going forward, it is expectedthat microwave and fiber will be two dominating backhaultechnologies.Fiber offers reliable connection with high peak data rates.However, 1) the deployment of fiber requires a noteworthyinitial investment for trenching and installation, 2) may takea long installation time, 3) may be even not allowed in, e.g.,metropolitan areas. Wireless backhaul using microwave is awell-established alternative to fiber, providing 10’s of Gbpsin commercial deployments . Importantly, microwave is ascalable and economical backhaul technique that can meet theincreasing requirements of 5G networks. Compared to fiber,wireless backhauling comes with significantly lower cost andtime-to-market as well as higher flexibility, with no digging,no intrusion or disruption of infrastructure, and possible todeploy in principle everywhere [2].With the same reasoning and motivated by availabilityof massive bandwidth in millimeter wave (mmWave) spec-trum/network densification, integrated access and backhaul(IAB) network has recently received considerable attention[7]–[9]. With IAB, the goal is to provide flexible wirelessbackhauling using 3GPP new radio (NR) technology in inter-national mobile telecommunications bands, and provide notonly backhaul but also the existing cellular services in thesame node and via the same hardware. This, in addition tocreating more flexibility and reducing the time-to-market, isgenerally to reduce the cost for a wired backhaul, whichin certain deployments could impose a large cost for theinstallation and operation of the BS. Importantly, • Our internal evaluations at Ericsson shows that, evenin the presence of dark fiber, the deployment of IAB Recent results demonstrate even more than 100 Gbps over MIMO backhaullinks [6]. a r X i v : . [ c s . N I] F e b XXXXX < network gives an opportunity to reduce the total cost ofownership in urban/suburban areas. This is partly becausethe same hardware can be used both for access andbackhaul, i.e., less extra equipment is required especiallyfor in-band backhauling. • An integrated access/backhaul solution improves the pos-sibilities for pooling of spectrum where it can be up tothe operator to decide what spectrum resources to use foraccess and backhaul, rather than having this decided inan essentially static manner by spectrum regulators.In this way, IAB serves as a complement to microwaveand fiber backhaul specially in dense urban and suburbandeployments.The performance of IAB networks have been studied fromdifferent perspectives. Particularly, [10]–[14], develop variousresource allocation schemes, and [15], [16] study the effectof time/frequency division duplex based resource allocationon the throughput of IAB networks. Moreover, [17]–[19]utilize infinite Poisson point processes (PPPs) to evaluate thecoverage probability of multi-hop IAB networks. Then, [20],[21] investigate the feasibility/challenges of mmWave-basedIAB networks via end-to-end simulations. Also, [22] and [23]evaluate the potentials of using IAB in fixed wireless accessand unmanned aerial vehicle-based communication setups,respectively. In [24] and [25], we provide an overview of3GPP Rel-16 discussions on IAB. Moreover, [25] uses aFHPPP (FH: finite homogeneous), i.e., a PPP with a constantdensity and random distributions of the nodes in a finite region,to develop a tractable framework and thereby analyzes theperformance between the IAB, fiber-connected networks, andverifies the robustness of IAB to various environmental effects.Finally, [26] develops simulated annealing algorithms for jointscheduling and power allocation, and [27] designs a jointprecoder design and power allocation scheme maximizing thenetwork sum rate.Although IAB can in principle operate in every spectrumfor which NR operation is specified, the focus of the 3GPPwork on IAB has been on mmWave spectrum. This is intu-itive because of the access to wide bandwidth in mmWavespectrum, while the existing LTE spectrum is very expensiveto be used for backhauling. With a mmWave spectrum, how-ever, blockage and tree foliage may be challenging, as theyreduce the achievable rate significantly. Properly planned andoptimized networks could reap higher performances, and savecosts to network operators as they can avoid static blockagessuch as buildings/trees [28], [29]. On the other hand, alongwith enabling traffic-based load balancing, routing can wellcompensate for temporal blockages, e.g., busses/trucks passingby. However, as the network size increases in dense areas,which is the main point of interest in IAB networks, derivingclosed-form solutions for optimal network topology/routingbecomes infeasible. It should mentioned that the networkdeployment optimization can be done offline, and recalcu-lated whenever there are substantial changes in the blockingsituations, service rate requirements, and addition of newset of BSs. Still, the optimization problem quickly becomesvery large, thus motivating a potentially suboptimal machinelearning approach, since an exhaustive search over all possible deployment options quickly becomes infeasible (see SectionV for further details). In such cases, machine learning tech-niques give effective (sub)optimal solutions with reasonableimplementation complexity. These are the motivations for thispaper, in which we design routing and machine learning-basedtopology optimization schemes for IAB networks.In this paper, we study the problem of topology optimizationand routing in IAB networks. We study the problem fromdifferent points of views: • We design effective genetic algorithm (GA)-based tech-niques not only for IAB node placement but also for ded-icated non-IAB backhaul connection distribution. Here,concentrating on the characteristics of mmWave commu-nications, we present the results for the cases with anFHPPP-based stochastic geometry model [17], [30]. Asthe metric of interest, we consider the network servicecoverage probability which is defined as the probability ofthe event that the UEs’ minimum data rate requirementsare satisfied. • We study the effect of temporal blockages and routing onthe coverage probability. In this way, one can avoid boththe long-term and temporal blockages via topology opti-mization and routing, respectively. Also, the setup giveshints on the effectiveness of mesh-based communicationin IAB networks, although it is not yet considered by3GPP IAB standardization. • We summarize the main 3GPP Rel-16 agreements aswell as the upcoming Rel-17 discussions on routing, andhighlight the main challenges which need to be solvedbefore meshed IAB can be implemented. • Finally, we study the effect of different parameters suchas antenna gain, blockage and tree foliage on the systemperformance in both cases with well-planned and randomnetwork deployments.As we show, machine learning techniques provide effectivesolutions for deployment optimization which can be easilyadapted for different channel models, constraints and metricsof interest with no need for mathematical analysis. More-over, compared to random deployment, deployment planningincreases the coverage probability of the IAB networks signifi-cantly. On the other hand, with a well-planned network and fora broad range of blockage/tree foliage densities, the networkcan well handle these blockages with small routing updates.Finally, while the service coverage probability of the IABnetwork is slightly affected by stationary/temporal blockagesin urban areas, for a broad range of parameter settings, theblockage is not problematic for well-planned routing-enabledIAB networks, in the sense that its impact on the coverageprobability is negligible. On the other hand, high levels of treefoliage may reduce the coverage probability of the network insuburban areas; the problem which can be solved by properdeployment planning.It should be mentioned that the problem of routing in IABnetworks has been previously studied in the cases with differ-ent numbers of hops [10]–[14], [31]–[33]. Also, [24] developsa cost-optimal node placement scheme, and [34] proposes ajoint node placement and resource allocation scheme maxi-
XXXXX < mizing the downlink sum rate of IAB networks. On the otherhand, with different (non-IAB) network topologies/use-cases,various machine-learning based solutions have been previouslyproposed for topology optimization. For instance, [35]–[44]develop deep reinforcement (DR) algorithm-based solutionsfor topology optimization of different network configurations.Deep Q-learning is used in [45] to evaluate the cumulativetransmission rate in vehicular networks. Spectrum allocationand access mode selection evaluations are considered in [46],while the potential of using K -means clustering algorithm todesign ultra-reliable and low-latency wireless sensor networksis evaluated in [47]. In addition, [48] and [49] use DR learning-based algorithms to solve the large-scale load balancing prob-lem for ultra-dense networks. However, these works considerdifferent algorithms and network configurations as those westudy in this work. Moreover, our discussions on the effectof environmental parameters on the system performance aswell as the 3GPP agreements on IAB-based routing have notbeen presented before. These makes our discussions and theconclusions completely different from those presented in thestate-of-the-art works.II. IAB IN • A Centralized Unit (CU) including the packet data con-vergence protocol (PDCP) and radio resource control(RRC) protocols, • One or several Distributed Units (DUs) consisting ofradio link control (RLC), medium access control (MAC),and physical layer protocols.IAB specifies two types of network nodes (see also Fig. 1): • The IAB-Donor-node is the node consisting of CU andDU functionalities, and connects to the core network vianon-IAB, for example fiber, backhaul. A donor node DUmay serve devices, like a conventional UEs, but will alsoprovide wireless connectivity to IAB nodes via backhaullink(s). • The IAB node includes two modules, namely, DU andmobile terminal (MT). IAB-DU serves UEs as well as,potentially, downstream IAB nodes in case of multi-hopwireless backhauling. At its other side, an IAB-MT isthe unit that connects an IAB node with the DU of theparent/upstream node. Note that the parent node couldeither be an IAB-Donor-node or another IAB node incase of multi-hop backhauling.
Figure 1. Types of network nodes.Figure 2. IAB protocol stack of the user-plane.
The IAB architecture is thus based on a hierarchical or, atleast, a-cyclical structure where it is well-defined if a certainnode is “above” or “below” a certain other node and whereinformation flows in well-defined down-stream and up-streamdirections. The possibility for a more mesh-like structure withno well-defined hierarchy was briefly discussed during theinitial phase of the 3GPP work on IAB. However, majorityof the companies discard the idea owing to its complexity andno clear benefits.When it comes to the connectivity with the parent node,the IAB-MT connects to the DU of its parent node essentiallyas a normal UE. The link (i.e., the Uu interface) between theparent node DU and the MT of the IAB node then provides thelower-layer functionality and relays the F1 messages betweenthe donor-node CU and the IAB-node DU. The specificationof the F1 interface only defines the higher-layer protocols, forexample, the signaling messages between the CU and DU, butis agnostic to the lower-layer (i.e., transport network layer)protocols. With IAB, the NR radio-access technology (theRLC, MAC, and physical layer protocols) together with someIAB-specific protocols, provides the lower-layer functionalityon top of which the F1 interface is implemented. See Fig. 2showing the user-plane protocols of a multi-hop IAB network(the control plane has a similar structure).
A. Backhaul Adaptation Protocol
Backhaul adaptation protocol (BAP) is a new IAB-specificprotocol responsible for routing and bearer mapping of packetsin the IAB network. More specifically, the BAP layer isresponsible for forwarding of the packets in the intermediatenodes/hops between the IAB-donor-DU and the access IAB-node. For this purpose, the IAB-Donor-CU assigns a distinctBAP address to each IAB-node during the integration process,which facilitates the unique identification of each IAB-node inthe network. For the downstream traffic, the BAP layer of the
XXXXX < Figure 3. Structure for the BAP header.
IAB-Donor-DU will add a BAP header to packets receivedfrom the upper layer. Similarly, for the upstream traffic, theBAP layer of the access IAB-node will add a BAP header tothe upper layer packets. Figure 3 shows the structure for theBAP header, which contains a 10-bit BAP address field and a10-bit BAP path ID field apart from 1-bit flag and 3 reservebits for future use. Note that 3GPP specifications use the BAPRouting ID as a cover term for BAP address and BAP pathID fields. The purpose of the BAP address field is to carrythe address of the destination IAB-node, while the Path IDfield contains the path identity to be used for traversing thepackets towards the destination IAB-node. This latter field isimportant for situations where multiple paths are configuredfor an IAB-node to improve network robustness/resilience andachieve load balancing by transporting a part of the traffic viaeach path towards the IAB-node.To illustrate the above concept, figure 4 shows an exampletopology for IAB network where two paths (i.e., Path 1 andPath 2) have been configured for IAB-node 5 by the IAB-donor-CU. This means that the routing tables in the BAP layerof all the intermediate IAB-nodes (i.e., IAB1, IAB2, IAB3,etc.) are properly configured with next-hop link informationfor all the BAP addresses and BAP path IDs carried inthe packets BAP header that these nodes will route in thenetwork. Furthermore, the IAB-donor-DU will have mappingrules (configured by the donor-CU) how to select the BAPaddress and BAP path ID fields for packets from the upperlayer based on the information in the IP address fields (i.e.,DS/DSCP) of the F1-AP signaling.Suppose the IAB-donor-DU receives a packet with IP ad-dress fields marked with information that is mapped to BAPaddress 5 and BAP path ID 1, the donor-DU will add a BAPheader with proper field values (i.e., address 5 and path ID 1)and will forward the packet to IAB2. Once IAB2 receives thepacket, the node will examine the BAP header of the packetand based on the BAP address (carried in the packet) andits routing table information will transmit the packet towardsIAB4. Similarly, IAB4 will route the packet to IAB5, wherethe IAB5 upon examining the BAP header field of the packetwill notice that the packet is destined for it. Hence, IAB4 willremove the BAP header before delivering the packet to its
Figure 4. Example of routing in IAB network.Figure 5. Intra-CU dual connectivity vs inter-CU dual-connectivity upper layer for further processing.In another scenario, if the IAB1 receives a packet (fromIAB-donor-DU) with BAP header containing BAP address 5and path ID 2, IAB1 will forward the packet towards IAB3instead of IAB2, and so on IAB3 will forward the packet toIAB4. When it comes to the upstream traffic, the BAP layer ofIAB5 will add a BAP header containing IAB-donor-DU BAPaddress and appropriate path ID (either path ID 1 or path ID2 based on the configuration information) to packets receivedfrom the upper layer. Next, IAB5 will forward the packetsto IAB4, which will be further forwarded by IAB4 either toIAB1 or IAB2 depending on the path ID field value carried inthe packets BAP headers. Once the packets reach IAB-donor-DU, the DU will remove the BAP header before deliveringthe packets to the upper layer for subsequent processing.
B. IAB extensions in 3GPP release 17
XXXXX < It may also enable additional possibilities for load-balancingwithin the wireless backhaul, i.e., the possibility to moredynamically route data via different paths depending on theinstantaneous load conditions on different links.III. S
YSTEM M ODEL
This section presents the system model, including the chan-nel model, the considered UE association rule as well as theachievable data rates in the backhaul and access links. TableI summarizes the parameters used in the analysis. Considera dense urban area with a two-tier heterogeneous network(HetNet), i.e., a two-hop IAB network, where multiple MBSs(M: macro) and SBSs (S: small) serve the UEs (see Fig. 6).In this way, following the 3GPP definitions (see Section II),the MBSs and the SBSs represent the donor and the childIABs, respectively, and throughout the paper we may use theterminologies MBS/SBS and donor IAB/IAB interchangeably.With the IAB setup, both the MBSs and the SBSs are usedfor both access and backhaul. However, the donor IABs, i.e.,the MBSs, are non-IAB backhauled to the core network.
Note.
In practice, a majority of the SBSs receive IAB-typebackhaul from the MBSs wirelessly. However, a fraction of theSBSs may have access to non-IAB type dedicated backhaulconnections, where such a backhaul can be provided either byfiber or a wireless radio link operating on a different frequencythan the IAB network (see Fig. 6). In this paper, along with op-timizing the SBSs’ locations, one of our goals is to determinethe proper nodes with non-IAB type backhauling such that thecoverage probability is maximized. In terms of performanceevaluation, our analysis does not depend on if these non-IABbackhaul links are provided by fiber or wirelessly. In practice,however, the network performance may depend on the typeof such non-IAB links; Wireless radio backhaul link is quiteflexible, and one can provide an SBS with wireless non-IABbackhaul as long as there is a strong LoS connection betweenthe SBS and an MBS. Fiber connections, on the other hand,may be available in specific areas, and there may be lowflexibility in fiber distribution between the nodes. In summary,depending on the type of the non-IAB backhaul links, tothe SBSs, our non-IAB backhaul link distribution method,presented in Algorithm 1, may give an ultimate opportunisticpotential of the network performance.Finally, note that considering the two-hop network is moti-vated by the fact that, although the 3GPP standardization doesnot limit the possible number of hops, as also reported in, e.g.,[24], [52]–[54], traffic aggregation in the backhaul links andlatency becomes challenging as the number of hops increases.We model the IAB network using an FHPPP based randomdistribution of the nodes in a finite region [17]–[19], [27].Particularly, without topology optimization, which we useas the benchmark to evaluate the performance of plannednetworks, the FHPPPs χ M , χ S , χ U with densities, λ M , λ S and λ U , respectively, are used to model the spatial distributionsof the MBSs, the SBSs and the UEs, respectively. With ourtopology optimization, however, while the number of nodes arestill determined based on some random process, the locationsof the SBSs as well as the location of the non-IAB backhaul connections to a fraction of the SBSs are optimized, in termsof network service coverage probability (see Section V for thedetails).In our setup, in-band communication is considered whereboth the access and backhaul links share and operate inthe same mmWave spectrum band. This is motivated by thefact that in-band communication gives better flexibility forresource allocation, at the cost of coordination complexity.For simplicity, assume the network to be distributed over acircular disk D . However, the model can be well applied onevery arbitrary region D .For the blockage, we use the well-known germ grain modeldescribed in [55, Chapter 14], which provides accurate blindspot prediction, compared to stochastic models that assumeindependant blocking. Particularly, the model takes inducedblocking correlation into account and, thus, suits well forenvironments with large obstacles. Here, an FHPPP χ bl modelsthe blockage distribution in the area D with λ bl denoting thedensity. The blockings are considered to be walls of length l bl and independantly and identically distributed orientation θ bl . Later, we use a GA-based approach to optimize the SBSslocations inside the region D to preferably avoid blockage inthe backhaul links.Following the state-of-the-art mmWave channel model, e.g.,[56], the received power at each node can be expressed as P r = P t h t,r G t,r γ (1 m ) γ t,r || x t − x r || − κ t,r . (1)Here, P t represents the transmit power in each link, and h t,r denotes the independant small-scale fading in individuallinks. Particularly, in our study Rayleigh fading is consideredfor small-scale fading. Thereby, G t,r denotes the combinedantenna gain of the transmitter and receiver in the link, γ t,r is the propagation path loss, and γ (1 m ) is the reference pathloss at one meter distance while κ t,r is the tree foliage loss.The total path loss, in dB, is characterized according to the5GCM UMa close-in model described in [57]. Here, the pathloss is characterized byPL = 32 . ( r ) α + 20 log ( f c ) , (2)where f c is the carrier frequency, r is the propagation distancebetween the nodes, and α is the path loss exponent. Dependingon the blockage, line-of-sight (LoS) and NLoS (N: Non) linksare affected by different path loss exponents. The propagationloss of the path loss model is given by γ t,r = (cid:40) r α L , if LoS, r α N , if NLoS, (3)where α L and α N denote path loss exponents for the LoS andNLoS scenarios, respectively.5G and beyond systems are equipped with large antennaarrays which are used to minimize the propagation loss. Weuse the sectored-pattern antenna array model to characterizethe beam pattern and antenna gain, which is given by G t,r ( ϕ ) = (cid:40) G − θ HPBW ≤ ϕ ≤ θ HPBW g ( ϕ ) otherwise. (4) XXXXX < Figure 6. Schematic of the considered IAB network. A majority of the SBSs rely on IAB for backhauling. A small fraction of the SBSs, however, may havenon-IAB type backhaul where such a backhauling can be provided by fiber or wirelessly.
Here, ϕ represents the angle between the transmit and receiveantennas. Furthermore, θ HPBW is the half power beamwidth,and G denotes the main lobe gain of the antenna while g ( ϕ ) is the side lobe gain [56]. Finally, as in, e.g., [17], [19],[25], for tractability we assume that the UE antenna gain tobe 0 dB due to its omni-directional beam pattern, althoughUE beamforming in mmWave is an interesting future work toincorporate.Unless otherwise stated and in harmony with, e.g., [17]–[19], [25], we assume that the backhaul links are noise-limited.This assumption, which has been verified in [25], is motivatedby the high beamforming capacity in the inter IAB backhaullinks and the fact that simultaneous transmission/receptionis not considered in our setup. Then, Section V validatesthis assumption, and we verify the effect of the backhaulinterference on the coverage probability. Also, the inter-UEinterference is neglected with the assumption of sufficientisolation and low power of the UEs [22]. Particularly, theinterference model focuses on the aggregated interference onthe access links, caused by the neighbouring interferers, whichfor UE u is expressed as I u = (cid:88) i , u ∈ χ i,u \{ w u } P i h i,u G i,u γ (1 m ) γ x i ,x u (cid:107) x i − x u (cid:107) − , (5)where i denotes the set of BSs excluding the associated BS w u of user u . Also, for SBS s , the aggregated interference on the backhaul links is given by I s = (cid:88) j , s ∈ χ j,s \{ w s } P j h j,s G j,s γ (1 m ) γ x j ,x s (cid:107) x j − x s (cid:107) − , (6)where j denotes the set of transmitting BSs excluding theassociated BS w s of SBS s .We use an FHPPP denoted by χ T with density λ T tomodel the spatial distribution of the tree lines of length l T [58]. The tree foliage loss is estimated using the FittedInternational Telecommunication Union-Radio (FITU-R) treefoliage model [59, Chapter 7]. The model is well knownfor its applicability in cases with non-uniform vegetation andfrequency dependancy within 10-40 GHz range. Particularly,considering both in-leaf and out-of-leaf, vegetation states, thetree foliage loss in (1) is expressed as κ = (cid:26) . f . c d . , in-leaf . f . c d . , out-of-leaf, (7)where d is the vegetation depth measured in meter.In our setup, each UE has the ability to be connected toeither an MBS or an SBS depending on the maximum averagereceived power. Let a u ∈ { , } be a binary variable indicatingthe association with 1, while 0 representing the opposite. Thus, XXXXX < Table IT HE D EFINITION OF THE P ARAMETERS .Parameter Definition Parameter Definition χ M FHPPP of the MBSs χ U FHPPP of the UEs χ S FHPPP of the SBSs χ bl FHPPP of the blockings χ T FHPPP of the tree lines λ U UEs density λ M MBSs density λ S SBSs density λ bl Blocking density λ T Tree density ρ Service coverage probability d Vegetation depth θ Orientation of the blocking wall ϕ Angle between transmitter and receiver D Circular disk R Radius of the disk P t Transmission power P r Received power h Fading coeficient G Antenna gain γ (1 m ) Reference path loss at 1 meter distance γ Propagation path loss x Location of the node r Propagation distance between the nodes α Path loss exponent N Number of connected UEs f c Carrier frequency R th Minimum data rate threshold w u Associated cell w s Associated BS in backhaul link N f Number of non-IAB backhauled SBSs κ Tree foliage loss l T Tree line length l h Hop length λ temp Temporal blocking density K Number of random sets in the GA J Number of sets around the Queen in GA N it Number of iterations in the GA B Bandwidth ψ Percentage of bandwidth resources on backhaul for the access links a u = if P i G z,x h z,x γ (1 m ) γ z,x ( (cid:107) z − x (cid:107) ) − ≥ P j G j h z,y γ (1 m ) γ z,y ( (cid:107) z − y (cid:107) ) − , ∀ y ∈ χ j , j ∈ { m , s }| x ∈ χ i , , otherwise, (8)where i , j denote the BS indices, i.e., MBS or SBS. As in (III)for each UE u , the association binary variable a u becomes 1for the cell giving the maximum received power at the UE,while for all other cells it is 0, as the UE can only be connectedto one IAB node.Since the MBSs and the SBSs have large antenna arraysand can beamform towards the desired direction, the antennagain over the backhaul links can be assumed to be the same,and backhaul link association can be well determined basedon the minimum path loss rule, i.e., by a b,m = if γ b m ( (cid:107) z − x (cid:107) ) − ≥ γ b m ( (cid:107) z − y (cid:107) ) − , ∀ y ∈ χ m | x ∈ χ m , , otherwise. (9)For resource allocation, on the other hand, the mmWavespectrum available is partitioned into the access and backhaullinks such that (cid:40) B Backhaul = ψB,B Access = (1 − ψ ) B, (10)In practice, along with the MBSs which are non-IABbackhaul-connected, a portion of the SBSs may have dedicatednon-IAB backhaul connections, resulting in a hybrid IABnetwork. Therefore, in our deployment, some of the SBSs areIAB backhauled wirelessly and the others are connected todedicated non-IAB backhaul links.Let us initially concentrate on the IAB-type backhauledSBSs. Also, let, B backhaul and B access denote the backhaul andthe access bandwidths, respectively, while total bandwidth is B = B backhaul + B access . The bandwidth allocated for eachIAB-type wirelessly backhauled SBS, namely, child IAB, bythe MBS, i.e., IAB donor, is proportional to its load and thenumber of UEs in the access link. The resource allocation isdetermined based on the instantaneous load where each IAB-type backhauled SBS informs its current load to the associatedMBS each time. Thus, the backhaul-related bandwidth for the j -th IAB node, if it does not have dedicated non-IAB backhaulconnection, is given by B backhaul ,j = ψBN j (cid:88) ∀ j N j , ∀ j, (11)where N j denotes the number of UEs connected to the j -thIAB-type backhauled node and ψ ∈ [0 , is the fraction of thebandwidth resources on backhauling. Therefore, the bandwidthallocated to the j -th IAB-type backhauled node is proportionalto the ratio between its load, and the total load of its connectedIAB donor. Meanwhile, the access spectrum is equally sharedamong the connected UEs at the IAB node according to B access ,u = (1 − ψ ) B (cid:88) ∀ u N j,u , ∀ u, (12)where u denotes the UEs indices, and j represents each IAB-type backhauled node. Moreover, N j,u is the number of UEsconnected to the j -th IAB-type backhauled node to which UE u is connected. Finally, the signal-to-interference-plus-noiseratio (SINR) is obtained in accordance with (5) bySINR = P r / ( I u + σ ) , (13)where σ is the noise power.With our setup, the network may have three forms of accessconnections, i.e., MBS-UE, IAB-type backhauled SBS-UE,non-IAB backhauled SBS-UE, and the individual data rateswill behave according to the form in which the UE’s connec-tion has been established. Particularly, the rates experienced XXXXX < by the UEs in access links that are connected to MBSs or tothe IAB type-backhauled SBSs are given by R u = (1 − ψ ) BN m log(1 + SINR ( x u )) , if w u ∈ χ m , min (cid:18) (1 − ψ ) BN (cid:88) ∀ u N j,u log(1 + SINR ( x u )) , ψBN (cid:88) ∀ j N j log(1 + SINR ( x b )) (cid:19) , if w u ∈ χ s , (14)where j represents each IAB-type backhauled SBS connectedto the MBS. Then, m gives the associated MBS, s denotes theSBS, and u represents the UEs’ indices. Unlike an MBS whichshares some of its bandwidth with IAB-type backhauled SBSs,a non-IAB backhauled SBS has a bandwidth of B for access,and does not need to share its bandwidth for backhauling.Thus, the UEs connected to a non-IAB backhauled SBSexperience the rate given by R u = BN u log(1 + SINR ( x u )) , if w u ∈ χ s , (15)where N u denotes the total number of UEs connected tothe non-IAB backhauled SBS of which the considered UE isassociated. Depending on the associated cell, there are threepossible cases for the data rate of the UEs. First is the casewhen the UEs are connected to the MBSs, i.e., IAB donor, asdenoted by w u ∈ χ m in (14). Since the MBSs have non-IABbackhaul connection, the rate will only depend on the accessbandwidth available at the UE. In the second case, the UEsare connected to the IAB-type backhauled SBSs, as denotedby w u ∈ χ s in (14). Here, the SBSs have shared backhaulbandwidth from the IAB-Donor-nodes i.e., MBSs, and thus theUEs data rates depend on the backhaul rate of the connectedIAB-type backhauled SBS as well. Thus, in this case the UEis bounded to get the minimum between backhaul and accessrate. Then, the third case is when the UEs are associates withthe non-IAB backhauled SBSs as denoted in (15). Unlike inthe previous case, here the SBSs have full bandwidth B whichis not shared with backhauling.In the following, we present the GA-based schemes tooptimize the locations of the SBSs as well as the non-IABbackhaul link distribution to a fraction of the SBSs such thatthe network service coverage probability is maximized.IV. P ROPOSED A LGORITHM
In general, Rel-16 IAB network supports NLoS back-hauling. However, the performance of the IAB networks isconsiderably affected by the quality of the backhaul links,where, if possible, it is preferred to have IAB-IAB channelswith strong LoS signal strength. Also, in hybrid networkswhere a fraction of the SBS nodes may be backhauled viadedicated non-IAB backhaul links, it is important to obtain theset of SBSs that are critical to be non-IAB backhaul-connectedfor optimal performance. However, depending on the networksize, it may be difficult to obtain the appropriate location ofthe SBSs and/or the non-IAB backhaul link placement schemefor SBSs analytically. For instance, with N s SBSs and a budget of having N f non-IAB backhauled SBSs, there are (cid:0) N s N f (cid:1) possible combinationsof non-IAB backhauled SBS selections. Therefore, the optimalset of SBSs suitable for non-IAB backhaul link placement canindeed be obtained via exhaustive search for the cases with fewSBSs. However, as the network size increases, it is not feasibleto search over all possible solutions. The problem becomeseven more challenging with determining the optimal locationsof the SBSs as they can be distributed in the whole networkarea. Thus, it is important to design efficient algorithms toobtain the (sub)optimal SBS locations as well as dedicatednon-IAB backhaul link placement with low complexity.With this background, the state-of-the-art works mainlyconcentrate on either modeling the network by placing theBSs on a grid or distribute them randomly based on stochasticgeometry models. However, none of these models are accurate,as they give an optimistic or a pessimistic estimate of the net-work performance, respectively. Also, in practice, the networkmay be well planed such that, at least, high-quality backhaullinks are guaranteed. This is the motivation for our GA-based approach in which we propose a fairly simple networkdeployment optimization algorithm with no need for detailedmathematical analysis. This is important specially because • as we show in the following, with a well-planned networktopology the need for routing, to compensate for temporalblockages, decreases which results in considerable imple-mentation complexity reduction. • Moreover, with our proposed GA-based approach it ispossible to scale the network with proper deployment asmore IAB nodes/non-IAB backhaul link connections areadded to the network. • Finally, due to the generic characteristics of machinelearning schemes, one can apply the same technique asour proposed GA method for both non-IAB backhaullink placement and SBS location optimization, as wellas for the cases with different channel models/metrics ofinterest.It should be noted that, we are interested in the potential ofoptimal partial non-IAB type backhaul connections in order tofind an upper bound on the performance of any real networkthat might be constrained. Such a constrained partial non-IABtype backhaul link deployment optimization would also bean interesting extension of this work, and any such networkperformance would be in between the optimized and therandom partial non-IAB type backhaul link deployment.Particularly, in this paper, we propose two GA-based ap-proaches [60] to identify the optimal SBSs to be non-IABbackhaul-connected and the optimal locations for the SBSs, asexplained in Algorithms 1 and 2, respectively. The algorithmsare used to maximize the service coverage probability definedas the fraction of the UEs which have instantaneous UE datarates higher than or equal to a threshold R th . That is, using(14) and (15), the service coverage probability is given by ρ = Pr( R U ≥ η ) . (16)In words, both algorithms are based on the procedure de-scribed below. As shown in Fig. 7, we start the algorithm XXXXX < Figure 7. An example of the proposed GA in Algorithm 1 for non-IABbackhaul link distribution between a fraction of the SBSs. by considering K possible selection strategies. For instance,Algorithm 1 considers K possible SBS sets for non-IAB back-haul link placement and Algorithm 2 considers K possiblelocation sets for the SBSs. Then, in each iteration, we findthe best strategy, i.e., selected solution, that maximizes theconsidered utility function, compared to the other K − se-lected strategies. This best strategy is referred to as the Queen.The Queen is considered as one of the possible solutions inthe next iteration to guarantee the monotonic improvement ofthe algorithm performance in successive iterations. Also, foreach iteration we create J < K sets around the Queen. Thesematrices are created by applying slight modifications to theQueen, i.e., as a kind of mutation. For example, changingfew SBSs of the set associated with the Queen generatesthese new sets needed for optimal SBS selection for non-IAB backhaul link placement in Algorithm 1. Also, in eachiteration K − J − sets of selection strategies are generatedrandomly, to avoid the network to be trapped in a localminimum, and the iterations continue for N it iterations decidedby the network designer depending on the problem at hand(See Section V). After running all considered iterations, theultimate Queen is returned as the best selection rule for thecurrent network instance. Particularly, Algorithm 1 returnsthe optimal SBS selection rule for non-IAB backhaul linkplacement, while Algorithm 2 returns the optimal SBS locationselection rule. The suitable parameter setting for K , J and N it in the algorithms can be obtained by the designer.Considering Algorithms 1 and 2, the following points areinteresting to note: • Our proposed algorithms result in significantly lowercomplexity, in comparison with the exhaustive search, asit only checks KN it number of possible solutions (seeSection V). • Moreover, due to Step 7 of the algorithms, where K − J − random possible solutions are checked in each iteration,the proposed algorithms mimic the exhaustive search if N it → ∞ , and they reach the globally optimal selectionrule if asymptotically many iterations are considered[60]. • Unlike typical GAs, we do not use the crossover op-eration. This is because the proposed algorithms workwell with no need for the additional complexity of thecrossover operation, and converge with a few iterations(see Section V). However, it is straightforward to include the crossover into the proposed algorithms where, forinstance, the Queen and the next best solutions arecombined to generate new possible solutions. • The proposed algorithms optimize the network deploy-ment off-line. However, it is straightforward to scale thenetwork and adapt the algorithm in an on-line manner. Forinstance, adding new set of SBSs to an already-plannednetwork deployment, one can rerun the algorithm for onlya few iterations with the initial considered solutions notrandomly but based on the Queen of the already-plannednetwork.Finally, it should be noted that: • Depending on the infrastructures and the availability ofnon-IAB backhaul link connection, in practice it maynot be possible to provide some SBSs with a non-IABbackhaul link connection (either fiber or a dedicatedLoS nonIAB wireless backhaul). This is because thethose connections may be available in specific areas.In this way, as explained in Section III, Algorithm 1gives an optimistic ultimate network performance, aswe consider no limitation for non-IAB backhaul linkdistribution among the SBSs. Then, depending on thespecific network deployment, it is straightforward toadapt Algorithm 1 to consider restrictions on non-IABbackhaul link distribution among the SBSs. • According to the 3GPP discussions, one can considertwo different, namely, wide-area and local-area, IABnetwork deployments. Local-area IAB deployment refersto the cases with an unplanned network where the mobileterminal (MT) module of the IAB nodes have UE-typefunctionality, in terms of transmit power etc. Wide-areaIAB network, on the other hand, refers to the cases withwell-planned deployment and gNB-type functionalitiesfor the IAB nodes. In this way, the proposed schememainly concentrates on the wide-area IAB network de-ployment, as the main use-case of the IAB networks.
XXXXX < Algorithm 1
GA-based non-IAB Backhaul Link Placement.
In each network instance with a budget for N f non-IABbackhaul-connected SBSs, and N s > N f SBSs, do thefollowings:
I. Consider K sets of N f non-IAB backhaul-connectedSBSs, F k , and for each set create the correspondingchannel matrix. Then, for each matrix H k , k = 1 ..., K ,implement the system model in Section III.II. For each selected possible solution F k , evaluate theobjective function U k , k = 1 , ...., K. For instance,considering the service coverage probability ρ as theobjective function, U k is given by (16).III. Find the set of the SBSs among the considered so-lutions F k , ∀ k, which gives in the best value of theobjective function, service coverage probability (theQueen), e.g., F i where ρ ( H k ) ≤ ρ ( H i ) , ∀ k = 1 , ..., K .IV. F ←− F i V. Generate
J < K, sets of SBSs F new j , j = 1 , ..., J, around the Queen, i.e., F i . These sets of SBSs aregenerated by making small changes to the Queen, forinstance, by replacing few SBSs with other SBSs.VI. F j +1 ←− F new j , j = 1 , ..., J. VII. Use the same procedure as in Step 1 and regeneratethe remaining sets F j , j = J + 2 , ..., K , randomly.VIII. Proceed to Step 2 and continue the process for N it iterations pre-considered by the network designer. Return the Queen as the optimal SBS selection rulefor non-IAB backhaul link placement.
Algorithm 2
GA-based SBS Location Selection.
In each network instance with N s SBSs, from all possiblelocations in the space, do the followings:
I. Consider K sets of L k locations, and for each set cre-ate the corresponding channel matrix H k , k = 1 ..., K ,according to the system model in Section V.II. Evaluate the objective function for each set, i.e., U k , k = 1 , ...., K. For instance, considering the servicecoverage probability ρ as the objective function, U k isgiven by (16).III. Find the Queen, i.e., the set of locations which givesthe best value of the objective function, i.e., servicecoverage probability, among the considered sets, e.g., L i where ρ ( H k ) ≤ ρ ( H i ) , ∀ k = 1 , ..., K ,IV. L ←− L i V. Generate
J < K, sets of locations L new j , j = 1 , ..., J ,around L i . These sets of locations are generated bymaking small changes to the Queen, for instance, byreplacing few locations with another sets of locations.VI. L j +1 ←− L new j , j = 1 , ..., J. VII. Use the same procedure as in Step 1 and regeneratethe remaining sets L j , j = J + 2 , ..., K , randomly.VIII. Proceed to Step 2 and continue the process for N it iterations pre-considered by the network designer. Return the Queen as the optimal SBS location selectionrule.
Table IIS
IMULATION P ARAMETERS .Parameters ValueCarrier frequency 28 GHzBandwidth 1 GHzIAB node and UEs density { MBS, SBS, UE } = (2, 50, 500) /km Blocking density 500 /km Path loss exponents { LoS, NLoS } = (3, 4)Main lobe antenna gains { MBS, SBS, UE } = (18, 18, 0) dBiSide lobe antenna gains { MBS, SBS, UE } = (-2, -2, 0) dBiHalf power beamwidth 30Noise power 5 dBPercentage of non-IAB back-hauled SBS nodes 10%In-leaf percentage 15%Tree depth 7.5 m V. P
ERFORMANCE E VALUATION O F D EPLOYMENT O PTIMIZATION
The simulation results and discussions are divided into threemain areas in which 1) we evaluate the convergence behaviourof the proposed algorithms, and we study their effect onoptimizing the IAB network performance, 2) verify the effectof environmental parameters on the coverage probability, and3) evaluate the system performance for different transmissioncapabilities of the nodes. Then, in Section VI, we investigatethe effect of routing on the performance of IAB networksexperiencing temporal blockings.The general system parameters are presented in Table IIand, in each figure, we give the detailed system parametersin the figure captions. The IAB network is deployed in a 2Ddisk, in which the blockage, and the tree distributions are alsomodelled using statistical models described in Section III. Inparticular, the network is a hybrid IAB deployment, of whicha fraction of the SBSs will be non-IAB backhaul-connectedwhile the others are backhauled using IAB. In all figures,except for Fig. 13 which studies the system performance insuburban areas, we focus on dense areas as the most importantuse-case in IAB networks. Also, in all figures, except in Fig.15, we ignore interference in the backhaul links, and assumethem to be noise-limited. In Figs. 8,12, 13-10, we studythe system performance in the cases with non-IAB backhaullink placement optimization (Algorithm 1). Figures 9, 14, 11and 16 present the results for the cases with Algorithm 2optimizing the SBSs locations.
A. On the Performance of the Proposed Algorithms
In Figs. 8-9, we study the convergence performance of theproposed algorithms, and compare the results with the caseshaving only MBSs or random network deployment. Figure. 8shows the service coverage probability achieved for differentnumbers of iterations in Algorithm 1 with optimal non-IABbackhaul link connection distribution and different algorithmparameters K and J . Here, the results are presented for thecases with 10% of the SBSs having the possibility to benon-IAB backhaul-connected. Then, Fig. 9 demonstrates theIAB network service coverage probability as a function of thenumber of iterations in Algorithm 2, and compares the results XXXXX < Figure 8. Service coverage probability as a function of the number ofiterations in Algorithm 1 with non-IAB backhaul link connection distribution,and P m , P s , P u = (40 , , dBm. The parameters are set to λ M = 2 km − , λ S = 50 km − and λ U = 500 km − .Figure 9. Service coverage probability as a function of the number ofiterations in Algorithm 2 with IAB node placement optimization, and P m , P s , P u = (40 , , dBm. The parameters are set to λ M = 2 km − , λ S = 50 km − and λ U = 500 km − . with the benchmark schemes using only MBSs or randomnetwork deployment of which 10% of the SBSs are non-IABbackhaul-connected.As seen in Figs. 8 and 9, the developed Algorithms 1and 2 converge rapidly to give a maximum service cover-age probability. For example, Fig. 8 converges with almost N it = 20 iterations which, with K = 6 , leads to a total of 120possible solution checkings. As a result, the proposed algo-rithm reduces the complexity compared to exhaustive searchsignificantly because with λ s = 50 km − and the networkarea of km − exhaustive search requires (cid:0) (cid:1) (cid:39) × solution checkings, i.e. (cid:39) times larger search thanthose in our proposed scheme. In particular, the proposedalgorithms have improved the service coverage probability,compared to the IAB network with random node locationsand random non-IAB backhaul connections, significantly. Forinstance, with the parameter settings of Fig. 8, optimizing thenon-IAB backhaul link distribution among of the SBSs Figure 10. CDF of the achievable rates with P m , P u = (40 , dBm for non-IAB backhaul link location optimization, λ M = 2 km − , λ S = 20 km − and λ U = 500 km − . increases the coverage probability from 40% with random non-IAB backhaul link distribution to 85%. Moreover, with theparameter settings of Fig. 9, optimizing the SBSs locationleads to a coverage probability increment from 40% withrandom network deployment to 55%. Finally, as expected andalso demonstrated in Figs. 8-9, as the UEs density increases,MBSs alone can not support the UEs’ coverage probabilityrequirements, and indeed we need to densify the network using(IAB) nodes of different types. In this way, as also experiencedin practical network implementations, a well-planned networkdeployment results in significant performance improvement,which reduces the need for high network node density as wellas the implementation cost.Note that, while Figs. 8-9 show monotonic improvementof the system performance in successive iterations, in someiterations the proposed algorithms may follow a ladder-shapeconvergence pattern. This is because the service coverageprobability does not necessarily improve in each iteration,and there is a possibility to reach a local optimum in someiterations. However, we always elude the local minima due toStep 7 of Algorithms 1 and 2. Thus, given that sufficientlylarge number of iterations are carried out, the algorithmconverges to a (sub)optimal solution.In Fig. 10 and 11, we study the cumulative distributionfunction (CDF) of the UEs achievable data rates in the caseswith non-IAB backhaul link distribution and SBS location op-timization, respectively, and compare the results with randomnetwork deployment. Here, the parameters are set to λ M = 2 km − , λ S = 20 km − and λ U = 500 km − , and in all cases of the SBSs are non-IAB backhaul-connected. As can beseen in Fig. 10, with a random deployment and the parametersettings of the figure, (almost) all UEs maximum achievablerates are below Mbps, the result which holds for bothconsidered values of the IAB nodes transmit powers. On theother hand, optimizing the non-IAB backhaul link distributionbetween the SBSs gives the chance to support higher accessdata rates, depending on the UEs position and their associatedbackhaul links qualities. For instance, as opposed to the cases
XXXXX < Figure 11. CDF of the achievable rates with P m , P u = (40 , dBm for SBSlocation optimization, λ M = 2 km − , λ S = 20 km − and λ U = 500 km − .Figure 12. Service coverage probability of the IAB network as a function ofthe blocking density λ B , with P m , P s , P u = (40 , , dBm and differentmethods of non-IAB backhaul link connection distribution among of theSBSs. The parameters are set to λ M = 2 km − , λ S = 50 km − , λ U = 500 km − and N it = 20 . with random network deployment, with P s = 24 dBm around of the UEs may experience > Mbps access rates,if the non-IAB backhaul link distribution among of theSBSs is properly planned. The same point is also observedin Fig. 11 where, as opposed to the cases with randomdeployment, the SBS location-optimized network can supportUEs data rates up to 1200 Mbps.
B. Effect of Blocking and Tree Foliage
In contrast to the non-IAB backhaul-connected networks,IAB networks may be affected by environmental effects spe-cially the blockage and the tree foliage . In Figs. 12 and13, we respectively study the effect of the blockage andtree foliage on the coverage probability of the IAB network As reported in [25], with the typical hop lengths of the IAB networks and28 GHz, the effect of the rain on the coverage probability of IAB network isnegligible. Figure 13. Service coverage probability of the IAB network as a function oftree density λ T , with P m , P s , P u = (40 , , dBm and different methodsof non-IAB backhaul link connection distribution among of the SBSs.The parameters are set to λ M = 2 km − , l h = 450 m, λ U = 500 km − and η = 50 Mbps. with random deployment or GA-optimized non-IAB backhaullink distribution. Here, the results are presented for differentrate thresholds of the UEs, i.e., η in (15). In particular,Fig. 12 shows the service coverage probability consideringthe FHPPP-based germ-grain blockage model for differentblocking densities.Although urban areas are the main point of interest for IABnetwork, to study the potentials of its usage in suburban areas,in Fig. 13 we demonstrate the service coverage probability asa function of the tree density in the suburban areas. Here,we present the results for the average hop distance l h = 450 mcorresponding to SBSs density, λ S = 8 km − . This is motivatedby, e.g., [61], reporting the tree foliage as one of the mainchallenges of IAB in suburban areas. According to Figs. 12-13, the following points can be concluded: • The GA-based planned deployment shows significantimprovement and resilience to blockage and tree foliage,compared to random deployment, where the coverageprobability is not much affected by the blockage (Fig.12). For instance, with the given system configurationin Fig. 12, the GA optimized setup shows 0.85 servicecoverage probability at η = 150 Mbps, λ B = 1000 km − ,while the coverage probability reduces to 0.72 at λ B =2000 km − , i.e., only 15% coverage loss by doublingthe blockage density. On the other hand, with a randomnetwork deployment, η = 150 Mbps and λ B = 1000 ,the coverage probability is only 0.4 and it is dropped to0.23, i.e., 42% performance degradation, as the blockagedensity increases to λ B = 2000 km − . • In suburban area and with a random network deployment,the coverage probability is considerably affected by thetree foliage, especially when the trees density and/orlength increase. However, we note that the introductionof GA optimization on selecting the SBSs with non-IABbackhaul-connection has brought resilience to the treefoliage. This is due to the fact that the algorithm finds
XXXXX < Figure 14. Service coverage probability of the IAB network as a function ofthe SBSs transmit power P s , with P m , P u = (40 , dBm for both non-IABbackhaul link location and node placement optimization methods, λ M = 2 km − , λ U = 500 km − and η = 150 Mbps. the optimum set of nodes minimizing the SBS links withhigh losses due to tree foliage. For instance, consideringthe settings of Fig. 13 and the random FHPPP model with l T = 15 m, the service coverage probability drops from0.75 to 0.55 (26% coverage degradation) when the treedensity is increased from 250 to 1250 km − . However,the same tree density increase at l T = 15 m in GA-optimized network gives a drop only from 0.88 till 0.83,i.e., only performance drop, the result which is almostindependent of the tree length.In general, the robustness of IAB in the presence of treefoliage is hard to predict due to the fact that the link qualitycan vary depending on the characteristics of the tree lines.Particularly, the backhaul links quality may change due towet trees, snow on the trees, wind and varying percentageof leaves in different seasons. However, we conclude that,although the IAB is prone to medium/highly densified treefoliage in suburban areas, network planning can reduce muchof its adverse effect, and the mmWave IAB is expected towork well for areas with low/moderate foliage level. C. Effect of Antenna Gain and Transmit Power
In Fig. 14, we demonstrate the service coverage proba-bility as a function of the SBS transmit power for threescenarios, namely, random FHPPP-based deployment, GA-based non-IAB backhaul link distribution and GA-based SBSlocation optimization. Also, Fig. 15 shows the service coverageprobability as a function of the SBS antenna gain for ran-dom FHPPP-based deployment with non-IAB backhaul-connected SBSs, macro-only network and GA-optimized non-IAB backhaul link distribution between of the SBSs. Inaddition, to verify the effect of interference in the backhaullinks, the figure shows the service coverage probability inthe presence of both noise-limited and noise plus interferencelimited backhaul links. Here, we increase the SBS antennas’main lobe gain, while fixing the side lobe gain at -2 dB.
Figure 15. Service coverage probability of the IAB network as a functionof the SBSs antenna gain G , with P m , P s , P u = (40 , , dBm for non-IAB backhaul link location optimization, λ M = 2 km − , λ U = 500 km − , η = 150 Mbps and N it = 20 . As we observe in Fig. 14, both GA-optimization methodsused for selecting the dedicated non-IAB backhauled nodesand selecting SBSs locations have significantly increased thesystem coverage probability, compared to random deployment,and the relative effect of network planning increases withthe SBSs’ transmit power (Fig. 14). Moreover, with differentdeployment conditions and the considered range of transmitpowers, the coverage probability increases almost linearly withthe SBSs transmit power, while the relative benefit of the trans-mit power increment increases in the cases with a well-plannednetwork (Fig. 14). Also, Fig. 15 demonstrates that, for theconsidered parameter setting of the figure and moderate/highantenna gains, the system performance is almost insensitiveto the antenna gain specially if the network is well planned.Finally, as seen in Fig. 15, the impact of the interference inthe backhaul links is negligible, and thus, the backhaul linkscan be well assumed to be noise-limited (Also, see [17]–[19],[25] for further discussions).VI. O
N THE E FFECT OF R OUTING
As demonstrated, deployment planning can compensate forstationary blockages/tree foliage. On the other hand, depend-ing on, e.g., the height of the SBSs, the (backhaul) links maybe temporally blocked by, for instance, trucks passing by. Insuch cases, routing can be used to reduce the coverage prob-ability degradation. For this reason, in this section, we studythe effect of routing on the performance of IAB networks (seeSection II for 3GPP standardization agreements on routing).Note that, in general, routing can be utilized not only fortemporal blockages but also for load balancing in the caseswith varying data traffic. In this paper, we concentrate ontemporal blockage, and load balancing-based routing is out ofthe scope of our work. Note that, here, the network deploymentis first optimized based on static blockages/tree foliage. Then,by temporal blockage we refer to the blockages that are addedto the network after the deployment optimization is performed.
XXXXX < Figure 16. Service coverage probability of the IAB network as a functionof the temporal blocking density λ temp , with P m , P s , P u = (40 , , . Theparameters are set to λ M = 2 km − , λ S = 50 km − , λ U = 500 km − and λ B = 700 km − Figure 17. Percentage of routing updating of the IAB network as a functionof the temporal blocking density λ temp with P m , P s , P u = (40 , , . Theparameters are set to λ M = 2 km − , λ S = 50 km − , λ U = 500 km − and λ B = 700 km − Figure 16 shows the service coverage probability consid-ering a static blocking density λ B = 700 km − for differenttemporal blocking densities. In addition, to understand the IABsensitivity for temporal blockings and the effect of the routing,in Fig. 17, we plot the percentage of the links that have beenupdated by routing as a function of the density of the temporalblockings added to the network. The results are presented forvarious cases with random deployment, GA-optimized non-IAB backhaul link distribution or GA-optimized SBS loca-tions. Here, by routing, the received powers are recalculated,the association matrix is re-updated and thereby the datarates are calculated again, i.e., (8), (9), (13) and (14), areadapted based on the presence of temporal blockages such thatthe coverage probability degradation is minimized. Also, bypercentage of routing update we refer to the fraction of links inthe network that have changed their associated BS, both in theaccess and backhaul links. Here, the results are demonstratedfor different transmit powers of the SBSs. According to Fig. 16-17 the following points can be concluded. • Unless for high densities of temporal blockings, theservice coverage probability of the IAB network is notdegraded much by temporal blockage (Fig. 16). Also,the introduction of GA to optimize the dedicated non-IAB backhaul connections and node locations has broughtfurther resilience in the network to temporal blockage.Finally, the sensitivity to temporal blockage increasesslightly at low SBS transmit powers (Fig. 16). • As demonstrated in Fig. 17, with temporal blockage,the routing scheme may update the access links of theUEs to the IAB nodes. However, 1) for a broad oftemporal blockage densities for both non-IAB backhaulconnection-optimized and random deployments, the ac-cess links updates are less than . Also, 2) in allconsidered cases, the backhaul links do not need to beupdated due to temporal blockage. This is intuitivelybecause the IAB donor-IAB backhaul links are strong tosupport the required rates and the presence of temporalblockage does not affect their efficiency much unlessfor high temporal blockage densities. Finally, comparedto random network deployment, optimizing the non-IABbackhaul link distribution among of the SBSs withGA has slightly reduced the percentage of routing updatewith the addition of temporal blockings. For instance,with the parameter settings of Fig. 17, λ temp = 50 km − ,and P s = 28 dBm, in random network deployment onemay need a routing update of 3.6%, while in the GA-optimized network it is only 2.9%.In this way, the results indicate that, while deploymentoptimization can well robustify the network to static block-ages, with a well-planned network the system performanceis almost insensitive to low/moderate temporal blockages,and routing may not be required unless for high temporalblockage densities/severe coverage probability requirements.On the other hand, depending on the data traffic variation andthe number of hops in the IAB network, the routing may beof interest in load balancing. XXXXX < VII. C
ONCLUSION
We studied the problem of deployment optimization androuting in IAB networks to guarantee high coverage proba-bility in the presence of tree foliage/blockage. Moreover, wereviewed the recent 3GPP agreements on IAB-based routing,as well as the key challenges to enable meshed IAB.As we showed, machine-learning techniques can be effec-tively utilized for deployment optimization, with no need formathematical analysis and with the capability to be adaptedfor different channel models/constraints/metrics of interest.Particularly, the proposed algorithm reduces the complexitycompared to exhaustive search significantly because with typ-ical network area our proposed scheme requires orders of mag-nitude less solution checkings compared to exhaustive search.Also, while deployment planning boosts the coverage proba-bility of IAB networks, compared to random deployment, sig-nificantly, for a broad range of coverage constraints/blockagedensities, the impact of routing to increase redundancy maybe negligible. Indeed, routing may be of interest in the caseswith severe availability constraints/high blockage densities aswell as for load balancing. Finally, in practice, deploymentplanning may be affected by, e.g., the availability of non-IAB backhaul connection in specific areas, and the designermay consider, e.g., seasonal tree foliage variations, rental costsand/or foreseen infrastructure changes.VIII. A
CKNOWLEDGMENT
This work was supported in part by VINNOVA (SwedishGovernment Agency for Innovation Systems) within the VINNExcellence Center ChaseOn.R
Commun. Surv. Tutor. , vol. 18,no. 3, pp. 1617–1655, Feb. 2016.[4] S. Dang, O. Amin, B. Shihada, and M.-S. Alouini, “What should 6Gbe?”
Nature Electronics , vol. 3, no. 1, pp. 20–29, Jan. 2020.[5] X. Gao, O. Edfors, F. Tufvesson, and E. G. Larsson, “Massive mimoin real propagation environments: Do all antennas contribute equally?”
IEEE Trans. Wirel. Commun. , vol. 63, no. 11, pp. 3917–3928, July 2015.[6] C. Czegledi, M. H¨orberg, M. Sj¨odin, P. Ligander, J. Hansryd, J. Sand-berg, J. Gustavsson, D. Sjo¨oberg, D. Polydorou, and D. Siomos,“Demonstrating 139 Gbps and 55.6 bps/Hz Spectrum Efficiency Using8x8 MIMO over a 1.5 km Link at 73.5 GHz,” accepted for presentationin IMS2020, CA, USA, Aug. 2020.[7] C. Dehos, J. L. Gonz´alez, A. D. Domenico, D. Kt´enas, and L. Dussopt,“Millimeter-wave access and backhauling: the solution to the exponentialdata traffic increase in 5g mobile communications systems?”
IEEECommun. Mag. , vol. 52, no. 9, pp. 88–95, Sept. 2014.[8] Y. Li, E. Pateromichelakis, N. Vucic, J. Luo, W. Xu, and G. Caire,“Radio resource management considerations for 5g millimeter wavebackhaul and access networks,”
IEEE Commun. Mag. , vol. 55, no. 6,pp. 86–92, Jun. 2017.[9] N. Rajatheva, I. Atzeni, S. Bicais, E. Bjornson, A. Bourdoux, S. Buzzi,C. D’Andrea, J.-B. Dore, S. Erkucuk, M. Fuentes et al. , “Scor-ing the terabit/s goal: Broadband connectivity in 6g,” arXiv preprintarXiv:2008.07220 , 2020. [10] M. N. Islam, S. Subramanian, and A. Sampath, “Integrated AccessBackhaul in Millimeter Wave Networks,” in
Proc. IEEE WCNC’2017 ,CA, USA, Mar. 2017, pp. 1–6.[11] M. N. Islam, N. Abedini, G. Hampel, S. Subramanian, and J. Li, “Inves-tigation of performance in integrated access and backhaul networks,” in
Proc. IEEE INFOCOM WKSHPS’2018 , HI, USA, Apr. 2018, pp. 597–602.[12] Y. Li, J. Luo, R. A. Stirling-Gallacher, and G. Caire, “Integrated accessand backhaul optimization for millimeter wave heterogeneous networks,” arXiv , 2019. [Online]. Available: https://arxiv.org/abs/1901.04959.[13] Y. Liu, A. Tang, and X. Wang, “Joint Incentive and Resource AllocationDesign for User Provided Network Under 5G Integrated Access andBackhaul Networks,”
IEEE Trans. Netw. Sci. Eng. , vol. 7, no. 2, pp.673–685, April-June 2020.[14] N. Tafintsev, D. Moltchanov, M. Simsek, S. Yeh, S. Andreev, Y. Kouch-eryavy, and M. Valkama, “Reinforcement learning for improved uav-based integrated access and backhaul operation,” in
Proc. IEEE ICCWorkshops’2017 , 2020, pp. 1–7.[15] M. N. Kulkarni, J. G. Andrews, and A. Ghosh, “Performance ofDynamic and Static TDD in Self-Backhauled Millimeter Wave CellularNetworks,”
IEEE Trans. Wireless Commun. , vol. 16, no. 10, pp. 6460–6478, Oct. 2017.[16] B. Makki, M. Hashemi, L. Bao, and M. Coldrey, “On the Performance ofFDD and TDD Systems in Different Data Traffics: Finite Block-LengthAnalysis,” in
Proc. VTC-Fall’2018 , IL, USA, Aug. 2018, pp. 1–5.[17] C. Saha, M. Afshang, and H. S. Dhillon, “Integrated mmWave Accessand Backhaul in 5G: Bandwidth Partitioning and Downlink Analysis,”in
Proc. ICC’2018 , MO, USA, May 2018, pp. 1–6.[18] S. Singh, M. N. Kulkarni, A. Ghosh, and J. G. Andrews, “TractableModel for Rate in Self-Backhauled Millimeter Wave Cellular Networks,”
IEEE J. Sel. Areas Commun. , vol. 33, no. 10, pp. 2196–2211, Oct. 2015.[19] C. Saha and H. S. Dhillon, “Millimeter Wave Integrated Access andBackhaul in 5G: Performance Analysis and Design Insights,” arXiv ,2019. [Online]. Available: https://arxiv.org/pdf/1902.06300.pdf[20] M. Polese, M. Giordani, A. Roy, S. Goyal, D. Castor, and M. Zorzi,“End-to-End Simulation of Integrated Access and Backhaul atmmWaves,” in
Proc. IEEE CAMAD’2018 , Barcelona, Spain, Sept. 2018,pp. 1–7.[21] M. Polese, M. Giordani, T. Zugno, A. Roy, S. Goyal, D. Castor, andM. Zorzi, “Integrated Access and Backhaul in 5G mmWave Networks:Potential and Challenges,”
IEEE Commun. Mag. , vol. 58, no. 3, pp.62–68, Mar. 2020.[22] M. Hashemi, M. Coldrey, M. Johansson, and S. Petersson, “IntegratedAccess and Backhaul in Fixed Wireless Access Systems,” in
Proc. VTC-Fall’2017 , Toronto, Canada, Sept. 2017, pp. 1–5.[23] N. Tafintsev, D. Moltchanov, M. Gerasimenko, M. Gapeyenko, J. Zhu,S. Yeh, N. Himayat, S. Andreev, Y. Koucheryavy, and M. Valkama,“Aerial Access and Backhaul in mmWave B5G Systems: PerformanceDynamics and Optimization,”
IEEE Commun. Mag. , vol. 58, no. 2, pp.93–99, Feb. 2020.[24] O. Teyeb, A. Muhammad, G. Mildh, E. Dahlman, F. Barac, andB. Makki, “Integrated Access Backhauled Networks,” in
Proc.VTC2019-Fall’2019 , Honolulu, HI, USA, Sept. 2019, pp. 1–5.[25] C. Madapatha, B. Makki, C. Fang, O. Teyeb, E. Dahlman, M. Alouini,and T. Svensson, “On integrated access and backhaul networks: Currentstatus and potentials,”
IEEE Open Journal of the CommunicationsSociety , pp. 1–1, Sept. 2020.[26] F. G´omez-Cuba and M. Zorzi, “Twice simulated annealing resourceallocation for mmwave multi-hop networks with interference.” in
Proc.ICC’2020 , Dublin, Ireland, 2020, pp. 1–7.[27] A. Fouda, A. S. Ibrahim, I. Guvenc, and M. Ghosh, “UAV-Based In-Band Integrated Access and Backhaul for 5G Communications,” in
Proc.IEEE VTC-Fall’2018 , IL, USA, Aug. 2018, pp. 1–5.[28] K. N. R. S. V. Prasad, H. K. Rath, and A. Simha, “Wireless mobilenetwork planning and optimization,” in
Proc. COMSNETS’2014 , 2014,pp. 1–4.[29] A. Al-Dulaimi, S. Al-Rubaye, J. Cosmas, and A. Anpalagan, “Planningof ultra-dense wireless networks,”
IEEE Network , vol. 31, no. 2, pp.90–96, March/April 2017.[30] S. M. Azimi-Abarghouyi, B. Makki, M. Haenggi, M. Nasiri-Kenari, andT. Svensson, “Coverage analysis of finite cellular networks: A stochasticgeometry approach,” in
Proc. IEEE IWCIT’2018 , Tehran, Iran, Apr.2018, pp. 1–5.[31] A. HasanzadeZonuzy, I. Hou, and S. Shakkottai, “Broadcasting real-time flows in integrated backhaul and access 5g networks,” in
Proc.WiOPT’2019 , Avignon, France, 2019, pp. 1–8.
XXXXX < [32] M. Polese, M. Giordani, A. Roy, D. Castor, and M. Zorzi, “Dis-tributed Path Selection Strategies for Integrated Access and Backhaulat mmWaves,” in Proc. IEEE GLOBECOM’2018 , Abu Dhabi, UnitedArab Emirates, Dec. 2018, pp. 1–7.[33] B. Zhai, M. Yu, A. Tang, and X. Wang, “Mesh architecture for efficientintegrated access and backhaul networking,” in
Proc. IEEE WCNC’2020 ,Seoul, Korea (South), May 2020, pp. 1–6.[34] J. Y. Lai, W. Wu, and Y. T. Su, “Resource allocation and node placementin multi-hop heterogeneous integrated-access-and-backhaul networks,”
IEEE Access , vol. 8, pp. 122 937–122 958, July 2020.[35] N. Chen, T. Qiu, X. Zhou, K. Li, and M. Atiquzzaman, “An intelligentrobust networking mechanism for the internet of things,”
IEEE Commun.Mag. , vol. 57, no. 11, pp. 91–95, Nov. 2019.[36] X. Meng, H. Inaltekin, and B. Krongold, “Deep reinforcement learning-based topology optimization for self-organized wireless sensor net-works,” in
Proc. GLOBECOM’2019 . HI, USA: IEEE, 2019, pp. 1–6.[37] X. Fu, F. R. Yu, J. Wang, Q. Qi, and J. Liao, “Dynamic service functionchain embedding for nfv-enabled iot: A deep reinforcement learningapproach,”
IEEE Trans. Wirel. Commun. , vol. 19, no. 1, pp. 507–519,Oct. 2019.[38] P. Sun, Y. Hu, J. Lan, L. Tian, and M. Chen, “Tide: Time-relevant deepreinforcement learning for routing optimization,”
Future GenerationComputer Systems , vol. 99, pp. 401–409, Oct. 2019.[39] M. Wang, Y. Cui, S. Xiao, X. Wang, D. Yang, K. Chen, and J. Zhu,“Neural network meets dcn: Traffic-driven topology adaptation withdeep learning,”
Proceedings of the ACM on Measurement and Analysisof Computing Systems , vol. 2, no. 2, pp. 1–25, Jun. 2018.[40] S. Zhang, B. Yin, and Y. Cheng, “Topology aware deep learning forwireless network optimization,” arXiv preprint arXiv:1912.08336 , 2019.[41] P. V. Klaine, M. A. Imran, O. Onireti, and R. D. Souza, “A survey of ma-chine learning techniques applied to self-organizing cellular networks,”
Commun. Surv. Tutor. , vol. 19, no. 4, pp. 2392–2431, July 2017.[42] Q. Mao, F. Hu, and Q. Hao, “Deep learning for intelligent wirelessnetworks: A comprehensive survey,”
Commun. Surv. Tutor. , vol. 20,no. 4, pp. 2595–2621, Jun. 2018.[43] S. Troia, R. Alvizu, and G. Maier, “Reinforcement learning for servicefunction chain reconfiguration in nfv-sdn metro-core optical networks,”
IEEE Access , vol. 7, pp. 167 944–167 957, Nov. 2019.[44] S. Wang and T. Lv, “Deep reinforcement learning for demand-awarejoint vnf placement-and-routing,” in
Proc. GC Wkshps’2019 , HI, USA,2019, pp. 1–6.[45] X. Du, H. Van Nguyen, C. Jiang, Y. Li, F. R. Yu, and Z. Han, “Virtualrelay selection in lte-v: A deep reinforcement learning approach toheterogeneous data,”
IEEE Access , May 2020.[46] S. Yan, X. Zhang, H. Xiang, and W. Wu, “Joint access mode selectionand spectrum allocation for fog computing based vehicular networks,”
IEEE Access , vol. 7, pp. 17 725–17 735, Jan. 2019.[47] Y. Chang, X. Yuan, B. Li, D. Niyato, and N. Al-Dhahir, “Machine-learning-based parallel genetic algorithms for multi-objective optimiza-tion in ultra-reliable low-latency wsns,”
IEEE Access , vol. 7, pp. 4913–4926, Dec. 2018.[48] Y. Xu, W. Xu, Z. Wang, J. Lin, and S. Cui, “Load balancing forultradense networks: A deep reinforcement learning-based approach,”
IEEE Internet Things J. , vol. 6, no. 6, pp. 9399–9412, Aug. 2019.[49] X. Chen, R. Proietti, H. Lu, A. Castro, and S. J. B. Yoo, “Knowledge-based autonomous service provisioning in multi-domain elastic opticalnetworks,”
IEEE Commun. Mag. , vol. 56, no. 8, pp. 152–158, Aug.2018.[50] 3GPP RP-201756, “Revised WID: Integrated Access and Backhaul forNR,” 3rd Generation Partnership Project (3GPP), Tech. Rep. MeetingRAN arXiv , 2018.[Online]. Available: https://arxiv.org/abs/1805.01040[53] F. Gomez-Cuba and M. Zorzi, “Optimal link scheduling in millimeterwave multi-hop networks with space division multiple access,” in
Proc.IEEE ITA’2016 , La Jolla, CA, Feb. 2016, pp. 1–9.[54] J. Y. Lai, W. Wu, and Y. T. Su, “Resource allocation and node placementin multi-hop heterogeneous integrated-access-and-backhaul networks,”vol. 8, Jul. 2020, pp. 122 937–122 958.[55] B. Błaszczyszyn, “Lecture Notes on Random Geometric Models—Random Graphs, Point Processes and Stochastic Geometry,” 2017.[Online]. Available: https://hal.inria.fr/cel-01654766/document [56] S. M. Azimi-Abarghouyi, B. Makki, M. Nasiri-Kenari, and T. Svensson,“Stochastic Geometry Modeling and Analysis of Finite Millimeter WaveWireless Networks,”
IEEE Trans. Veh. Technol. , vol. 68, no. 2, pp. 1378–1393, Feb. 2019.[57] T. S. Rappaport, Y. Xing, G. R. MacCartney, A. F. Molisch, E. Mellios,and J. Zhang, “Overview of Millimeter Wave Communications for Fifth-Generation (5G) Wireless Networks—With a Focus on PropagationModels,”
IEEE Trans. Antennas Propag. , vol. 65, no. 12, pp. 6213–6230, Dec. 2017.[58] P. J. Diggle,
Some statistical aspects of spatial distribution models forplants and trees , 1982, no. 162.[59] M. A. Abu-Rgheff,
5G Physical Layer Technologies . John Wiley &Sons, Nov. 2019, pp. 303-313.[60] B. Makki, A. Ide, T. Svensson, T. Eriksson, and M. Alouini, “A geneticalgorithm-based antenna selection approach for large-but-finite mimonetworks,”
IEEE Trans. Veh. Technol.
Charitha Madapatha is currently pursuing thePh.D. in Communication Systems and InformationTheory from Chalmers University of Technology,Gothenburg, Sweden. He received his MSc. degreein Communication Engineering from Chalmers in2020 and the BSc. degree in TelecommuncationsEngineering from Asian Institute of Technology,Pathumthani, Thailand, in 2016.Charitha is the recipient of Swedish InstituteScholarship for Global Professionals, Sweden, 2018and the AIT Fellowship grant, Thailand, 2014-2016.He was also involved in LTE network planning and optimization in Thai-land and Cambodia. His current research interests include integrated accessand backhaul, multi antenna systems, mmWave communication, analysis ofphysical layer algorithms, resource allocation. He has co-authored severalinternational scientific publications in the field of wireless networks.
Behrooz Makki [M’19, SM’19] received hisPhD degree in Communication Engineering fromChalmers University of Technology, Gothenburg,Sweden. In 2013-2017, he was a Postdoc researcherat Chalmers University. Currently, he works as Se-nior Researcher in Ericsson Research, Gothenburg,Sweden.Behrooz is the recipient of the VR Research Linkgrant, Sweden, 2014, the Ericsson’s Research grant,Sweden, 2013, 2014 and 2015, the ICT SEED grant,Sweden, 2017, as well as the Wallenbergs researchgrant, Sweden, 2018. Also, Behrooz is the recipient of the IEEE best revieweraward, IEEE Transactions on Wireless Communications, 2018. Currently,he works as an Editor in IEEE Wireless Communications Letters, IEEECommunications Letters, the journal of Communications and InformationNetworks, as well as the Associate Editor in Frontiers in Communications andNetworks. He was a member of European Commission projects “mm-Wavebased Mobile Radio Access Network for 5G Integrated Communications”and “ARTIST4G” as well as various national and international researchcollaborations. His current research interests include integrated access andbackhaul, hybrid automatic repeat request, Green communications, millime-ter wave communications, free-space optical communication, NOMA, finiteblock-length analysis and backhauling. He has co-authored 63 journal papers,46 conference papers and 60 patent applications.
XXXXX < Ajmal Muhammad received his PhD degree inInformation Coding from Link¨oping University,Link¨oping, Sweden in 2015. During 2015-2017, heworked as a PostDoc researcher at ONLab, KTH-Royal Institute of Technology, Stockholm, Sweden.Currently, he is working as Experienced Researcherat Ericsson Research, Kista. Ajmal has co-authored41 research publications including journal and con-ference papers, and 30 patent applications.
Dr. Erik Dahlman is currently Senior Expert in Ra-dio Access Technologies within Ericsson Research.He has been deeply involved in the developmentof all 3GPP wireless access technologies, from theearly 3G technologies (WCDMA(HSPA), via 4GLTE, and most recently the 5G NR technology. Hecurrent work is primarily focusing on the evolutionof 5G as well as technologies applicable to futurebeyond 5G wireless access.Erik Dahlman is the co-author of the books 3GEvolution – HSPA and LTE for Mobile Broadband,4G – LTE and LTE-Advanced for mobile broadband, 4G – LTE-AdvancedPro and The Road to 5G and, most recently, 5G NR – The Next GenerationWireless Access Technology.In 2009, Erik Dahlman received the Major Technical Award, an awardhanded out by the Swedish Government, for his contributions to the technicaland commercial success of the 3G HSPA radio-access technology. In 2010, hewas part of the Ericsson team receiving the LTE Award for “Best Contributionto LTE Standards”, handed out at the LTE World Summit. In 2014 he wasnominated for the European Inventor Award, the most prestigious inventoraward in Europe, for contributions to the development of 4G LTE.
Mohamed-Slim Alouini (S’94, M’98, SM’03, F’09)was born in Tunis, Tunisia. He received the Ph.D.degree in Electrical Engineering from the CaliforniaInstitute of Technology (Caltech), Pasadena, CA,USA, in 1998. He served as a faculty memberin the University of Minnesota, Minneapolis, MN,USA, then in the Texas A & M University at Qatar,Education City, Doha, Qatar before joining KingAbdullah University of Science and Technology(KAUST), Thuwal, Makkah Province, Saudi Arabiaas a Professor of Electrical Engineering in 2009.His current research interests include the modeling, design, and performanceanalysis of wireless communication systems.