A Survey on Integrated Access and Backhaul Networks
AA Survey on Integrated Access and BackhaulNetworks
Yongqiang Zhang, Mustafa A. Kishk, and Mohamed-Slim Alouini
Abstract
Benefiting from the usage of the high-frequency band, utilizing part of the large available bandwidthfor wireless backhauling is feasible without considerable performance sacrifice. In this context, integratedaccess and backhaul (IAB) was proposed by 3GPP to reduce the fiber optics deployment cost of 5G andbeyond networks. In this paper, we first give a brief introduction of IAB based on the 3GPP release.After that, we survey existing research on IAB networks, the integrations of IAB to cache-enablednetwork, optical communication transport network, and the non-terrestrial network. Finally, we discussthe challenges and opportunities that might arise while developing and commercializing IAB networks.
Index Terms
Integrated access and backhaul; millimeter wave (mmWave) communication; UAV-assisted commu-nication; satellite-terrestrial communication; 3GPP; 5G NR.
I. I
NTRODUCTION
Due to the dramatic growth in both the end-users population (e.g., smartphones and tablets) anddemand for information service (e.g., video streaming and cloud computing), global mobile datatraffic has skyrocketed in recent years[1]. By 2030, the global mobile traffic volume is predictedto increase 670 times compared with it in 2010 [2]. In order to deal with such exponential trafficgrowth, the density of base stations (BSs) is expected to be substantially higher in the future. Thefeasibility of denser BSs deployments (a.k.a., Network densification) was discussed in [3] withthe anticipated requirement of coping with the increasing traffic growth. As a promising approach
The authors are with Computer, Electrical and Mathematical Science and Engineering Division, King Abdullah University ofScience and Technology (KAUST), Thuwal 23955-6900, Kingdom of Saudi Arabia. Email: { yongqiang.zhang.2, mustafa.kishk,slim.alouini } @kaust.edu.sa. a r X i v : . [ c s . N I] J a n o extend cell area and meet the high capacity demand, network densification is aiming to providea reliable access channel by reducing the distance from mobile users to BSs and increasing thespectrum reuse [4].However, in conventional network densification, one of the major drawbacks is the capi-tal/operational cost of the optical fiber deployment for BSs. For instance, one-meter optical fiberdeployment is estimated to cost approximately ∼ USD in the downtown area, nearly85% of the total expense results from the trenching and installing operations [5]. In this context,there is an attractive implementation solution by using wireless backhaul instead of conventionalwired-optical-fiber backhaul. Compared with wired-fiber backhaul, wireless backhaul can notonly provide almost the same transmitting rate as optical fiber but also bring with considerablecost decline and more flexible/timely deployment (e.g., no intrusion) [6], [7].On the other hand, one of the objectives of 6G communications is to achieve worldwideconnectivity [8]. To this end, there is an emerging need for providing reliable communicationsservice in remote and rural areas. For the same reasons, it might be economically appealing touse wireless backhaul rather than wired-optical-fiber backhaul to provide connectivity in suchunderserved areas.In Long-Term Evolution (LTE) Rel-10, 3GPP first introduced the study item involved wirelessbackhaul, known as LTE relaying [9]. However, since spectrum resource in LTE is too valuableto be used for backhauling, it gained little commercial interest. As several nationwide 5Gmobile networks have been already launched, many other countries authorities have announcedplans to grante 5G licenses for commercial use. One of the key features of 5G is the usageof high frequencies transmitting carries (e.g., millimeter wave), with the aim to facilitate theprospective larger spectrum. The large available spectrum of the these bands empowers a dramaticimprovement of transmission date rates, which support the carrier frequencies up to 52.6 GHzin millimeter wave (mmWave). Consequently, limited by physical properties, high-frequencycarriers lead to limited coverage area and need higher density of BSs deployment.Benefiting from the wide bandwidth in 5G new radio (NR), operator is capable of partitioningthe total radio resource into two parts for wireless backhauling and access, respectively. Thistechnique is known as integrated access and backhaul (IAB) networks, which has recentlyattracted more research attention [6], [10], [11]. For 5G NR, IAB has already been standardizedand recognized as a cost-effective alternative to the wired backhauling [12]. Compared with LTErelaying, IAB NR has more potential to receive considerable industrial attention. For example, ection II • Introduction to IAB • NR Architecture • Key Design Objective and Features • Field Trial Experiments
Section III • OBHD Mode SG-based Analysis • IBHD Mode SG-based Analysis
Section IV • Interference Mitigation • Energy Consumption Minimization • System Throughput Maximization • Utility Optimization • Spectral or Energy Efficiency Maximization
Section IX-X • Challenges and Opportunities • Conclusions
Section V • Deployment Design • Routing • Fairness • Incentive Mechanism Design • Network Coding-Aware
Section VI-VIII • Cache-enabled IAB Network • Optical IAB Network • Non-Terrestrial IAB Network
Fig. 1: Organization of the survey.in case of mmWave network, the shortage of narrow coverage will result in an urgent needto increase the density of BSs deployment. Consequently, there is a much higher demand forwireless backhauling. Meanwhile, by exploiting the much larger bandwidth, it is less expensivefor operator to perform self-backhauling. Besides, a large number of antennas can easily beemployed for mmWave-enabled BSs due to their small wavelength, which can enhance thesignal directional gain and link reliability for backhauling.As shown in Fig. 1, the rest of the paper is organized as follows. In Section II, we give a briefintroduction of IAB architecture. The stochastic geometry-based analysis for IAB network is dis-cussed in Section III. Section IV and V focus on the resource allocation and scheduling researchin IAB network. Studies on the integration of IAB network with cache-enable network, opticalcommunication, and non-terrestrial communication are surveyed in Section VI. Section VII, andSection VIII, respectively. The challenges and opportunities of IAB are discussed in Section IX.The paper is concluded in Section X.II. I
NTEGRATED A CCESS AND B ACKHAUL T ECHNIQUE
Over the course of the last 20 years, wireless backhauling in wireless networks has beenstudied extensively [13]. However, there is no tight integration of access and backhaul in thestudies of LTE backhaul network, since it specifies that only one single hop with fixed parent BSis supportable for LTE relay and strict resources partitioning between access and backhaul [14].In contrast, IAB network is a more flexible deployment solution without the heavy overhead of
AB-donor IAB-nodeCore Network Internet Mobile User
Fig. 2: IAB system architecture.wired-fiber installation[15]. Fig. 2 demonstrates a simple example for an IAB network. Basedon a wired connection to the core network IAB-donor can provide access communication tomobile users and wireless backhaul to IAB-nodes, respectively. IAB-node is able to providenetwork access service to mobile user and backhaul the access traffic wirelessly. Therefore,IAB-node can be regarded as a wireless relay to extend the coverage of an IAB-donor. Thisfunctionality is helpful for network to achieve a robust coverage performance when Line-of-sight (LoS) propagation is blocked by environmental obstacles (e.g., Buildings).
A. NR IAB Architecture
With the objective to cope with the increasing demand for backhaul, 3GPP first proposed astudy item on IAB in [16]. The physical-layer specification of IAB were completed at 2019, andhigher-layer protocols and architecture was completed in 3GPP Rel-16 at July 2020 [12]. Furtherenhancements (e.g., mobile IAB) has been carried out in 3GPP Rel-17, which is expected to befrozen in December 2020. We will give a brief overview of the IAB NR in the following basedon 3GPP specifications.Fig. 3 and 4 illustrate the IAB protocol stack for user plane and control plane, receptively. Theoverall architecture is based on the functionality split. In particular, the IAB-donor consists of acentral unit (CU) and no less than one distributed unit (DU). The DU includes the Radio Link
LCBAP
IAB-node 2 IAB-node 1 IAB-donor-DU
MACRLCBAPMACUDP
BH RLC channel
GTPIP
BH RLC channel
PHYPHY RLCBAPMACPHY IPIAB-MTIAB-DU IAB-MTIAB-DU
F1-U
UDPGTPIPIP
IAB-donor-CU-UP
RLC UE MACSDAPPHYPDCP SDAPPDCPRLCBAPMACRLCBAPMAC PHYPHY
NR-Uu
Fig. 3: IAB user plane protocol stack.
RLCBAP
IAB-node 2 IAB-node 1 IAB-donor-DU
MACRLCBAPMACSCTP
BH RLC channel
F1-APIP
BH RLC channel
PHYPHY RLCBAPMACPHY IPIAB-MTIAB-DU IAB-MTIAB-DU
F1-C
SCTPF1-APIPIP
IAB-donor-CU-UP
RLC UE MACRRCPHYPDCP RRCPDCPRLCBAPMACRLCBAPMAC PHYPHY
NR-Uu
Fig. 4: IAB control plane protocol stack.Control (RLC), Medium Access Control (MAC), and Physical layer (PHY) protocols. Apart fromPacket Data Convergence Protocol (PDCP), the CU includes Service Data Adaptation Protocol(SDAP) or Radio Resource Control (RRC) protocol in control plane or user plane, respectively.The interface between CU and DU is standardized as F1 interface, which defines the higherlayer protocols. The IAB-donor connect the core network via non-IAB backhaul, and uses theIAB-donor-DU to serve UEs and connected IAB-nodes wirelessly.The IAB-node consists of mobile termination (MT) functionality and DU functionality. TheAB-nodes rely on IAB for backhauling, and also provide service for UEs and IAB-nodes via theDU functionality. The MTs act as normal devices and associate with the DU of parent IAB-nodeor IAB-donor. The message transmission is based on the lower layer functionality provided bythe link between IAB-node MT and its parent node DU. Besides, IAB-node can be backhauled tothe IAB-donor through more than one intermediate IAB-nodes, which means that the multi-hopbackhauling is supportable in IAB network.The lower three protocol stacks up to RLC, are referred to as NR-Uu interface. The middlethree layers between Backhaul Adaptation Protocol (BAP) and PDCP provide the F1 interfacefor user plane (F1-U) and control plane (F1-C), respectively.The BAP is a novel protocol defined by IAB, and responsible for routing information packetsfrom IAB-donor to target IAB-node and vice versa. A typical IAB-node has its own BAPaddress. For downlink (DL) scenario, the BAP layer of IAB-donor first adds the BAP headerto the information packets. The BAP header consists of the routing ID for the destination BAPaddress and the path ID includes the path to the destination node. Also, there is a flag indicatorin the BAP header which is determined by the type of the packet (i.e., control plane or userplane).Once the typical IAB-node receives an information packet, the BAP layer will first check therouting ID in BAP header. If the IAB-node is the destination node, the information packet willbe forwarded to higher layers. For instance, this packet will be elevated to GTP-U or F1-APwhen the it is intended for UE served by the IAB-node or it is control plane packet for theIAB-node. Otherwise, the IAB-node will deliver the packet to its DU and transmit the packetto next node based on the routing table.
B. Key Design Objective and Features
The advantage of IAB is supporting flexible and high-density deployments of NR cells withoutrelying on costly wired backhaul network deployment. Also, IAB envisions a diverse range ofdeployment scenarios, including outdoor or indoor dense small-cell deployments and coverageextension.
1) Spectrum:
Benefiting from the large bandwidth and small wavelength, mmWave IABenables massive beamforming and using a portion of inexpensive bandwidth to do backhaul.Since high-band spectrum is generally organized as an unpaired spectrum, for operation purposes,3GPP proposed that IAB enables wireless in-band and out-of-band relaying.
In-band relaying: Both the access link and the backhaul link are transmitted at the samefrequency band simultaneously. • Out-of-band relaying: The access link and backhaul link transmissions are conducted inorthogonal channels. In other words, the access link transmission is conducted at certainfrequency band while the backhaul link utilizes the remaining frequency band.
2) Network Topology:
IAB enables multi-hop backhauling, which makes a flexible rangeextension. In Rel-16, directed acyclic graph (DAG) based and spanning tree (ST) based multi-hop topology is supported. Besides, network topology adaptation and redundant connectivity aresupported in IAB, which enables better backhaul performance and fast adaptation to deal withlink overloads or failures.
3) Radio Link:
With the sharing of resources between access and backhaul, it may result inlimitations on the end-user quality of service (QoS) (e.g. rate, latency). And the new interferenceresults from the aceess/backhaul-to-backhaul/access needed to mitigate. One of the targets inradio link design is to ensure the QoS requirements of users are fulfilled even in a multi-hopsetting. Another important issue is that the deployment of IAB network should be transparentto UEs (i.e. no additional requirement for UE features/standardization).
C. Field Trial Experiments
The advantage of mmWave IAB network was verified by a field experiment in [17]. Forinstance, in [17], coverage ratio and user throughput in an out-door IAB network with time-division multiplexing were investigated. The simulation results showed that coverage ratio canbe improved to about ∼ with only one IAB node deployment, and achieve maximumcoverage ratio gain up to compared to non-IAB-node deployment. The system throughputimproved significantly by IAB-node deployment, compared with non-IAB-node deployment, upto . throughput gain is obtained.TABLE I: Taxonomy of Literature in Section III. Objective Network Reference
Coverage Probability Multi macro-cells;TDD; mmWave [18]
Continued on next page
ABLE I –
Continued from previous page
Objective Network Fourth Reference hybrid UHF & mmWave;OBHD [19]Multi macro-cells;IBFD [20]Single macro-cell;OBHD; mmWave [21]Multi macro-cells;OBHD; mmWave [22]massive MIMO; OBFD [23]Ergodic Capacity Multi macro-cells;OBHD [24]
III. S
TOCHASTIC G EOMETRY - BASED A NALYSIS OF
IAB N
ETWORKS
Stochastic geometry (SG) has been regarded as a powerful mathematical tool for modeling,analyzing, and designing wireless network [25]. In this section, we will introduce some paperswith the integration of SG and IAB network.
A. Out-of-band HD (OBHD) Mode
In [19], Singh et al. developed a tractable mmWave IAB network analysis model, with theobjective to characterize the rate distribution. Moreover, the authors considered a more generalsetup with the mmWave IAB network co-existing with conventional ultra-high frequency (UHF)cellular network. The proposed analytical framework was verified by two realistic environmenttopologies. It was shown that the density of BSs can improve the rate coverage probabilitydrastically. Compared with the increasing of BSs density, the improvement of bandwidth has aless positive effect on the rate of cell edge mobile users.Due to the fact that the sharing of bandwidth between access and backhaul link in IAB-donormode, the allocation of bandwidth plays an important role in the performance of network. In [21],he authors studied a setup that assumes a single macro-cell consisting of one anchored BS(i.e., IAB-donor) surrounded by small-cell BSs (i.e., IAB-nodes) while the IAB-nodes employedOBHD mode. By using SG tools to capture the locations of small-cell BSs (SBSs) and mobileusers as well as the load of BSs, Saha et al . presented an analytical framework to characterize therate coverage probability accurately. For instance, the authors first assumed the bandwith of IAB-donor is splitted into two parts: access bandwidth and backhaul link bandwidth. Further, Saha et al . investigated the impact of three different backhaul link bandwidth allocation strategies: 1)equal partition: all IAB-nodes share the bandwidth equally; 2) instantaneous load-based partition:the backhaul link bandwidth allocated to an IAB-node is proportional to its instantaneous load;and 3) average load-based partition: the backhaul link bandwidth allocated to an IAB-node isproportional to its average load. The simulation results showed that the coverage probability w . r . t the bandwidth splitting ratio at the IAB-donor behaves like a concave function (i.e., thereexists only one optimal splitting ratio in all considered strategies). Moreover, in terms of bothcoverage probability and the median rate, the performance of three strategies can be sorted in adescending order as: instantaneous load-based > average load-based > equal.In the follow-up work, Saha et al. considered a more general model setting which consists ofmulti macro-cells [22], the locations of the fiber-wired macro-cell BSs (MBSs) were modeled bya Poission point process. The authors proposed two bandwidth allocation strategies: 1) IntegratedResource Allocation (IRA): the bandwidth allocated equally for users; 2) Orthogonal ResourceAllocation (ORA): the MBS reserved a fixed fraction of bandwidth to allocate to its directlyserved users, and allocate the rest of the bandwidth for the backhaul link to SBSs proportional tothe load at each small-cell. In terms of rate coverage probability, the simulation results showedthat the IRA outperforms ORA, and the improvement increases with increasing the density ofSBSs. The reason behind these performance trends is that the fixed fraction reserved bandwidthbackhaul link cannot deal with the increasing backhaul load results from increasing density ofIAB-nodes.A semi-closed-form expression for the ergodic throughput of the two-tier IAB network waspresented in [24]. Based on the proposed analytical model, the authors formulated and solvedan optimization problem in order to maximize the number of offloaded mobile users, whilesatisfying the requirement for SINR. Numerical results showed that the proposed method canachieve two times of improvement for the average user and SBS throughput. The joint uplink(UL) and DL coverage probability under different resource allocation schemes in an mmWave-nabled IAB network was studied in [18]. In particular, the authors investigated the impact ofstatic and dynamic time-division duplexing (TDD) scheduling policies at the access link, as wellas the synchronized or unsynchronized time alignment at the backhaul link. Simulation resultsrevealed that the combination of dynamic TDD and unsynchronized time alignment can achievethe best performance in terms of average DL and UL transmitting rate. B. In-band FD (IBFD) Mode
IBFD IAB-node is the network framework in which IAB-nodes can use the same timeand frequency resource block to conduct reception and transmission. Therefore, there is noneed to consider the bandwidth allocation in the IAB-node operation. However, this mode willintroduce the undesired interfering transmitted signal at IAB-node’s MT from the transmissionfrom another IAB-node’s DU to the users, which is known as self-interference (SI). Since thedistance of the interference source (DU) from the MT is much larger than the distance betweenthe desired signal’s source and the MT, SI will be enough powerful interference which leads tocritical performance loss. Although there are numerous state-of-art self-interference cancellationtechnologies, residual self-interference (RSI) always exists [26]–[29].An analytical expression for coverage probability in a massive multiple-input multiple-output(MIMO) wireless self-backhaul networks was derived in [23] by using SG tools. The consideredtwo-tier network consists of a mixture of operating small-cells, each small-cell is assumed toadopt in-band or out-of-band backhaul mode with a certain probability. Both the existence ofself-interference and cross-tier interference were considered for the signal-to-interference ratio(SIR) at mobile users. Simulation results revealed that implementing the considered mixturedeployment consideration outperforms adopting only either in-band or out-of-band backhaulingSBS option.In [20], the authors developed a framework to analyze the coverage probability and averageDL transmission rate for a two-tier IAB network, where the MBS (i.e., IAB-donor) is OBFD-enabled, and SBSs are IBFD-enabled. The trade-off between increasing interference and spectrumefficiency was investigated by the end-to-end analysis of the backhaul link and access link jointly.Numerical results showed that the DL rate in the considered system setting could achieve nearlytwo times gain compared to a traditional TDD or frequency-division duplexing (FDD) self-backhauling network. Meanwhile, due to the higher interference in IBFD mode, the coveragerobability in the considered network is approaching half of its value in conventional TDD orFDD operation. TABLE II: Taxonomy of Literature in Section IV.
Objective Methodology Network Reference
InterferenceMitigation Random matrix massive MIMO;IBFD [30]Heuristic method MIMO;IBFD [31]Random matrix;Heuristic approach mmWave;meshbackhauling [32]Convex optimization MIMO; TDD [33]EnergyConsumptionMinimization Heuristic method DRX [34]Heuristic method mmWave;multi-hopbackhaul [35]Convex optimization massive MIMOIBFD [36]Alternative optimization NOMA; IBFD [37]ThroughputMaximization DRL OFDMA [38]Heuristic method TDD [39]Water-filling algorithm IBFD [40]Random matrix MIMO;IBFD [41]SCP mmWave;IBFD [42]Convex optimization;bisection search hybrid IBFD& OBHD [43]
Continued on next page
ABLE II –
Continued from previous page
Objective Methodology Network Fourth Reference
Game theory mmWave [44]Game theory OBFD [45]Game theory;DRL IBFD [46]Random matrix;Alternative optimization mmWave;hybridbeamforming [47]Random Matrix HybridmmWave& Sub-6 GHz [48]Markov approximation OBFD [49]CCP NOMA [50]Riemannian optimization IBFD [51]SCP IBFD [52]UtilityOptimization Lyapunov optimization mmWave [53]RL ;Lyapunov optimization mmWave;multi-hopbackhual [54]Alternative optimization Network [55]Lyapunov optimization virtualization [56]EEMaximization CCP massive MIMO;EH; IBFD [57]SEMaximization SQP OBFD;Self-organizingnetwork [58]SCP IBFD;massive MIMO [59]
V. R
ESOURCE A LLOCATION IN
IAB N
ETWORKS
A. Interference Mitigation
With the increasing mobile traffic and the existence of wireless backhaul hop, the number oftransmission links in the IAB network is much larger than the past wireless networks that wouldintroduce severe inter or intra interference at access and backhaul link. Thus, the performance ofinterference management can be an important issue in IAB network design. In [30], the authorsstudied the average sum rate performance for three different duplex schemes in a massive MIMO-enabled IAB network. For instance, a TDD based duplex, an IBFD mode, and an IBFD modewith interference rejection (IBFD-IR) were considered. Based on zero-forcing (ZF) schemes, theauthors formulated the beamforming matrix for UL and DL transmission in all the consideredschemes. The simulation results showed that IBFD and IBFD-IR schemes outperform TDDschemes when the distance between IAB-donor and IAB-nodes exceeds some certain levels.The performance for IBFD-IR is better than IBFD but the performance gap decreases as thedistance between IAB-donor and IAB-nodes increases.Different from the case in [30], the authors in [31] investigated the incorporation of beam-forming into interference mitigation based on the usage of limited-feedback information. Inparticular, for the DL transmission in an out-band IAB network, two different schemes based onantenna selection (AS) or quantized phase information (QCP) were proposed. The correspond-ing beamforming and interference mitigation weight vector for AS and QCP were determinedby the received signal power and SNR, respectively. System-level simulation results showedthat the combination of beamforming and interference mitigation can achieve nearly 9% and6.4% compared with corresponding purely beamforming schemes. The performance of QCP-based beamforming and interference mitigation was found to be the best among the consideredtechniques.In [32], a comprehensive mechanism aiming at managing the out-band backhaul intra-channelinterference for a mesh-architecture mmWave IAB network was proposed. The proposed mech-anism consists of node placement, sub-channel alignment, and transmitting power control. Bysolving a SIR maximization problem, the authors derived the optimal power allocation policy.It was shown that the proposed method can reduce interference up to a distance of 250 m awayfrom the wired-backhaul node.Flexible TDD based resource scheduling is able to manage the frequency-time resources morefficiently. In [33], an iterative beamformer design for a TDD based IAB network with oneIAB-donor and multiple IAB-nodes was proposed. For a given time slot, IAB-donor and IAB-nodes operate in different UL/DL modes. By considering the traffic dynamics at each nodeand assuming each user and access point (AP) has a specific UL/DL queue, Jayasinghe et al. formulated and solved an optimization problem with the objective to minimize the weighted l p -norm queue minimization of the UL and DL users during two time slots. By comparing theproposed method with the half-duplex (HD) IAB system, the results showed that the performanceof the proposed scheme is better for all considered traffic arrival rates. B. Energy Consumption Minimization
In [34], the authors proposed a coordinated mechanism to minimize network power consump-tion in an IAB network. By utilizing the information of QoS in the whole network, the controllercascaded with MBS can determine and inform which BSs or mobile users should enter to sleepmode along with the possible sleep time by discontinuous reception configuration (DRX) . Bysimulation based on realistic network parameters, the presented mechanism is able to achieveup to 50% power saving in comparison with LTE. The problem to find optimal time allocationand power control for the mmWave IAB network was addressed in [35]. The authors considereda multi-hop transmission system model, with one MBS and mesh connected SBSs. Each hopequipped with a status indicator to guarantee the hop transmitting priority. By formulating anoptimization problem with the objective to minimize the overall energy consumption underthroughput constraints, Meng et al. first established space/time-division multiple access groups,and then proposed a heuristic algorithm to derive the time allocation and power control policies.The proposed scheme was shown to be achieving a considerable performance boost comparedwith the STDMA scheme [60] and the MQIS scheme [61].The energy consumption minimization problems in full-duplex (FD) IAB network was studiedin [36] and [37]. Korpi et al. proposed a resource allocation scheme to minimize the energyconsumption of a two-tier massive MIMO-enabled IAB network under QoS constraint in [36].In addition to the FD IAB scheme, a closed-form expression for the optimal transmitting powerfor HD and hybrid FD schemes were derived. In order to compare the performance of a threeconsidered schemes, the authors further derived feasibility boundaries for the DL and UL raterequirements in closed-form. It was shown that both FD and hybrid relay schemes are not ableto guarantee the QoS constraints even as the transmitting power approaches infinity. Numericalesults reveal that the FD schemes can achieve the lowest energy consumption among the threeconsidered schemes. With the consideration of using non-orthogonal multiple access (NOMA)to enhance the spectrum efficiency, Lei et al. presented an effective algorithm to minimizethe energy consumption for a two-tier NOMA-enabled FD IAB network based on fixed-pointiteration in [37]. Numerical results showed that the performance gap between the proposedmethod and the orthogonal multiple access (OMA) based networks increases as the transmissionrate requirements increases.
C. System Throughput Maximization
The rate of the mobile user associated with IAB-node is determined by the minimum ratebetween the backhaul link at this IAB-node and the access link, which is obviously sensitiveto the bandwidth partitioning strategies. The spectrum allocation can significantly influence theuser’s data rate, as analyzed in [22]. Lei et al. derived the optimal bandwidth allocation policyfor a two-tier IAB network, by solving a mixed-integer nonlinear programming problem withthe objective to maximize the sum log-rate, with the usage of deep reinforcement learning(DRL) to derive the optimal solution in [38]. Benefiting from the DRL, it is tractable to obtainthe optimal allocation policy for the large-scale time-varying IAB network. Compared with thefull-spectrum reuse strategy, simulation results showed that the proposed method can achieveconsiderable performance gain especially when the density of IAB networks is large. Withthe consideration of UL-DL and access-backhaul link transmission rate requirements, the jointUL and DL resource allocation mechanism for a TDD IAB-network was studied in [39]. Inthis mechanism, the optimization problem aims at maximizing the overall throughput and wasdecomposed into two subproblems on time slot orchestration and sub-band scheduling. Numericalresults showed that total system throughput improves significantly compared with conventionalRound Robin (RR) and Proportional Fair (PF) mechanisms.As for in-band IAB network, the trade-off between self-interference and spectrum efficiencywas investigated in [40]–[43]. The authors in [40] obtained the optimal power allocation policyby solving the DL sum rate maximization problem in an IBFD IAB network. Based on thestandard optimization method, a closed-form expression for optimal transmitting power wasderived in [41] by solving the optimization problem with the objective to maximize the networksum rate of DL and UL. The authors considered both the IAB-donor and IAB-node are workingin IBFD mode, where the mobile users adopt HD mode. Simulation results showed that theroposed solution can achieve the best performance when UL and DL rate requirements areclose. In [42], the optimal spectrum and power allocation policies were derived by sequentialconvex programming (SCP) method. The authors considered a mmWave in-band IAB networkwith one IAB-donor surrounded by multiple IAB-nodes, where partial users could reuse thebackhaul bandwidth resources. The formulated problem was under the heterogeneous user’s DLtransmission rate requirements. Compared with the simple power allocation optimization, theproposed algorithm can achieve a nearly 25 % gain in terms of system throughput. Similar to [36],the authors in [43] investigated the sum-rate maximization problem in IBFD, OBFD, and hybridIBFD/OBFD backhauling network. By using the convex optimization method and bisectionsearch algorithm, the authors derived the solutions for optimal spectrum allocation policiesfor three considered scenarios. With two heuristic methods based on received signal power,three distributed algorithms and coverage probability were derived for corresponding backhaulingschemes. It was shown that IBFD outperforms OBFD when self-interference cancellation exceedsa certain level. Moreover, IBFD tends to allocate more spectrum resources for the SBSs closeto MBS while OBFD is just the opposite.Using tools from game theory, the authors in [44] obtained the joint optimal power control andtransmission slot allocation policies for a two-tier mmWave IAB network. For the formulatednon-cooperative game with the objective of sum rate maximization, the authors first proved theexistence and feasibility of the Nash equilibrium then designed a centralized resource allocationalgorithm as well as a decentralized algorithm based on the functionality splitting architectureat IAB-nodes to solve it. Simulation results showed that the proposed algorithms can achieveat least 17.94% improvement compared with pure optimization for power control. The jointoptimal backhaul and access link resource allocation policies for a two-tier IAB network werestudied in [45]. In order to maximize the overall DL data rate, the problem was formulatedinto a Stackelberg game form. The wired-backhaul MBS plays the role of the leader, while theSBSs act as the followers. For instance, the optimal sub-carriers allocating policies for MBSand SBSs were derived from the leader’s and followers’ optimization problems, respectively.An optimal power control strategy was obtained from the follower’s problem. Numerical resultsproved that the presented algorithm can improve the performance up to 14.2% and 24.9 %. Byintegrating game theory tool and reinforcement learning (RL), the authors in [46] presented agame-theoretic leaning mechanism to maximize the UL sum rate in an in-band IAB network.The proposed mechanism is supportable for self-organizing, which means that the MBS enableshroughput balancing in a decentralized way. In comparison with non-IAB mechanism, theproposed mechanism can achieve up to 40% performance improvement.With the objective to maximize the weighted sum-rate in a mmWave IAB heterogeneousnetwork (HetNet), the joint optimal strategies for transmitting power allocation, time splittingratio, mobile users association, and beamforming design were proposed in [47]. Time splittingmanner is considered between backhaul and access links. The access links design includes themobile users association and beamforming design by using limited channel state information(CSI). By simulation, the performance of the proposed limited feedback hybrid beamformingscheme was shown to be able to approach the performance of the digital beamforming schemewith full CSI. The spatial multiplexing gain increases up to the number of radio frequency (RF)chains for IAB-node.The DL sum rate maximization problem in an enhanced hybrid IAB network was studiedin [48]. For instance, with the aim to extend the coverage of IAB-nodes, the authors firstconsidered that the access link is working in sub-6 Hz while the backhaul link is working inmmWave, then proposed a hybrid precoder to improve the backhaul transmission rate. Numericalresults revealed that the performance of the presented precoder is close to conventional blockdiagonalization precoder, with a significant decline for the required number of the RF chains.In [49], the authors considered a network with multi-antenna IAB-nodes, and the transmissionlink can be allocated with an orthogonal sub-channel when it suffers from severe interference. Byintroducing the penalty function and factors, Pu et al. first transformed the sum rate maximizationproblem into 0-1 integer programming without any inequality. Further, a resource allocationalgorithm based on Markov approximation with polynomial time complexity was proposed. Theperformance evaluation showed the sum rate for the proposed scheme increases faster than OBFDscheme as the number of antennas increases.In [50], the authors addressed the sum rate maximization problem in a NOMA-enabled out-band IAB network, where a typical mobile user is only assumed to be served by two SBSs.Two decoding strategies based on different SIC decoding order were considered at DL accessand backhaul links. By using convex-concave procedure (CCP) method, the authors derived theoptimal power allocation policies for UL and DL transmissions. It was shown that decoding betterchannel quality signals first outperforms the other decoding strategy. Compared with OMA, theperformance improvement of considered NOMA-enabled network increases as the number ofmobile users increases. With the consideration that mobile users can be served by a cluster ofBSs cooperatively, the authors in [51] addressed the joint beamforming and SBSs clusteringproblem for IBFD IAB network. In order to maximize the overall DL transmission rate, amanifold optimization problem was formulated under the transmitting power constraints. Byusing Riemannian optimization technique, a heuristic algorithm was presented to derive theoptimal policies for the SBSs clustering and beamforming vectors. Extensive simulations showedthat optimal cluster size is related to the transmitting power constraint for the IAB-donor. Theextension of [51] from full CSI to partial CSI is presented in [52]. To this end, a stochasticsuccessive lower-bound maximization algorithm and a deterministic algorithm with lower timecomplexity was proposed to solve the modified sum-rate maximization problem. Moreover, theperformance gap between [52] and [51] was shown to be decreasing as the cluster numberincreases.
D. Utility Optimization
The maximization of network utility with interference management in the DL transmission ofa mmWave IAB HetNet was considered in [53]. The authors developed a Lyapunov optimizationframework to decouple the primal optimization problem and used convex optimization andsuccessive convex approximation method to derive the optimal solution for user association,beamforming design, and power allocation, respectively. Simulation results revealed that theproposed scheme can achieve 5.6 times gain in cell-edge users throughput compared with un-optimized user allocation scheme with a density of SBSs equal to / Km . A more generalsetting for multi-hop multipath scheduling is contemplated in [54]. Vu et al. formulated a networkutility function under low latency and network stability in order to achieve Ultra-low latency andreliable communication (ULLRC). A RL-based algorithm and a Lyapunov optimization algorithmwere proposed to derive the optimal solution for path selection and rate allocation, respectively.Compared to [53], it was shown in the simulation that the proposed scheme can provide reliablecommunication with a guaranteed probability near to 100% and latency reduction at least 50%.Wireless network virtualization empowers the resources at network infrastructure provider(InP) to be sliced into several virtual parts, which can be shared by multiple service providers(SPs) [62]. In [55], the authors investigated the application of wireless network virtualizationtechnique into a FD IAB network. They assumed both MBS and SBS from multiple InPscan be virtualized and shared by mobile users from different mobile virtual network operators(MVNOs). To solve the utility maximization problem with the aim to maximize all the profitf MVNOs and cost of resource consumption, a two-stage iterative algorithm based on convexprogramming was proposed. Numerical results showed that the proposed algorithm can achieveconvergence within 15 iterations and significant average throughput improvement for small-cells. The authors in [56] addressed the time-average utility maximization for an IAB networkwith wireless network virtualization. In particular, they considered the requirements for networkqueue stability, average throughput of different SPs, and the backhaul link capacity. A Lyapunovoptimization-based algorithm was proposed to find the optimal strategy for bandwidth portioningratio, user association, and percentage of resources allocated by the associated BS. In comparisonwith the traditional networks without virtualization techniques, the proposed scheme can resultin at least 60% increment of average total utility. E. Spectral and Energy Efficiency Maximization
In [57], the authors considered integrating energy harvesting (EH) technique into an IBFD IABnetwork with the aim to improve energy efficiency (EE). The MBS is equipped with massiveMIMO antennas array, and the SBS is equipped with a multi-antenna array and can harvest energyfrom renewable resources. The authors first proposed a precoder to mitigate the self-interferenceand cross-tier interference, then used CCP method to obtain the optimal power control and mobileusers association policies by solving the energy efficiency maximization problem. Comparedwith the conventional OBFD network, it was found that the proposed method can dramaticallyimprove EE for dense SBSs deployment.The spectral efficiency maximization problem in the IAB network was studied in [58], [59]. Inparticular, the joint optimization of backhaul and access links with the aim to maximize spectralefficiency (SE) in OBHD IAB networks was studied in [58]. By using a sequential quadraticprogramming (SQP) algorithm, the optimal antenna tilt angle deployment policies were derived.The proposed schemes can achieve up to 50% performance improvement compared with fixed-tiltdeployment. With the consideration of IBFD mode, Chen et al. addressed the SE optimizationproblem which under DL backhaul capacity requirements in [59]. After the transformation ofthe primal optimization problem, the optimal power allocation policy was derived using SCP-based algorithm. IBFD IAB network was shown to be able to bring about 25% SE performanceimprovement. V. S
CHEDULING IN
IAB N
ETWORKS
ABLE III: Taxonomy of Literature in Section V.
Objective Methodology Network Reference
DeploymentDesign Heuristic method massive MIMO;ad-hoc [63]Convex optimization OFDMA;MIMO [64]Heuristic method meshmulti-hopbackhaul;mmWave [65]K-means clustering;Genetic algorithmalgorithm [66]branch-and-boundalgorithm multi-hopbackhaul;mmWave; [67]Genetic algorithm IBFD [68]Routing Virtual-networkmethod meshmulti-hopbackhaul;mmWave [69]Heuristic method meshmulti-hopin-bandbackhaul;mmWave [70]Convex optimization multi-hopbackhaul;mmWave [71]Heuristic method one-hopbackhaul;TDD;mmWave [72]
Continued on next page
ABLE III –
Continued from previous page
Objective Methodology Network Fourth Reference
Distributed greedyalgorithm multi-hopbackhaul;mmWave [73]Backpressure algorithm multi-hopbackhaul;mmWaveMU-MIMO [74]Simulated annealingalgorithm [75]DRL multi-hop;mmWave [76]RL; CGmethod [77]Semi-distributedlearning algorithm DynamicmmWavenetwork [78]Heuristic algorithm mmWave [79]Fairness Matching theory TDMA mmWave [80]Heuristic method HybridIBFD&OBHD [81]WPF algorithm DAGmulti-hopbackhaul [82]Backpressure algorithm IBFD [83]IncentiveMechanism Game theory Two-hopbackhaul [84]
Continued on next page
ABLE III –
Continued from previous page
Objective Methodology Network Fourth Reference
Game theory UPN;multi-hopbackhaul [85]NetworkCoding-AwareScheduling Convex optimization XOR networkcoding [86]Coding design Linear networkcoding;multi-hopbackhaul [87]
The flexible deployment of the IAB network brings an emerging demand for efficient schedul-ing to realize high throughput and low latency while guaranteeing users fairness. This sectionsurveys current research related to the scheduling in the IAB network.
A. Deployment Design
Due to the higher pathloss and penetration loss, the coverage area of the network operatedin high frequents is limited by both the number and locations of BSs. In [63], the authorsinvestigated the ad-hoc deployments in a massive MIMO-enabled IAB network, where theSBSs were positioned in proximity to mobile users. In [65], a self-optimizing deploymentframework for an IAB mesh network was developed. Based on the neighbor discovery, theproposed framework enables autonomous deployment for the newly added nodes. Taking intoaccount the height of nodes, the authors investigated the joint resource allocation and nodedeployment problem for a MIMO-orthogonal frequency-division multiple access (OFDMA)based IAB network in [64]. Since only IAB-donor is connected to the core network via wiredfiber in IAB network, the location of IAB-donor in a set of BSs can determine the quality in multi-hop wireless backhauling. A genetic algorithm (GA) in combination with the K-means clusteringmethod to select the location of IAB-donor to maximize the backhaul capacity was proposed in66]. Through extensive Monte Carlo simulations, the authors evaluated the performance of thisalgorithm in comparison with the conventional genetic algorithms and K-means clustering interms of the average number of hops and backhaul network capacity at different node densities.Simulation results showed that the proposed algorithm can achieve at least 20.2% and 19.8% aperformance improvement in the average number of hops and backhaul capacity, respectively.The trade-off between deployment cost and network performance brings more challengesin deployment design in IAB network. In [67], the authors formulated a mixed-integer linearprogram problem with the objective to minimize the cost of required fiber-wired BSs deploymentand used the branch-and-bound algorithm to obtain the optimal solutions of deployment, routing,and resource allocation on each flow in a single-tier IAB network. Simulation results showedthat IAB is able to reduce the amount of fiber-wired BSs deployment significantly. The BSsdeployment problem for a two-tier IBFD IAB network was studied in [68]. With the aim tominimize deploying cost and maximize coverage, the optimal number of IAB donors and IABnodes were derived from a non-dominated sorting genetic algorithm. Numerical results confirmedthe trade-off between BSs deployment costs and coverage. For realistic network topology, authorsshowed that the optimal deployment solution can be derived from the Pareto front of the multi-objective optimization problem according to operator policy.
B. Routing
With the introduction of wireless backhaul links and the support of multi-hop backhaul, thedesign of routing policies in the IAB network should be able to scale together with the increasingcomplexity. The end-to-end ergodic capacity for an IAB network was shown to be decreasingas the number of hops increases in [88]. In order to maximize the overall sum rate as well asminimize the average latency, the joint route selection and radio resource allocation optimizationfor an indoor IAB network was studied in [69]. Based on the virtual-network method, the designof routing and radio resource allocation among different links can be optimized independently.As expected, by ray-tracing simulations, the number of IAB-donors has a positive impact on theperformance of the proposed centralized scheme. The authors in [70] considered an outdoorin-band mesh-architecture IAB network. Under the constraints of QoS, the centralized andsemi-centralized joint routing and power control mechanism were presented. This mechanism isrealized by a logical controller which manages the network across the MAC layer. It was shownhat the proposed mechanism does not need intense modifications on current standards whileproviding robust performance for real-time flows.The optimal flow link resource allocation and routing policies to maximize the geometric meanof users in a single-tier multi-hop network mmWave IAB network is proposed in [71]. The authorsinvestigated different routing patterns, and found that the only top 20 percentile user rates in IABwith RSRP-based ST outperforms IAB with mesh. The suboptimal allocation of the subcarriersand the transmitting power are derived by the dual-decomposition method, and a systematicapproach to determine the optimal node locations was developed. Simulation results revealedthat the performance gap between the optimal and the proposed suboptimal method is negligible.In [72], the authors addressed the dynamic scheduling problem for a mmWave DAG network,where BSs are equipped with multiple antenna panels. By considering flexible access/backhaulTDD and dynamic route selection, a heuristic based centralized scheduling scheme was presented.Both median and th percentile of UL and DL rate for the proposed scheme was shown to beimproved significantly by system-level simulation.Different from the centralized solution in [69], [71], four distributed greedy path selectionstrategies: 1) Highest-quality-first (HQF), 2) Wired-first (WF) policy, 3) Position-aware (PA)policy, and 4) Maximum-local-rate (MLR) were studied in [73]. The HQF and MLR policies areimplemented by selecting the node with the best SNR and achievable rate, respectively. As for theWF policy, the IAB-node prefers to connect with the reachable IAB-donor if the SNR exceeds agiven threshold. Otherwise, the IAB-node will connect with the node with the highest SNR. ThePA policy is based on the position information, the IAB-node will connect with the highest SNRnode selected from the candidate nodes which are closer to the target IAB-donor than currentIAB-node. Moreover, in order to maintain a small number of hops for the proposed HQF, WF,and MLR policies, the authors introduced two wired bias functions which are polynomial andexponential in the number of hops, respectively. The wired bias function is applied to the SNRof the wired IAB-donor, and takes the value which is larger than the fixed tolerable SNR levelwhen the number of traveled hops reaches the threshold. Simulations revealed that aggressivebias function is able to decrease the number of hops needed to connect to an IAB-donor.As shown in [89], mesh-architecture IAB network can achieve 6.70% ∼ C. Fairness
From the user’s point of view, the service provider needs to focus on fairness to meet theirsatisfaction requirements. In this subsection, we survey the scheduling studies in the IAB networkintending to guarantee users fairness.In [80], Yuan et al. addressed the maximum throughput fair scheduling (MTFS) problemin a TDMA mmWave HetNet with one multiple-RF-chains-equipped IAB-donor and multiplesinge-RF-chain-equipped IAB-nodes. The topology of the considered system is represented bya directed graph, corresponding to an adjacency matrix. A two-step approach was presentedto solve the MTFS problem: first, maximize the max-min throughput and then find the optimalscheduling solution to maximize network throughput for the given max-min throughput. Based onmatching theory and ellipsoid algorithm, the authors first proposed an optimal MTFS algorithmwith polynomial time complexity. Next, they developed an edge-color approximation algorithmwith improved runtime efficiency. Simulation results showed that the proposed optimal algorithmcould converge with a few minutes for over 200 IAB-nodes. The edge-color approximationmethod can be 5 to 100 times faster with a performance decline up to 20%.The authors in [81] investigated the scheduling problem in a hybrid IAB network. Theconsidered network enables an IAB-node to switch its operating mode between HD and IBFD.Analytical expressions for DL and UL transmission rates for users were first derived. Basedon the overall fairness of users, IAB-node can select the mode which contributes the largestvalue. For dense IAB-node deployment scenario, simulation results showed that IAB-nodesprefer to select IBFD mode to improve users’ fairness. PF scheduling was proved to be notsuitable for multi-hop network in [91]. The impact of applying a modified PF scheduling ina multi-hop IAB network was studied in [82]. By adding a weight related to the numbers ofusers served by IAB-node, the authors proposed a weighted proportional fair (WPF) algorithm.Moreover, an IAB-aware flow control algorithm aiming at improving the system throughput andmitigating congestion were given. Compared with the original PF integrated with traditionalend-to-end flow control, the proposed scheme can result in a 30% and 2% increment of fairnessndex and system throughput, respectively. Under a strict fairness requirement, the joint flowcontrol, user association, and power control for an in-band IAB network was studied in [83]. Theoptimal power control and flow control policies were derived by using a backpressure algorithmand geometric programming, respectively. In comparison with the out-band IAB network, thethroughput at the in-band IAB-node was shown to be doubled.
D. Incentive Mechanism Design
As a promising approach to address the multi-objective optimization problems in wirelessnetworks, the application of economic and pricing models in the wireless network has attractedmuch attention from the research community [92]. The trade-off between the revenue of IAB-nodes and mobile users satisfaction in an IBFD network was investigated in [84]. The authorsproposed a price-based scheme by using Stackelberg game theory. For instance, IAB-nodes actas the leaders, while their mobile users act as the followers. The mobile user can be regardedas a buyer with the aim to maximize the difference between its gained transmission rate and thepayment, while the IAB-node aims at maximizing its revenue under a specific power allowance.An iterative algorithm was proposed to obtain the price for the resource and the power allocationpolicy which reaches the state of proved unique Stackelberg equilibrium.In a user-provided network (UPN), users can serve as providers, directly offering connectivityto other users [93]. This architecture can provide users with good channel condition whenthe link to the BS suffers from poor quality, in a device-to-device (D2D) mode. However, inpractice, users would have no incentive to be such a provider unless they receive satisfyingrewards from the operator to compensate for their transmitting cost. By considering a D2Dmode UPN under mmWave IAB HetNet, joint incentive and resource allocation design wasstudied in [85]. The authors formulated a Nash bargaining problem under the user utility, thesensitivity of battery energy, the incentive compensation, and the limitation of network resourcesconstraints. A centralized algorithm and a distributed algorithm were proposed to derive optimalNash bargaining solution and decoupled sub-problem, respectively. The performance of theproposal is evaluated in terms of pay off, download data, and the operator’s revenue. It wasshown that the joint optimization scheme can outperform the optimized resource or incentiveonly at least in every performance metric. . Network Coding-Aware Scheduling
Network coding enables the router to broadcast a mixture of information packets. Due to itspotential to provide energy-efficient and high throughput performance, network coding has foundwider acceptance in the academia and industry [94]–[96]. For a HD hexagonal wireless backhaulnetwork where the MBS transmits a linear combination of the messages to a set of SBSs, theper user degrees of freedom was shown to be approaching / in [97]. In [86], the authorsconsidered the integration of XOR network coding with the IAB network. It is assumed that themessage sent by an IAB-donor can be split into a common part and a private part. An IAB-nodecan XOR its decoded signal and broadcast the result to the mobile user and the IAB-donor forDL and UL transmissions, respectively. With the aim to minimize the energy consumption ofthe IAB-donor, the authors proposed the optimal policies for message splitting ratio, private parttransmitting power, and common part beamformer design. Meanwhile, the energy consumptionminimization problem for the IAB-nodes was formulated and solved by the convex optimizationmethod.The drawback of [86] is that only one-hop wireless backhaul is supportable and two mobileusers are considered. The properties of multi-hop backhaul and multi-route topology for theIAB network bring new challenges. In [87], with the consideration of more complex topologyin IAB network, the authors proposed a linear network coding solution to enhance the perfor-mance robustness and reduce the latency. A rate-proportional mechanism and an adaptive coded-forwarding mechanism were proposed to balance the traffic load among routes and determinethe ratio of coded messages, respectively. Compared with the round-robin mechanism, the rate-proportional mechanism can improve 25% spectral efficiency performance. In comparison withARQ retransmission, for a given input data rate, the adaptive coded-forwarding mechanism wasproved to be able to double the success rate.VI. C ACHE - ENABLED
IAB N
ETWORKS
TABLE IV: Taxonomy of Literature in Section VI.
Objective Methodology Network Reference
SCDPAnalysis Stochastic geometry massive MIMO;OBFD;Caching [98]
Continued on next page
ABLE IV –
Continued from previous page
Objective Methodology Network Fourth Reference
ASPAnalysis IBFD;mmWave;Caching [99]ThroughputAnalysis OBHD;mmWave;Caching [100]CachingStrategyDesign Matching theory;ARIMA Single-hopbackhaul [101]MAB-basedlearning Unknown contentpopularity [102]Game theory;RL HybridmmWave&Sub-6 Ghz [103]DRL MDS-codedcaching [104]Heuristic algorithm mmWave [105]
By caching popular contents from remote servers in network devices positioned closely tomobile users, the burden of wireless backhaul can be alleviated. In this section, we survey theresearch focused on cache-enabled IAB network.
A. Successful Probability and Throughput Analysis
The authors in [98] considered a two-tier edge caching system, with limited storage resourceSBSs and MIMO-enabled MBS. The analytical framework for the successful content deliveryprobability (SCDP) was developed using tools from SG. Further, a one-dimension search al-gorithm was proposed to obtain the minimum BSs density and maximum load at a small-cellunder the SCDP requirement, respectively. Simulation results revealed that the hit probabilityshould be less than a certain level to achieve a given SCDP. The reason behind this requirements that the larger hit probability could introduce more interference from SBS-tier and degradethe performance of cached content delivery. Also, it was shown that the delay of non-cachedcontent requests can be reduced close to the cached content request with the aid of MIMO-enabled backhaul.With the consideration of using IBFD wireless backhaul, the authors investigated the averagesuccess probability (ASP) for a cache-enabled mmWave IAB network in [99]. The IAB nodesserve multiple users in IBFD mode with hybrid beamforming. IAB-nodes equipped with astorage memory to pre-cached some popular contents in order to alleviate network traffic loads.The authors assumed a tuning parameter to control the amount of RSI, and investigated twocaching strategies: uniform (UC) and M most popular files (MC), then derived the average rate,latency, and upper-bound for the average success probability (ASP) of file delivery. In additionto investigating the latency and mean throughput, the authors derived a lower-bound for the ASPof content delivery. Simulation results showed that MC significantly outperforms UC for ASPand the performance gap of rate and latency between IBFD and OBHD decreases as the storagespace at IAB-nodes increases.In [100], Zhang et al. derived the average potential throughput (APT) of a two-tier out-bandmmWave IAB cache-enabled network. The IAB-donors are connected to the core network viaoptical fiber links and contain all kinds of files. Similar to [21] and [22], the total spectrumresource at IAB-donor for the DL is partitioned into access and backhaul with a fixed ratiofor the whole system. By using a SG tool, the authors developed an analytical framework tocapture the SINR statistics and APT of users associated with each tier. Simulation results showedthat the considered cache-enable setting can achieve 80% APT improvement compared with thetraditional setting. B. Caching Strategy Design
The cached content placement decision model is captured by fixed hit probability in [98]–[100]. As suggested in [98] and [99], the hit probability plays an important role in the networkperformance. In order to find the optimal content caching decision at IAB-nodes and the mobileusers association policy, the authors in [101] formulated a mixed-integer optimization problemwith an objective function formulated as the weighted sum of transmission time and content hitnumber. By using matching theory and Autoregressive Integrated Moving Average (ARIMA)method, an iterative algorithm was proposed to solve the joint problem. Compared with aonventional network, the proposed scheme brings up to 82% performance improvement. Withthe assumption that cache-enabled IAB-nodes have three different backhaul options: 1) wiredfiber, 2) mmWave wireless backhaul, and 3) sub-6 GHz wireless backhaul, authors proposed agame-theoretic learning algorithm to solve the content caching problem in [103]. A minoritygame was formulated, where IAB-nodes act as players with independent decisions for backhauloption choice and types of caching contents. In comparison with the greedy algorithm, theproposed scheme can achieve up to 85% performance improvement.Both [101] and [103] assumed that the content popularity is stable over time, this assumptioncould lead to performance degradation in reality when the content popularity is unpredictable. Inthis context, with the consideration of dynamic content popularity, the authors in [102] proposeda learning algorithm based on multi-armed bandit (MAB) theory to maximize the bandwidth alle-viation for backhaul links. The proposed algorithm is able to learn the content placement decisionwithout the need for file popularity information. In comparison with the random cache method,the proposed algorithm can achieve up to 7 times performance improvement. By introducinga virality parameter to represent the change of users’ content requests over time, the authorsin [105] studied the cached content placement problem in a mmWave IAB network. Moreover,a hierarchical caching architecture was considered, where the IAB-donor and IAB-nodes storedifferent finite kinds of contents. The request for content stored in the central storage server willbe forwarded by IAB-donor via a wireless backhaul. In order to minimize backhaul load andaverage delay, the authors proposed a heuristic method to derive the optimal content placementdecision. Simulation results showed that the proposed method can achieve improvement for hitprobability by 23%, and reduction in backhaul load and average delay by 48% and 27%. Thecoded cooperation of cache-enabled IAB-nodes was considered in [104]. The authors applied themaximum distance separable (MDS) code at IAB-nodes. By leveraging the Q-learning model,the content caching decision is derived. Compared with the DDPG method, the proposed methodcan reduce the time complexity significantly.VII. O
PTICAL
IAB N
ETWORKS
ABLE V: Taxonomy of Literature in Section VII.
Objective Methodology Network Reference
InterferenceMitigation Heuristic algorithm VLC& NOMA [106]PowerConsumptionMinimization&SINR Analysis Projected subgradient& Stochastic analysis VLC [107], [108]ThroughputMaximization [109], [110]Lagrange dualdecomposition FSO [111]Routing Heuristic algorithm [112]DelayMinimization [113]DeploymentDesign Sequentialcomputation [114]
Optical wireless communication (OWC) using wavelength band ranges from 350 nm to 1550nm is a promising alternative solution to RF transmission [115]. This section reviews thestudies for the incorporation of visible light communication (VLC) and free-space optical (FSO)communication into IAB networks, respectively.
A. VLC
In VLC, the transmission is based on the modulation of the intensity of the optical source,which uses the visible range of electromagnetic spectrum [116]. Due to the energy-efficientbenefit and long lifetime, Light-emitting diode (LED) is widely used in indoor illumination. Theauthors in [107] developed an analytical framework to investigate SINR and average spectralefficiency in a hexagonal VLC-based IAB network. Moreover, the effect of employing in-bandand out-band wireless backhaul was studied. Compared with out-band backhaul, it was shownhat the in-band mode can achieve better performance when the emission semiangle exceeds25 ◦ or falls below 20 ◦ . The work in [107] is extended in [108] to a two-tier multi-hop wirelessbackhaul scenario. For that setup, the authors derived the power control policy by solving thebackhaul transmitting power consumption minimization problem. It was shown that, in terms ofaverage sum rate, a small value of the emission semiangle can achieve better performance, andin-band mode significantly outperforms out-band mode.The joint backhaul spectrum resource and transmitting power allocation problem was addressedin [109]. The authors first derived the analytical expressions for access and backhaul rate, andthen used it to obtain the power and spectrum allocation coefficients under the backhaul capacityrequirement. From an optimization perspective, in [110], based on projected subgradient method,with the aim to maximize the sum rate of mobile users and the IAB-nodes, user-centric and cell-centric backhaul spectrum scheduling policies were derived for the considered two-tier multi-hopVLC-based IAB network. By integrating the RF transmission with VLC, the authors in [106]considered a hybrid network with RF-based IAB-donor and VLC-based IAB-nodes. In particular,the IAB-node employed NOMA to improve spectral efficiency. With the objective to minimizethe interference, a heuristic algorithm was proposed to obtain the bandwidth reuse decisionamong the IAB-nodes. Simulation results showed that the proposed dynamic scheme can achieveperformance improvement in both sum rate and spectrum efficiency. B. FSO
FSO enables high capacity transmission between access points (APs) that are separated byseveral kilometers, with a frequency that is above 300 GHz [117], [118]. Therefore, FSOcommunication is an attractive option for coping with the increased communication trafficdemands especially in the outdoor environment [119]. However, one of the inherent difficultiesfor FSO communication is the requirement for a reliable LoS. In [112], the authors first proposeda greedy algorithm to optimize the network sum rate and overall power cost, and to obtain thepolicies for connecting path and routing. Moreover, two network reconfiguration solutions to dealwith the link failure and dynamic traffic demand changement were proposed, respectively. Inorder to improve the reliability, the authors in [114] considered the usage of the mirror-aided linksto ensure the connectivity of two distinct nodes when their LoS is blocked. For the formulatedgraph optimization problem with the aim to minimize the sum of weighted cost and reliable linklength, they proposed a sequential computation algorithm to obtain the connecting paths as wells the number and locations of mirrors, IAB-donors, and FSO transceivers. Based on a realistictopology model, the simulation results showed that the proposed method can achieve optimal ornear-optimal performance with much lower time complexity.The authors in [113] considered an IAB network where the IAB-donor is equipped with aFSO transceiver. With the objective to minimize the overall delay under the requirement of allIAB-nodes are covered, they solved the multicast problem for IAB-nodes when they are static ormobile. By reformulating the problem as a time-dependent prize collecting traveling salesmanproblem, the authors proposed several heuristic algorithms to find the optimal scheduling ofthe directional optical links. A hybrid RF/FSO UL transmission for IAB networks was studiedin [111], where there are two types of links in the wireless backhaul. In particular, the authorsconsidered two transmission schemes at the IAB-nodes: 1) delay-tolerant, where it can storethe received information packets and forward them later; 2) delay-limited, where it forwards thereceived signal immediately. By solving the throughput maximization problem via Lagrange dualdecomposition method, the optimal UL block resource allocation policy was derived. Simulationresults showed that sharing block resources is required to improve the throughput when thequality of FSO links falls below a certain level.VIII. N ON -T ERRESTRIAL
IAB N
ETWORKS
TABLE VI: Taxonomy of Literature in Section VIII.
Objective Methodology Network Reference
DeploymentDesign Ray tracing UAV-terrestrial [120]Heuristic algorithm [121][122][123]Linear search UAV-marine [124]SCDPAnalysis Stochastic geometry Cache-enabledUAV [125]OPAnalysis UAV-terrestrial [126]Routing Linear programming [127]Heuristic algorithm [128], [129]
Continued on next page
ABLE VI –
Continued from previous page
Objective Methodology Network Fourth Reference
ThroughputMaximization Hungarian method Satellite-terrestrial [130]Swap matching [131]IncentiveMechanism Game theory Satellite-aerial-terrestrial [132][133]
Providing global connectivity is not a new ambition [134]. To this end, satellite and unmannedaerial vehicle (UAV) are attractive options to offer communication service over the areas thatare too expensive to reach or too difficult to deploy conventional terrestrial networks (e.g., ruraland remote regions, marine area) [135], [136]. This section focuses on the research related tothe non-terrestrial IAB networks. We classify these papers into two categories: (i) UAV-assistedIAB networks, and (ii) satellite-terrestrial IAB networks.
A. UAV-assisted IAB Networks
Due to the ability to their mobility and relocation flexibility, UAV-BSs could be deployedat any 3D position of interest. However, due to the non-negligible multipath propagation andlink blockage in a realistic environment, one important issue in a UAV-assisted network is theoptimization of UAV-BSs positions. In [120], based on ray-tracing simulations, the authors firstinvestigated the coverage gains of UAV-assisted mmWave IAB network, where the UAV-BSsact as IAB-nodes. With the consideration of amplify-and-forward (AF) OBFD and decode-and-forward (DF) IBFD modes, the optimal positions for UAVs were derived via the ray tracing-basedcoverage maps. The simulation results showed that the AF mode outperforms DF mode with31% DL coverage gains improvement.Zhang et al . addressed the joint optimization problem for the number and 3D positions ofUAV-BSs in an IBFD UAV-assisted IAB network [121]. A heuristic algorithm was proposed tosolve the optimization problem which aims to minimize the number of UAVs while maximizingthe overall transmission rate. Numerical results proved that the proposed algorithm can not onlyincrease the overall throughput but also can decrease data rate block ratio. In [122], the authorsolved the UAV-BSs placement problem in order to maximize the sum rate and the number ofserved mobile users. Moreover, the effect of the movement of mobile users was investigatedin the simulations. Simulation results showed that the proposed method can achieve robustperformance under the impact of users’ mobility. By defining a piecewise-linear function ofgained transmission rate that captures the profit of mobile users, the authors in [123] addressedthe joint UAV-BS position and bandwidth allocation optimization problem. In order to maximizethe total network profit, they proposed a heuristic algorithm to solve the formulated mixed-integernon-linear programming problem. Numerical results showed that the proposed solution is ableto significantly outperform the well-known BARON solver [137] and NEOS platform[138].In a realistic environment, the dynamic change for both the multipath propagation and linkblockage are non-negligible. The authors in [126] developed analytical frameworks to charac-terize the performance of a UAV-assisted mmWave IAB network with terrestrial BSs (TBSs)and UAV-BSs. With the consideration of the dynamic mobility of human blockers and UAV-BSs, the time-averaged and time-dependent performance metrics were derived. It was shownthat both outage probability (OP) and spectrum efficiency were improved as the intensity ofUAV-BSs traversals increases. In addition, it was also shown that the lower flight speed canachieve a better wireless backhaul performance. The authors in [124] considered a maritimecommunication system with cache-enabled UAV-BSs. The optimal horizontal positions of UAV-BSs were obtained by an iterative one-dimensional linear search algorithm to maximize thesum rate. In [125], the authors investigated the incorporation of cache-enabled UAV-BSs intoterrestrial cache-enabled IAB network. A tractable model for characterizing SCDP and energyefficiency was developed. Simulation results showed that the considered scheme can achieve26.6% SCDP performance improvement on average.The path selection problem for a UAV-assisted multi-hop IAB network was addressed in [127].The authors formulated and solved a binary linear optimization program aiming at maximizingthe total transmission rate, and obtain the optimal path scheduling policy. By integrating UAVsas drone BSs into the IAB network, Fouda et al . [129] presented a system model for forwardlink transmissions in an in-band IAB HetNet. Given that the backhaul link to UAV is providedby the IAB-donor, an optimization problem is formulated that aims to achieve the maximumsum rate of the users, under the constrains that the mutual interference between access andbackhaul links is below a given threshold. By using an alternative optimization method, theauthors derived the optimal 3D hovering positions of the UAVs, mobile users association policy,recoder design at the backhaul link, and transmitting power allocation policy. Further, a betterperformance algorithm based on particle swarm optimization (PSO) is proposed in [128].
B. Satellite-Terrestrial IAB Networks
One advantage of satellite-terrestrial communication is that it enables the network operators tocover wider areas at a lower cost. There are several on-going satellite-terrestrial communicationprojects, such as SpaceX and OneWeb, aiming to provide global-coverage and high data ratesvia low earth orbit satellite (LEO) networks. In this context, the design of satellite-terrestrialbackhaul link plays a key role in the service quality of users [139].The authors in [130] addressed the spectrum resource allocation problem for the DL transmis-sion in a satellite-terrestrial IAB network. The considered system assumes the terrestrial links andsatellite links can reuse the same frequency. By decomposing the sum rate maximization probleminto two sub-problems, the optimal carrier allocation strategy was derived by the Hungarianmethod sequentially. In [131], a satellite-terrestrial IAB network architecture for data offloadingwas studied. The satellite-terrestrial backhaul links share the Ka-band while the terrestrial linksshare the C-band. Terrestrial mobile users have access to the core network through the macro-cell, the small-cell, or the LEO-based small-cell. With the aim to maximize the sum rate, theauthors proposed a modified swap matching algorithm to obtain the optimal policies for usersassociation, subchannel allocation, and power control. Simulation results revealed that the totalcapacity is not a monotonic function of the projected area of the satellite, which means thatthere exists an optimal satellite deployment solution to maximize the total backhaul capacity.With the application of the game theory model into a satellite-terrestrial IAB network, theauthors in [132] proposed a data offloading pricing mechanism based on the Stackelberg gamemodel. To deal with the increasing backhaul demand at conventional APs, they assumed someuser’s demand from conventional APs can be migrated to LEO-based APs. In this context, theLEO-based APs act as the leaders while the conventional APs act as the followers. An iterativealgorithm was proposed to reach the Stackelberg equilibrium. In each iteration, the follower leveloptimization problem aims at optimizing user association is solved by fractional programming,then the leader level problem with the aim to optimize the service price and Ka-band spectrumallocation is solved by alternative optimization. The simulation results showed that there is anoptimal LEO satellite density to balance the trade-off between APs utility and cost.n addition to the application of the game theory model, the integration of UAV and satellite-terrestrial IAB networks was investigated in [133]. The IAB-nodes consist of the satellite,terrestrial SBSs, and UAVs, while the IAB-donor is the MBS. This joint backhaul and accessresource management problem was formulated as a competitive market. Hu et al. assumedthat some UAVs, satellite, some SBSs, and MBSs act as “sellers”, where others UAVs andSBSs act as “customers”. The communication services were regarded as “goods” and theirprices are determined by the QoS, while the cost is determined by the power consumption.The network seeks for achieving the Walrasian equilibrium, at which there are no good exits,and each role’s profit is maximized. The simulation results showed that the proposed algorithmcan reach Walrasian equilibrium within 200 iterations. Compared with random allocation, theproposed algorithm can achieve 3 to 4 times gain in terms of data rate.IX. C
HALLENGES AND O PPORTUNITIES
A. IAB Deployment
In reality, the height of blockages and the terrain information of land cannot be negligible forthe UHF channel. Therefore, the 3D position should be considered in the IAB deployment. More-over, the interaction between different layers of the protocol stack is another design challengefor IAB. Designing a more realistic and efficient deployment solution to improve the end-to-endperformance is still worth being studied.
B. Scheduling
DAG and ST limit the flexibility and efficiency due to the fixed parent-to-child relation betweentwo adjacent IAB nodes. Studies in both [89] and [71] proved that mesh-architecture based IABcan achieve a considerable improvement in comparison with DAG-architecture based IAB. Itis expected that mesh-architecture-based IAB will be introduced in the Rel-17. The solution toaddress congestion control and routing will be important in mesh-architecture-based IAB.
C. Mobility Management
Due to the usage of higher frequencies, the wireless backhaul link in IAB is vulnerable tomobility blockage (e.g., vehicles and human movements). Therefore, from a resilience perspec-tive, it is desired to ensure that an IAB network can provide reliable services to end-users whensome backhaul links are degraded or lost. . Intelligent IAB
In IAB, the relationship between access and backhaul will work more closely than ever, totalseparation of their resources may not be possible anymore, and joint operation is required.Designing a mechanism to realize an intelligent access-demand-aware backhaul system, where acentral or distributed controller optimizes backhaul capacity according to dynamic access networkdemand, may be a crucial issue in operator maintenance.X. C
ONCLUSIONS
IAB provides an economic-friendly and flexible network deployment solution in the era of5G and beyond network densification. In this paper, we first gave a brief introduction to IAB.Next, a discussion and comparison of recent research works on IAB networks were given. Inaddition, we introduced state-of-the-art studies on integrated IAB with cache-enabled network,optical transport network, and non-terrestrial communication, respectively. We also described thepossible challenges and opportunities in this promising research area.A
UTHOR C ONTRIBUTIONS
The work was developed as a collaboration among all authors. All authors conceived thework and suggested the outline of the paper. The manuscript was mainly drafted by YZ and wasrevised and corrected by all co-authors. All authors have read and approved the final manuscript.A
CKNOWLEDGEMENT
This work was supported by Office of Sponsored Research (OSR) at King Abdullah Universityof Science and Technology (KAUST).C
ONFLICT OF I NTEREST S TATEMENT
The authors declare that the research was conducted in the absence of any commercial orfinancial relationships that could be construed as a potential conflict of interest.R
EFERENCES [1] D. Soldani and A. Manzalini, “Horizon 2020 and beyond: On the 5G operating system for a true digital society,”
IEEEVeh. Technol. Mag. , vol. 10, no. 1, pp. 32–42, Mar. 2015.[2] I. Union, “IMT traffic estimates for the years 2020 to 2030,”
Report ITU , pp. 2370–0, 2015.3] N. Bhushan, J. Li, D. Malladi, R. Gilmore, D. Brenner, A. Damnjanovic, R. Sukhavasi, C. Patel, and S. Geirhofer,“Network densification: the dominant theme for wireless evolution into 5G,”
IEEE Commun. Mag. , vol. 52, no. 2, pp.82–89, Feb. 2014.[4] M. Agiwal, A. Roy, and N. Saxena, “Next generation 5G wireless networks: A comprehensive survey,”
IEEE Commun.Surveys Tuts. , vol. 18, no. 3, pp. 1617–1655, 2016.[5] H. A. Willebrand and B. S. Ghuman, “Fiber optics without fiber,”
IEEE Spectr. , vol. 38, no. 8, pp. 40–45, 2001.[6] O. Teyeb, A. Muhammad, G. Mildh, E. Dahlman, F. Barac, and B. Makki, “Integrated access backhauled networks,” in
Proc., IEEE 90th Vehicular Technology Conference (VTC-Fall) , Sep. 2019.[7] C. B. Czegledi, M. Horberg, M. Sjodin, P. Ligander, J. Hansryd, J. Sandberg, J. Gustavsson, D. Sjoberg, D. Polydorou,and D. Siomos, “Demonstrating 139 Gbps and 55.6 bps/Hz spectrum efficiency using 8×8 MIMO over a 1.5-km link at73.5 GHz,” in
Proc., IEEE/MTT-S International Microwave Symposium (IMS) , Aug. 2020.[8] S. Dang, O. Amin, B. Shihada, and M.-S. Alouini, “What should 6G be?”
Nature Electronics , vol. 3, no. 1, pp. 20–29,2020.[9] 3GPP, “Overview of 3GPP Release 10,” 3GPP, Tech. Rep., 2016, v10.2.1.[10] C. Dehos, J. L. Gonz´alez, A. De Domenico, D. Ktenas, and L. Dussopt, “Millimeter-wave access and backhauling: thesolution to the exponential data traffic increase in 5G mobile communications systems?”
IEEE Commun. Mag. , vol. 52,no. 9, pp. 88–95, 2014.[11] C. Madapatha, B. Makki, C. Fang, O. Teyeb, E. Dahlman, M.-S. Alouini, and T. Svensson, “On integrated access andbackhaul networks: Current status and potentials,”
IEEE Open Journal of the Communications Society , vol. 1, pp. 1374–1389, 2020.[12] 3GPP, “NG-RAN; Architecture description,” 3GPP, TS 38.401, Jul. 2020, v16.2.0.[13] H. Raza, “A brief survey of radio access network backhaul evolution: part I,”
IEEE Commun. Mag. , vol. 49, no. 6, pp.164–171, Jun. 2011.[14] M. Polese, M. Giordani, T. Zugno, A. Roy, S. Goyal, D. Castor, and M. Zorzi, “Integrated access and backhaul in 5GmmWave networks: Potential and challenges,”
IEEE Commun. Mag. , vol. 58, no. 3, pp. 62–68, Mar. 2020.[15] E. Dahlman, S. Parkvall, and J. Skold,
5G NR: The Next Generation Wireless Access Technology , 2nd ed. San Diego:Academic Press, 2020.[16] 3GPP, “NR; Study on Integrated Access and Backhaul; Release 15,” 3GPP, TR 38.874, 2018.[17] T. Tian, Y. Dou, G. Ren, L. Gu, J. Chen, Y. Cui, T. Takada, M. Iwabuchi, J. Tsuboi, and Y. Kishiyama, “Field trial onmillimeter wave integrated access and backhaul,” in
Proc., IEEE Vehicular Technology Conference (VTC2019-Spring) ,Apr. 2019.[18] M. N. Kulkarni, J. G. Andrews, and A. Ghosh, “Performance of dynamic and static TDD in self-backhauled millimeterwave cellular networks,”
IEEE Trans. Wireless Commun. , vol. 16, no. 10, pp. 6460–6478, Oct. 2017.[19] S. Singh, M. N. Kulkarni, A. Ghosh, and J. G. Andrews, “Tractable model for rate in self-backhauled millimeter wavecellular networks,”
IEEE J. Sel. Areas Commun. , vol. 33, no. 10, pp. 2196–2211, Oct. 2015.[20] A. Sharma, R. K. Ganti, and J. K. Milleth, “Joint backhaul-access analysis of full duplex self-backhauling heterogeneousnetworks,”
IEEE Trans. Wireless Commun. , vol. 16, no. 3, pp. 1727–1740, Mar. 2017.[21] C. Saha, M. Afshang, and H. S. Dhillon, “Bandwidth partitioning and downlink analysis in millimeter wave integratedaccess and backhaul for 5G,”
IEEE Trans. Wireless Commun. , vol. 17, no. 12, pp. 8195–8210, Dec. 2018.[22] C. Saha and H. S. Dhillon, “Millimeter wave integrated access and backhaul in 5G: Performance analysis and designinsights,”
IEEE J. Sel. Areas Commun. , vol. 37, no. 12, pp. 2669–2684, Dec. 2019.23] H. Tabassum, A. H. Sakr, and E. Hossain, “Massive MIMO-enabled wireless backhauls for full-duplex small cells,” in
Proc., IEEE Global Communications Conference (GLOBECOM) , Dec. 2015.[24] H. Zhang, Y. Chen, Z. Yang, and X. Zhang, “Flexible coverage for backhaul-limited ultradense heterogeneous networks:Throughput analysis and η -optimal biasing,” IEEE Trans. Veh. Technol. , vol. 67, no. 5, pp. 4161–4172, May 2018.[25] H. ElSawy, E. Hossain, and M. Haenggi, “Stochastic geometry for modeling, analysis, and design of multi-tier andcognitive cellular wireless networks: A survey,”
IEEE Commun. Surveys Tuts. , vol. 15, no. 3, pp. 996–1019, 2013.[26] M. Duarte and A. Sabharwal, “Full-duplex wireless communications using off-the-shelf radios: Feasibility and first results,”in
Proc., IEEE Conference Record of the Forty Fourth Asilomar Conference on Signals, Systems and Computers , Nov.2010.[27] N. Phungamngern, P. Uthansakul, and M. Uthansakul, “Digital and RF interference cancellation for single-channel full-duplex transceiver using a single antenna,” in
Proc., IEEE International Conference on Electrical Engineering/Electronics,Computer, Telecommunications and Information Technology (ECTICON) , May 2013.[28] D. Bharadia, E. McMilin, and S. Katti, “Full duplex radios,” in
Proc., ACM SIGCOMM , 2013, pp. 375–386.[29] D. Kim, H. Lee, and D. Hong, “A survey of in-band full-duplex transmission: From the perspective of PHY and MAClayers,”
IEEE Commun. Surveys Tuts. , vol. 17, no. 4, pp. 2017–2046, 2015.[30] B. Li, D. Zhu, and P. Liang, “Small cell in-band wireless backhaul in massive MIMO systems: A cooperation of next-generation techniques,”
IEEE Trans. Wireless Commun. , vol. 14, no. 12, pp. 7057–7069, Dec. 2015.[31] I. Ullah, B. B. Haile, E. Mutafungwa, and J. Hamalainen, “Use of beamforming and interference mitigation techniquesfor relay backhaul enhancement,” in
Proc., 39th International Conference on Telecommunications and Signal Processing(TSP) . IEEE, Jun. 2016.[32] M. Nakamura, G. K. Tran, and K. Sakaguchi, “Interference management for millimeter-wave mesh backhaul networks,”in
Proc., 16th IEEE Annual Consumer Communications & Networking Conference (CCNC) , Jan. 2019.[33] P. Jayasinghe, A. T¨olli, J. Kaleva, and M. Latva-aho, “Traffic aware beamformer design for integrated access and backhaulwith flexible TDD,” in
Proc., IEEE Wireless Communications and Networking Conference (WCNC) , 2020, pp. 1–6.[34] A. Prasad, M. A. Uusitalo, and A. Maeder, “Energy efficient coordinated self-backhauling for ultra-dense 5G networks,”in
Proc., IEEE 85th Vehicular Technology Conference (VTC Spring) , Jun. 2017.[35] D. Meng, X. Li, W. Pu, X. Yang, and D. Li, “An energy-saving scheme with multi-hop transmission for mmWave backhaulnetworks,” in
Proc., IEEE Vehicular Technology Conference (VTC Spring) , Jun. 2018.[36] D. Korpi, T. Riihonen, A. Sabharwal, and M. Valkama, “Transmit power optimization and feasibility analysis of self-backhauling full-duplex radio access systems,”
IEEE Trans. Wireless Commun. , vol. 17, no. 6, pp. 4219–4236, Jun.2018.[37] L. Lei, E. Lagunas, S. Chatzinotas, and B. Ottersten, “NOMA aided interference management for full-duplex self-backhauling HetNets,”
IEEE Commun. Lett. , vol. 22, no. 8, pp. 1696–1699, Aug. 2018.[38] W. Lei, Y. Ye, and M. Xiao, “Deep reinforcement learning-based spectrum allocation in integrated access and backhaulnetworks,”
IEEE Trans. on Cogn. Commun. Netw. , vol. 6, no. 3, pp. 970–979, Sep. 2020.[39] Q. Liu, H. Tian, G. Nie, and H. Wu, “Joint uplink and downlink scheduling in HetNets with wireless self-backhaul,” in
Proc., 24th Asia-Pacific Conference on Communications (APCC) . IEEE, Nov. 2018.[40] E. Lagunas, L. Lei, S. Maleki, S. Chatzinotas, and B. Ottersten, “Power allocation for in-band full-duplex self-backhauling,” in
Proc., 40th International Conference on Telecommunications and Signal Processing (TSP) . IEEE,Jul. 2017.[41] D. Korpi, T. Riihonen, and M. Valkama, “Self-backhauling full-duplex access node with massive antenna arrays: Powerllocation and achievable sum-rate,” in
Proc., 24th European Signal Processing Conference (EUSIPCO) . IEEE, Aug.2016.[42] S. Zhang, X. Xu, M. Sun, X. Tao, and C. Liu, “Joint spectrum and power allocation in 5G integrated access and backhaulnetworks at mmWave band,” in
Proc., IEEE 31st Annual International Symposium on Personal, Indoor and Mobile RadioCommunications , Aug. 2020.[43] U. Siddique, H. Tabassum, and E. Hossain, “Downlink spectrum allocation for in-band and out-band wireless backhaulingof full-duplex small cells,”
IEEE Trans. Commun. , vol. 65, no. 8, pp. 3538–3554, aug 2017.[44] Y. Liu, X. Fang, and M. Xiao, “Discrete power control and transmission duration allocation for self-backhauling densemmWave cellular networks,”
IEEE Trans. Commun. , vol. 66, no. 1, pp. 432–447, Jan. 2018.[45] M. Lashgari, B. Maham, and W. Saad, “Transmission rate maximization in self-backhauled wireless small cell networks,”in
Proc., IEEE 86th Vehicular Technology Conference (VTC-Fall) , Sep. 2017.[46] P. Blasco, M. Bennis, and M. Dohler, “Backhaul-aware self-organizing operator-shared small cell networks,” in
Proc.,IEEE International Conference on Communications (ICC) , Jun. 2013.[47] G. Kwon and H. Park, “Joint user association and beamforming design for millimeter wave UDN with wireless backhaul,”
IEEE J. Sel. Areas Commun. , vol. 37, no. 12, pp. 2653–2668, Dec. 2019.[48] S. Ni, J. Zhao, H. H. Yang, and Y. Gong, “Enhancing downlink transmission in MIMO HetNet with wireless backhaul,”
IEEE Trans. Veh. Technol. , vol. 68, no. 7, pp. 6817–6832, Jul. 2019.[49] W. Pu, X. Li, J. Yuan, and X. Yang, “Resource allocation for millimeter wave self-backhaul network using markovapproximation,”
IEEE Access , vol. 7, pp. 61 283–61 295, 2019.[50] X. Yang, C. Hua, P. Gu, and W. Xu, “Joint power allocation for non-orthogonal multiple access in wireless backhaulnetworks,” in
Proc., IEEE International Conference on Fog Computing (ICFC) , Apr. 2020.[51] E. Chen, M. Tao, and N. Zhang, “User-centric joint access-backhaul design for full-duplex self-backhauled cooperativenetworks,” in
Proc., IEEE Global Communications Conference (GLOBECOM) , Dec. 2018.[52] ——, “User-centric joint access-backhaul design for full-duplex self-backhauled wireless networks,”
IEEE Trans.Commun. , vol. 67, no. 11, pp. 7980–7993, Nov. 2019.[53] T. K. Vu, M. Bennis, S. Samarakoon, M. Debbah, and M. Latva-Aho, “Joint load balancing and interference mitigationin 5G heterogeneous networks,”
IEEE Trans. Wireless Commun. , vol. 16, no. 9, pp. 6032–6046, 2017.[54] T. K. Vu, M. Bennis, M. Debbah, and M. Latva-Aho, “Joint path selection and rate allocation framework for 5G self-backhauled mm-wave networks,”
IEEE Trans. Wireless Commun. , vol. 18, no. 4, pp. 2431–2445, Apr. 2019.[55] L. Chen, F. R. Yu, H. Ji, G. Liu, and V. C. M. Leung, “Distributed virtual resource allocation in small-cell networks withfull-duplex self-backhauls and virtualization,”
IEEE Trans. Veh. Technol. , vol. 65, no. 7, pp. 5410–5423, Jul. 2016.[56] L. Tang, X. Yang, X. Wu, T. Cui, and Q. Chen, “Queue stability-based virtual resource allocation for virtualized wirelessnetworks with self-backhauls,”
IEEE Access , vol. 6, pp. 13 604–13 616, 2018.[57] L. Chen, F. R. Yu, H. Ji, B. Rong, X. Li, and V. C. M. Leung, “Green full-duplex self-backhaul and energy harvestingsmall cell networks with massive MIMO,”
IEEE J. Sel. Areas Commun. , vol. 34, no. 12, pp. 3709–3724, Dec. 2016.[58] A. Imran, L. Giupponi, M. A. Imran, and A. Abu-Dayya, “Joint coverage and backhaul self-optimization in emergingrelay enhanced heterogeneous networks,” in
Proc., IEEE International Conference on Communications (ICC) , Jun. 2014.[59] L. Chen, F. R. Yu, H. Ji, V. C. Leung, X. Li, and B. Rong, “A full-duplex self-backhaul scheme for small cell networkswith massive MIMO,” in
Proc., IEEE International Conference on Communications (ICC) , May 2016.[60] Y. Zhu, Y. Niu, J. Li, D. O. Wu, Y. Li, and D. Jin, “QoS-aware scheduling for small cell millimeter wave mesh backhaul,”in
Proc., IEEE International Conference on Communications (ICC) , 2016, pp. 1–6.61] J. Qiao, L. X. Cai, X. Shen, and J. W. Mark, “STDMA-based scheduling algorithm for concurrent transmissions indirectional millimeter wave networks,” in
Proc., IEEE International Conference on Communications (ICC) , 2012, pp.5221–5225.[62] C. Liang and F. R. Yu, “Wireless network virtualization: A survey, some research issues and challenges,”
IEEE Commun.Surveys Tuts. , vol. 17, no. 1, pp. 358–380, 2014.[63] A. Bonfante, L. G. Giordano, D. Lopez-Perez, A. Garcia-Rodriguez, G. Geraci, P. Baracca, M. M. Butt, M. Dzaferagic,and N. Marchetti, “Performance of massive MIMO self-backhauling for ultra-dense small cell deployments,” in
Proc.,IEEE Global Communications Conference (GLOBECOM) , Dec. 2018.[64] J. Y. Lai, W.-H. Wu, and Y. T. Su, “Resource allocation and node placement in multi-hop heterogeneous integrated-access-and-backhaul networks,”
IEEE Access , vol. 8, pp. 122 937–122 958, 2020.[65] P. Wainio and K. Seppanen, “Self-optimizing last-mile backhaul network for 5G small cells,” in
Proc., IEEE InternationalConference on Communications Workshops (ICCW) , may 2016.[66] M. Raithatha, A. U. Chaudhry, R. H. Hafez, and J. W. Chinneck, “Locating gateways for maximizing backhaul networkcapacity of 5G ultra-dense networks,” in
Proc., Wireless Telecommunications Symposium (WTS) . IEEE, Apr. 2020.[67] M. N. Islam, S. Subramanian, and A. Sampath, “Integrated access backhaul in millimeter wave networks,” in
Proc., IEEEWireless Communications and Networking Conference (WCNC) , 2017, pp. 1–6.[68] A. L. Rezaabad, H. Beyranvand, J. A. Salehi, and M. Maier, “Ultra-dense 5G small cell deployment for fiber and wirelessbackhaul-aware infrastructures,”
IEEE Trans. Veh. Technol. , vol. 67, no. 12, pp. 12 231–12 243, Dec. 2018.[69] D. Hui and J. Axnas, “Joint routing and resource allocation for wireless self-backhaul in an indoor ultra-dense network,”in
Proc., IEEE Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC) , Sep.2013.[70] R. Favraud, N. Nikaein, and C.-Y. Chang, “QoS guarantee in self-backhauled LTE mesh networks,” in
Proc.,IEEE GlobalCommunications Conference , Dec. 2017.[71] M. N. Islam, N. Abedini, G. Hampel, S. Subramanian, and J. Li, “Investigation of performance in integrated access andbackhaul networks,” in
Proc., IEEE Conference on Computer Communications Workshops , 2018, pp. 597–602.[72] A. Lukowa, V. Venkatasubramanian, and P. Marsch, “Dynamic self-backhauling in 5G networks,” in
Proc., IEEE 29thAnnual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) , Sep. 2018.[73] M. Polese, M. Giordani, A. Roy, D. Castor, and M. Zorzi, “Distributed path selection strategies for integrated access andbackhaul at mmwaves,” in
Proc., IEEE Global Communications Conference (GLOBECOM) , 2018, pp. 1–7.[74] F. Gomez-Cuba and M. Zorzi, “Optimal link scheduling in millimeter wave multi-hop networks with MU-MIMO radios,”
IEEE Trans. Wireless Commun. , vol. 19, no. 3, pp. 1839–1854, Mar. 2020.[75] ——, “Twice simulated annealing resource allocation for mmWave multi-hop networks with interference.” in
Proc., IEEEInternational Conference on Communications (ICC) , Jun. 2020.[76] M. Gupta, A. Rao, E. Visotsky, M. Cudak, A. Ghosh, and J. G. Andrews, “Learning-based delay optimization for self-backhauled millimeter wave cellular networks,” in .IEEE, Nov. 2019.[77] B. Zhang, F. Devoti, and I. Filippini, “RL-based resource allocation in mmWave 5G IAB networks,” in
Proc.,Mediterranean Communication and Computer Networking Conference (MedComNet) . IEEE, Jun. 2020.[78] A. Ortiz, A. Asadi, G. H. Sim, D. Steinmetzer, and M. Hollick, “SCAROS: A scalable and robust self-backhaulingsolution for highly dynamic millimeter-wave networks,”
IEEE J. Sel. Areas Commun. , vol. 37, no. 12, pp. 2685–2698,Dec. 2019.79] A. U. Chaudhry, N. Jacob, D. George, and R. H. Hafez, “On the interference range of small cells in the wireless backhaulof 5G ultra-dense networks,” in
Proc., Wireless Telecommunications Symposium (WTS) . IEEE, Apr. 2020.[80] D. Yuan, H.-Y. Lin, J. Widmer, and M. Hollick, “Optimal joint routing and scheduling in millimeter-wave cellularnetworks,” in
Proc., IEEE Conference on Computer Communications (INFOCOM) , Apr. 2018.[81] U. Siddique, H. Tabassum, and E. Hossain, “Adaptive in-band self-backhauling for full-duplex small cells,” in
Proc.,IEEE International Conference on Communication Workshop (ICCW) , Jun. 2015.[82] Y. Zhang, V. Ramamurthi, Z. Huang, and D. Ghosal, “Co-optimizing performance and fairness using weighted pfscheduling and IAB-aware flow control,” in
Proc., IEEE Wireless Communications and Networking Conference (WCNC) ,2020, pp. 1–6.[83] S. Goyal, P. Liu, and S. Panwar, “Scheduling and power allocation in self-backhauled full duplex small cells,” in , May 2017.[84] A. Rahmati, A. Sadeghi, and V. Shah-Mansouri, “Price-based resource allocation for full duplex self-backhauled smallcell networks,” in
Proc., IEEE International Conference on Communications (ICC) , Jun. 2015.[85] Y. Liu, A. Tang, and X. Wang, “Joint incentive and resource allocation design for user provided network under 5Gintegrated access and backhaul networks,”
IEEE Trans. Netw. Sci. Eng. , vol. 7, no. 2, pp. 673–685, Apr. 2020.[86] H. Thomsen, E. de Carvalho, and P. Popovski, “Using wireless network coding to replace a wired with wireless backhaul,”
IEEE Wireless Commun. Lett. , vol. 4, no. 2, pp. 141–144, Apr. 2015.[87] W. Mao, M. Narasimha, M. Simsek, and H. Nikopour, “Network coding for integrated access and backhaul wirelessnetworks,” in
Proc., 29th Wireless and Optical Communications Conference (WOCC) . IEEE, May 2020.[88] X. Zhang, F. Liu, and H. Xia, “Ergodic capacity analysis for full-duplex integrated access and backhaul system,” in
Proc.,IEEE Global Conference on Signal and Information Processing (GlobalSIP) , Nov. 2019.[89] B. Zhai, M. Yu, A. Tang, and X. Wang, “Mesh architecture for efficient integrated access and backhaul networking,” in
Proc., IEEE Wireless Communications and Networking Conference (WCNC) , May 2020.[90] L. Tassiulas and A. Ephremides, “Stability properties of constrained queueing systems and scheduling policies formaximum throughput in multihop radio networks,”
IEEE Trans. Autom. Control , vol. 37, no. 12, pp. 1936–1948, 1992.[91] M. Andrews, A. Ghosh, R. N. Pupala, and S. Vasudevan, “Performance evaluation of self backhauled small cellheterogeneous networks,”
IEEE Trans. Wireless Commun. , vol. 16, no. 5, pp. 3102–3110, May 2017.[92] N. C. Luong, P. Wang, D. Niyato, Y.-C. Liang, Z. Han, and F. Hou, “Applications of economic and pricing models forresource management in 5G wireless networks: A survey,”
IEEE Commun. Surveys Tuts. , vol. 21, no. 4, pp. 3298–3339,2019.[93] R. Sofia and P. Mendes, “User-provided networks: consumer as provider,”
IEEE Commun. Mag. , vol. 46, no. 12, pp.86–91, 2008.[94] C. Fragouli, J. Widmer, and J.-Y. Le Boudec, “On the benefits of network coding for wireless applications,” in
Proc., 4thInternational Symposium on Modeling and Optimization in Mobile, Ad Hoc and Wireless Networks . IEEE, 2006, pp.1–6.[95] M. A. Iqbal, B. Dai, B. Huang, A. Hassan, and S. Yu, “Survey of network coding-aware routing protocols in wirelessnetworks,”
Journal of Network and Computer Applications , vol. 34, no. 6, pp. 1956–1970, 2011.[96] E. Magli, M. Wang, P. Frossard, and A. Markopoulou, “Network coding meets multimedia: A review,”
IEEE Trans.Multimedia , vol. 15, no. 5, pp. 1195–1212, 2013.[97] M. Bande and V. V. Veeravalli, “Design of a heterogeneous cellular network with a wireless backhaul,”
IEEE Trans.Wireless Commun. , pp. 1–1, 2020.98] L. Wang, K.-K. Wong, S. Lambotharan, A. Nallanathan, and M. Elkashlan, “Edge caching in dense heterogeneous cellularnetworks with massive MIMO-aided self-backhaul,”
IEEE Trans. Wireless Commun. , vol. 17, no. 9, pp. 6360–6372, Sep.2018.[99] T. Zhang, S. Biswas, and T. Ratnarajah, “An analysis on caching in full-duplex enabled mmWave IAB HetNets,” in
Proc.,IEEE International Symposium on Wireless Communication Systems (ISWCS) , Aug. 2019.[100] C. Zhang, H. Wu, H. Lu, and J. Liu, “Throughput analysis in cache-enabled millimeter wave HetNets with access andbackhaul integration,” in
Proc., IEEE Wireless Communications and Networking Conference (WCNC) , May 2020.[101] R. Haw, S. M. A. Kazmi, K. Thar, M. G. R. Alam, and C. S. Hong, “Cache aware user association for wirelessheterogeneous networks,”
IEEE Access , vol. 7, pp. 3472–3485, 2019.[102] P. Blasco and D. Gunduz, “Learning-based optimization of cache content in a small cell base station,” in
Proc., IEEEInternational Conference on Communications (ICC) , Jun. 2014.[103] K. Hamidouche, W. Saad, M. Debbah, J. B. Song, and C. S. Hong, “The 5G cellular backhaul management dilemma: Tocache or to serve,”
IEEE Trans. Wireless Commun. , vol. 16, no. 8, pp. 4866–4879, Aug. 2017.[104] S. Gao, P. Dong, Z. Pan, and G. Y. Li, “Reinforcement learning based cooperative coded caching under dynamicpopularities in ultra-dense networks,”
IEEE Trans. Veh. Technol. , vol. 69, no. 5, pp. 5442–5456, May 2020.[105] S. Ahangary, H. Chitsaz, M. J. Sobouti, A. H. Mohajerzadeh, M. H. Yaghmaee, and H. Ahmadi, “Reactive caching ofviral content in 5G networks,” in
Proc., 3rd International Conference on Advanced Communication Technologies andNetworking (CommNet) . IEEE, Sep. 2020.[106] S. Murugaveni and K. Mahalakshmi, “Optimal frequency reuse scheme based on cuckoo search algorithm in li-fi fifth-generation bidirectional communication,”
IET Commun. , vol. 14, no. 15, pp. 2554–2563, Sep. 2020.[107] H. Kazemi, M. Safari, and H. Haas, “A wireless backhaul solution using visible light communication for indoor li-fiattocell networks,” in
Proc., IEEE International Conference on Communications (ICC) , May 2017.[108] ——, “A wireless optical backhaul solution for optical attocell networks,”
IEEE Trans. Wireless Commun. , vol. 18, no. 2,pp. 807–823, Feb. 2019.[109] ——, “Bandwidth scheduling and power control for wireless backhauling in optical attocell networks,” in
Proc., IEEEGlobal Communications Conference (GLOBECOM) , Dec. 2018.[110] ——, “Multi-hop wireless optical backhauling for LiFi attocell networks: Bandwidth scheduling and power control,”
IEEE Trans. Wireless Commun. , vol. 19, no. 9, pp. 5676–5691, Sep. 2020.[111] V. Jamali, D. S. Michalopoulos, M. Uysal, and R. Schober, “Link allocation for multiuser systems with hybrid RF/FSObackhaul: Delay-limited and delay-tolerant designs,”
IEEE Trans. Wireless Commun. , vol. 15, no. 5, pp. 3281–3295, May2016.[112] Z. Gu, J. Zhang, Y. Ji, L. Bai, and X. Sun, “Network topology reconfiguration for FSO-based fronthaul/backhaul in 5G+wireless networks,”
IEEE Access , vol. 6, pp. 69 426–69 437, 2018.[113] M. Atakora and H. Chenji, “A multicast technique for fixed and mobile optical wireless backhaul in 5G networks,”
IEEEAccess , vol. 6, pp. 27 491–27 506, 2018.[114] Y. Li, N. Pappas, V. Angelakis, M. Pioro, and D. Yuan, “Optimization of free space optical wireless network for cellularbackhauling,”
IEEE J. Sel. Areas Commun. , vol. 33, no. 9, pp. 1841–1854, Sep. 2015.[115] Z. Ghassemlooy, S. Arnon, M. Uysal, Z. Xu, and J. Cheng, “Emerging optical wireless communications-advances andchallenges,”
IEEE J. Sel. Areas Commun. , vol. 33, no. 9, pp. 1738–1749, Sep. 2015.[116] Y. Wu, M. Pickavet, and D. Colle, “Analysis of the impact of misaligned wireless backhaul links on optical attocellnetworks,” in
Proc., 22nd International Conference on Transparent Optical Networks (ICTON) . IEEE, Jul. 2020.117] M. A. Khalighi and M. Uysal, “Survey on free space optical communication: A communication theory perspective,”
IEEECommun. Surveys Tuts. , vol. 16, no. 4, pp. 2231–2258, 2014.[118] A. Trichili, M. A. Cox, B. S. Ooi, and M.-S. Alouini, “Roadmap to free space optics,”
J. Opt. Soc. Am. B , vol. 37,no. 11, pp. A184–A201, Nov 2020. [Online]. Available: http://josab.osa.org/abstract.cfm?URI=josab-37-11-A184[119] M. Alzenad, M. Z. Shakir, H. Yanikomeroglu, and M.-S. Alouini, “FSO-based vertical backhaul/fronthaul framework for5G + wireless networks,” IEEE Commun. Mag. , vol. 56, no. 1, pp. 218–224, Jan. 2018.[120] A. Perez, A. Fouda, and A. S. Ibrahim, “Ray tracing analysis for UAV-assisted integrated access and backhaulmillimeter wave networks,” in
Proc., IEEE International Symposium on ”A World of Wireless, Mobile and MultimediaNetworks”(WoWMoM) , 2019, pp. 1–5.[121] L. Zhang and N. Ansari, “On the number and 3-D placement of in-band full-duplex enabled drone-mounted base-stations,”
IEEE Wireless Commun. Lett. , vol. 8, no. 1, pp. 221–224, Feb. 2019.[122] E. Kalantari, M. Z. Shakir, H. Yanikomeroglu, and A. Yongacoglu, “Backhaul-aware robust 3D drone placement in 5G+wireless networks,” in
Proc., IEEE International Conference on Communications Workshops (ICC Workshops) , May 2017.[123] C. T. Cicek, H. Gultekin, B. Tavli, and H. Yanikomeroglu, “Backhaul-aware optimization of UAV base station locationand bandwidth allocation for profit maximization,”
IEEE Access , vol. 8, pp. 154 573–154 588, 2020.[124] J. Zhang, F. Liang, B. Li, Z. Yang, Y. Wu, and H. Zhu, “Placement optimization of caching UAV-assisted mobile relaymaritime communication,”
China Communications , vol. 17, no. 8, pp. 209–219, Aug. 2020.[125] A. A. Khuwaja, Y. Zhu, G. Zheng, Y. Chen, and W. Liu, “Performance analysis of hybrid UAV networks for probabilisticcontent caching,”
IEEE Systems Journal , pp. 1–12, 2020.[126] M. Gapeyenko, V. Petrov, D. Moltchanov, S. Andreev, N. Himayat, and Y. Koucheryavy, “Flexible and reliable UAV-assisted backhaul operation in 5G mmWave cellular networks,”
IEEE J. Sel. Areas Commun. , vol. 36, no. 11, pp. 2486–2496, Nov. 2018.[127] A. Almohamad, M. O. Hasna, T. Khattab, and M. Haouari, “Maximizing dense network flow through wireless multihopbackhauling using UAVs,” in
Proc., International Conference on Information and Communication Technology Convergence(ICTC) . IEEE, Oct. 2018.[128] A. Fouda, A. S. Ibrahim, I. Guvenc, and M. Ghosh, “Interference management in UAV-assisted integrated access andbackhaul cellular networks,”
IEEE Access , vol. 7, pp. 104 553–104 566, 2019.[129] ——, “UAV-based in-band integrated access and backhaul for 5G communications,” in
Proc., IEEE Vehicular TechnologyConference (VTC-Fall) , 2018, pp. 1–5.[130] E. Lagunas, S. Maleki, L. Lei, C. Tsinos, S. Chatzinotas, and B. Ottersten, “Carrier allocation for hybrid satellite-terrestrialbackhaul networks,” in
Proc., IEEE International Conference on Communications Workshops (ICC Workshops) , May 2017.[131] B. Di, H. Zhang, L. Song, Y. Li, and G. Y. Li, “Data offloading in ultra-dense LEO-based integrated terrestrial-satellitenetworks,” in
Proc., IEEE Global Communications Conference (GLOBECOM) , Dec. 2018.[132] R. Deng, B. Di, S. Chen, S. Sun, and L. Song, “Ultra-dense LEO satellite offloading for terrestrial networks: How muchto pay the satellite operator?”
IEEE Trans. Wireless Commun. , pp. 1–1, 2020.[133] Y. Hu, M. Chen, and W. Saad, “Joint access and backhaul resource management in satellite-drone networks: A competitivemarket approach,”
IEEE Trans. Wireless Commun. , vol. 19, no. 6, pp. 3908–3923, Jun. 2020.[134] E. Yaacoub and M.-S. Alouini, “A key 6G challenge and opportunity—connecting the base of the pyramid: A survey onrural connectivity,”
Proc. IEEE , vol. 108, no. 4, pp. 533–582, Apr. 2020.[135] A. Talgat, M. A. Kishk, and M.-S. Alouini, “Stochastic geometry-based analysis of LEO satellite communication systems,”
IEEE Commun. Lett. , pp. 1–1, 2020.136] M. Kishk, A. Bader, and M.-S. Alouini, “Aerial base station deployment in 6G cellular networks using tethered drones:The mobility and endurance tradeoff,”
IEEE Veh. Technol. Mag. , vol. 15, no. 4, pp. 103–111, Dec. 2020.[137] M. Tawarmalani and N. V. Sahinidis, “Global optimization of mixed-integer nonlinear programs: A theoretical andcomputational study,”
Math. Program. , vol. 99, no. 3, pp. 563–591, Apr. 2004.[138] J. Czyzyk, M. Mesnier, and J. More, “The NEOS server,”
IEEE Computational Science and Engineering , vol. 5, no. 3,pp. 68–75, 1998.[139] X. Artiga, A. Perez-Neira, J. Baranda, E. Lagunas, S. Chatzinotas, R. Zetik, P. Gorski, K. Ntougias, D. Perez, andG. Ziaragkas, “Shared access satellite-terrestrial reconfigurable backhaul network enabled by smart antennas at mmWaveband,”