![Graphical illustration of backhaul evolution of 5G small cell networks with wireless backhauls and cloud-RAN architecture [2].](https://www.researchgate.net/profile/Uzma-Siddique-2/publication/283331419/figure/fig2/AS:667669544046597@1536196232505/Graphical-illustration-of-backhaul-evolution-of-5G-small-cell-networks-with-wireless_Q320.jpg)
Content uploaded by Uzma Siddique
Author content
All content in this area was uploaded by Uzma Siddique on Nov 05, 2015
Content may be subject to copyright.
Wireless Backhauling of 5G Small Cells:
Challenges and Solution Approaches
Uzma Siddique, Hina Tabassum, Ekram Hossain, and Dong In Kim
Abstract—Dense deployment of small cells over traditional
macrocells is considered as a key enabling technique for the
emerging 5G cellular networks. A fundamental challenge is
however to provide an economical and ubiquitous backhaul
connectivity to these small cells. There is a wide range of backhaul
solutions that together can address the backhaul challenges of
5G networks. In this context, this article provides an overview
of the different backhaul solutions and highlight the perceived
challenges in backhauling small cells. A qualitative overview
of the existing research studies and their critical assumptions
are then discussed. Next, for backhauling downlink traffic of
a small cell user, we characterize the cellular region in which
the downlink transmission capacity for a user served by a given
half-duplex (HD) small cell becomes limited by the backhaul
link capacity. We then illustrate solution techniques such as
full-duplex backhauling to improve the performance of wireless
backhauling for small cells.
Index Terms—5G small cells, wireless backhauling, half-
duplex communication, in-band full-duplex communication, self-
interference, downlink transmission, user capacity.
I. INTRODUCTION
To enable efficient spectral reuse, massive deployment of
small cells will be a key technique for 5G cellular net-
works [1]. However, provisioning of efficient and economical
backhauling solutions for these small cells is a challenging
problem. By definition, the small cell backhaul connections are
used to (i) forward/receive the end-user (small cell user) data
to/from the core network and (ii) exchange mutual information
among different small cells over X2 interface. The backhaul
evolution for 5G small cells will include wired and wireless
backhauling to and from core network aggregators (e.g., macro
base stations (MBSs)), cooperation through anchor-base sta-
tions (A-BSs)1, multi-hopping at short range links, and cloud-
based architecture as illustrated in Fig. 1. Since the backhaul
requirements can significantly vary depending on the location
of small cells, the cost of implementing backhaul connections,
traffic load intensity of small cells, latency and target quality
of service requirement of the small-cell users, there is not a
single optimal approach for the backhauling of small cells.
Although wired backhaul solutions ensure reliability with
high data rates, the cost of wired connections is highly
dependent on the offered capacity as well as the distance.
Moreover, highly reliable wired backhaul connectivity may
not be necessary for the small cells that are typically serving
a relatively reduced traffic load compared to a macrocell.
1These refer to the BSs with wired connectivity to MBSs and forward
backhaul data from MBSs to small cells wirelessly in downlink.
MBS
Core Network
Baseband Unit
SBS
Wireless Backhaul (F2 )
Wired Connectivity
X2 Interface
E-band Radio
Microwave Frontha ul
Half-Duplex
S
Full-Duplex
Multi-Hopping
Short Range Link
Wireless Access Lin k (F1)
Anchor-BS
Wired
Backhaul
Fig. 1. Graphical illustration of backhaul evolution of 5G small cell networks
with wireless backhauls and cloud-RAN architecture [2].
Nevertheless, the five-nines reliability2and capacity of
wired backhauling can not be completely overlooked. As such,
the backhaul transmission of 5G small cell networks will cer-
tainly leverage on the combination of wired and wireless back-
auling solutions. However, since the reliability, challenges, and
performance of wired backhaul solutions have been quite well-
investigated and the associated cost-complexity trade-offs are
well-known, this article focuses mainly on the investigation
and performance analysis of the wireless backhaul networks.
Wireless backhauling has been recently considered as a vi-
able and cost-effective approach that allows operators to obtain
end-to-end control of their network rather than leasing third
party wired backhaul connections. The key wireless backhaul
solutions leverage on exploiting the millimeter wave (mm-
wave) spectrum in 60 GHz and 70-80 GHz band, microwave
spectrum between 6 GHz and 60 GHz bands, sub 6 GHz
band, TV white spaces, and satellite technologies. However,
an optimal selection of the wireless backhaul solution depends
on the propagation environment as well as a number of system
parameters such as locations and deployment density of small
cells, desired backhaul capacity, interference conditions, cost,
coverage, hardware requirements, and the spectrum availabil-
ity.
In this context, this article focuses on a multi-tier radio
access network (RAN) where the small cells communicate
with the MBSs instead of communicating to a cloud for
2It means that the backhaul link remains reliable for 99.999% of the time.
1
backhauling3. The contributions of the article are listed herein:
•We provide a comprehensive overview of the existing
wireless backhaul solutions and list out their fundamental
features, benefits, drawbacks, application scenarios, and
the implementation challenges. A qualitative overview of
the existing research studies and their critical assumptions
are then discussed.
•Next, we focus on the wireless backhaul solutions and
mathematically characterize the cellular region in which
the downlink transmission capacity of half-duplex (HD)
small cell base station (SBS) becomes limited by its back-
haul capacity. To overcome this limitation, we propose
enabling full-duplex (FD) mode of operation and math-
ematically analyze the interference scenarios in which
the FD mode of operation is a potential solution. We
further investigate the performance enhancements offered
by deploying anchor-BSs in the presence of either HD or
FD SBS.
•The performance gains of the aforementioned solution
techniques are quantitatively analyzed through simula-
tions and insights are extracted related to the scenarios
in which the FD mode and deployment of A-BS are
advantageous. In the HD mode, an SBS operates on
the access link (link between SBS and its user) and the
backhaul link (link between SBS and MBS) in different
time slots, while in the FD mode, an SBS operates (i.e.,
receives and transmits) on both the access and backhauls
link simultaneously.
•Finally, we point out other design considerations that can
potentially tackle the challenges of sub-6 GHz wireless
backhauling in small cell networks.
The application of FD communication for wireless backhaul-
ing of small cells is novel and has not been investigated yet.
Wireless backhauling of small cells using FD communication
results in new types of interferences such as self-interference
and backhaul interference. As such, in the context of wireless
backhauling, the benefits of FD communication over the
traditional HD communication are not evident. It is therefore
crucial to understand the fundamental performance limits of
the FD communication and analyze the scenarios in which FD
mode of operation is advantageous.
II. OVE RVIEW OF EXISTING WIRELESS BACK HAU L
SOLUTIONS AND THE IR KE Y CHALLENGES
This section first provides an overview of existing wireless
backhaul solutions. Challenges related to different backhaul
solutions are then discussed and a brief summary of their
fundamental features, benefits, and drawbacks is listed (in
Table I).
3In a cloud-based architecture, several small cells are connected to a pool
of baseband units (i.e., cloud) using a fronthaul link. Similar to backhaul,
fronthaul can be realized using wired/wireless solutions, such as optical fiber,
microwave, or even mm wave communication [3]
4Spectrum cost also includes the licensing cost.
5The term form-factor refers to the length of the antenna array.
A. Overview of Wireless Backhaul Solutions
1) Sub 6 GHz spectrum: Sub 6 GHz frequencies support
non-line-of-sight (NLOS) propagation and provide ubiquitous
coverage through obstacles. Due to NLOS feature, point-to-
multipoint (P2MP) backhaul connectivity is possible at the
cost of interference. For licensed sub 6 GHz spectrum, the
licensee is responsible to manage the interference within this
spectrum. Moreover, no new hardware is required to man-
age the access and backhaul links. Nonetheless, the wireless
backhaul solution using sub 6 GHz frequencies is highly
vulnerable to the interference, traffic congestions, and it has a
high licensing cost.
2) Microwave spectrum: The microwave frequency range
has been mentioned as 6-60 GHz [4]. The frequencies of
microwave links are typically reported as 10.5, 13, 15, 18, 23,
26, and 32 GHz [5]. However, in a number of countries, these
bands such as 13, 15 and 23 GHz are becoming congested. As
a result, the exploitation of higher frequencies are currently
under consideration. For instance, in the UK, an auction of
spectrum in the 10, 28, 32 and 40 GHz bands was conducted
in 2008 to meet the demand for microwave frequencies [5].
Due to shorter wavelengths, microwave spectrum is suitable
for LOS scenarios with fixed antenna alignments on both
transmitting and receiving ends. Since the signal attenuation
is high in microwave frequencies, they are favorable for
short-range communications, e.g., neighborhood backhauling
in ultra-dense small cell deployment scenarios.
3) Millimeter wave (mm-wave) spectrum: The propagation
properties of mm wave (60 GHz and 70-80 GHz) are attractive
for high capacity short-range links. This mm-wave spectrum
is spacious and can potentially minimize interference with
highly directive narrow beam-width antennas. Nevertheless,
mm-waves are affected by the atmospheric attenuation to a
greater degree compared to lower frequencies. The power
attenuation at 60 GHz is basically due to the oxygen or dry air,
whereas 70-80 GHz is more similar to conventional microwave
where attenuation is mainly caused by water molecules in the
air. As a result, 60 GHz is more heavily attenuated. However,
the license-exempt nature of 60 GHz makes it more cost-
effective from the operators’ perspective.
4) TV white spaces (TVWS): The TV band is divided
into two bands: VHF band (54-60 MHz, 76-88 MHz, 174-
216 MHz) and UHF band (470-698 MHz) [6]. With the
emerging digital TV (DTV) transmissions, large amount of
TV spectrum has become vacant which is referred to as TV
white spaces (TVWS). While TVWS are licensed for TV
transmissions, they can be exploited for backhaul provisioning
to small cells in a cognitive (unlicensed) manner. That is,
the backhaul interference caused to primary TV transmissions
should not exceed a prescribed threshold. The TVWS offers
larger footprint due to their longer wavelengths and unlicensed
nature help minimizing the cost. The channels in the TVWS
offer much better propagation characteristics compared to
low-frequency cellular bands. Nonetheless, the usefulness of
TVWS for small cell backhauling would be strictly limited by
the transmit power and location of primary TV transmitters.
5) Satellite frequency bands: In a satellite backhaul link,
the degree of attenuation due to weather or rain fade would
2
TABLE I
SUMMARY OF THE FUNDAMENTAL BACKH AUL SOLUTIONS FOR 5G SMA LL CE LLS
Backhaul spectrum
features [4]
Benefits Limitations Application
Sub 6 GHz
•800 MHz-6 GHz
•Licensed
•NLOS
•70 Mbps @20 MHz
•Urban: 1.5-2.5 km
•Rural: 10 km @3.5 GHz
•No additional spectrum
•No new hardware required
•Easy O& M
•Wider coverage
•Antenna alignment not re-
quired
•Low attenuation
•Limited spectrum
•High cost spectrum
•Interference issues
•Low mobility scenarios
•Rural and urban areas
•Conversational voice and
video (live streaming),
real-time gaming
Microwave
•6-60 GHz
•Licensed
•LOS
•1 Gbps+
•2∼4 km
•High capacity
•Medium coverage
•High directivity
•Additional spectrum cost4
•Hardware cost
•Require antenna alignment
•High attenuation
•Urban and rural areas
•Real-time as well as non-real
time services
Millimeter wave
•60 GHz
•Unlicensed
•LOS
•1 Gbps+
•∼1 km
•Bulk of unused spectrum
•High capacity
•Low coverage
•High directivity
•Small form-factor5
•Zero spectrum cost
•Noise limited
•Hardware cost
•Multi-hopping required
•High attenuation
•Multiple antennas required
•Require antenna alignment
•Dense urban areas
•Real-time as well as non-real
time services
Millimeter wave
•70-80 GHz
•Light licensed
•LOS
•1 Gbps
•∼3 km
•Bulk of unused spectrum
•High capacity
•Low coverage
•High directivity
•Small form-factor
•Noise limited
•Hardware cost
•Multi-hopping required
•High attenuation
•Multiple antennas required
•Require antenna alignment
•Dense urban areas
•Real-time as well as non-real
time services
TV white space
•600-800 MHz
•Unlicensed
•NLOS
•18 Mbps
•1∼5 km urban
•Antenna alignment not re-
quired
•Wider coverage
•Low attenuation
•Primary user constraints
•Hardware cost
•Opportunistic availability
•Interference issues
•Sparsely populated areas
•Conversational voice and
video (live streaming),
real-time gaming
Satellite
•4-6,10-12, 20-30 GHz
•Licensed
•LOS
•2-10 Mbps downlink
•1-2 Mbps uplink
•Wider coverage
•Supports high mobility
•Ubiquitous coverage
•Additional spectrum cost
•Hardware cost
•Antenna alignment issues
•Jitter, time delay
•Rural and remote areas
•Mobile situations
•Buffered streaming
depend on the frequency band selected. Lower frequency
bands, i.e., 4-6 GHz (also known as C band) are practically
unaffected by weather, while the Ku-band (10-12 GHz) is
slightly more affected. However, the highest currently used
band, i.e., Ka-band (20-30 GHz) could expect up to 24 dB
of rain fade. The main benefit of satellite-based backhauling
is, it becomes possible at any location from where a suitable
satellite is visible and also in high mobility scenarios. By high
mobility scenarios, we refer to those scenarios where the small
cells are located on aeroplanes (business or large-bodied jets),
ships (ranging from large yachts to commercial vessels and
cruise ships) and land deployed “cells on wheels” to provide
extra coverages. In such a case, the wireless backhaul solutions
should be selected such that they are capable of providing
continuous backhaul coverage to these mobile small cells.
Typically, a satellite backhaul for a small cell requires to install
a small parabolic dish and a remote satellite modem. For ships
and airplanes, stabilized antenna systems can be used to point
at the satellite and to switch between different satellites when
moving from one coverage area to another.
B. Key Challenges
1) Outdoor propagation impairments of mm-wave signals:
While the LOS nature of the mm-waves tend to limit the inter-
ference between small cells, the poor penetration (blocking)
through obstacles is a critical problem [7]. For example, a
100 m mm-wave outdoor link requires an additional 32 dB or
more gain to ensure reliable communication compared to an
indoor mm-wave link. To overcome this, large-sized phased-
array antennas can be used. However, the larger antenna arrays
are more sensitive to wind induced misalignments. Further,
due to short links and narrow beams, minor variations of
propagation geometry could result in severe pointing errors
that can degrade the backhaul performance.
2) Multi-hopping in µwave and mm-wave bands: For both
microwave and mm-wave links, a physically clear and un-
obstructed radio path is required between the SBS and its
backhaul gateway. This may require multiple hops to overcome
obstacles in the propagation path. Multi-hop routing can
significantly increase the overall capital expenditure and the
end-to-end delay. Further, the LOS requirements lead to more
complex installation, precise alignment, and commissioning of
3
the equipments compared to NLOS transmitters and receivers.
3) Spectral mask requirements in TVWS: While the TVWS
can be a potential candidate for wireless backhauling, the radio
design and interference threshold at primary user can be a
performance limiting factor. For instance, the spectral mask
requirements, i.e., adjacent channel leakage ratio (ACLR)
and adjacent channel selectivity (ACS) may result in costly
filter designs with increased power consumption [4]. ACLR
requirements impact the dynamic range of the digital-to-
analog converter. On the other hand, the ACS requirements
impact receiver linearity, power consumption and analog-to-
digital converter design. Thus the coexistence issue needs to
be efficiently resolved to enable robust as well as reliable
backhaul operation.
4) Backhaul interference: Wireless backhauling can be in-
band or out-of-band. With the former, the same channel is used
for access and backhaul links. With out-of-band backhauling,
there is no interference between access and backhaul links
due to the transmissions on different bands/spectrum or wired
connectivity. With in-band wireless backhauling, network op-
erators can upgrade their existing networks in a short time and
at a low cost. However, this gives rise to additional sources
of interference, which may severely degrade the benefits
of resource reuse in both transmission and backhaul links.
As such, intelligent cell association and resource allocation
strategies are required that can potentially mitigate backhaul
interference.
5) Backhaul signaling overhead: In dense small cell de-
ployments, the information exchange between small cells and
macrocells, as well as between neighboring small cells would
be much more frequent. This can be a direct consequence of
frequent handovers [8], the execution of interference manage-
ment, load balancing, and energy saving solutions, or other
collaborative communications methods. Moreover, if the joint
processing techniques like coordinated multi-point (CoMP) are
applied, the user data needs to be shared among multiple
BSs [9]. This data exchange can lead to a huge backhaul
signaling overhead. Consequently, small cells should be able
to dynamically manage and activate/deactivate different con-
nections, and adapt according to favorable conditions such that
backhaul signaling overhead can be minimized.
6) Backhaul delay: Transmission through a wireless
medium incurs a delay when retransmissions are required
due to transmission failures. This can happen, for example,
due to interference from concurrent transmissions and due
to channel fading. This delay, referred to as backhaul delay,
can significantly degrade system reliability and the end-user
performance. Therefore, characterizing the wireless backhaul
delay will be crucial while analyzing the performance of
different backhaul solutions [10].
7) Jitter and time delays in satellite backhauling: Satellite
backhauling leads to large time delays due to
•signal propagation delay between the ground station to
the satellite and that between the satellite and the ground
station. This delay can range from 240-260 ms.
•delays due to packetization and processing which can be
in the range of 35-50ms, yielding a typical one-way trip
time of 275-310 ms.
Moreover, the expected variations in delay (jitter) in the uplink
and downlink would be in the range of 5-25 ms and 10-50 ms,
respectively [4].
8) Non-uniform user traffic: A direct consequence of small
cell densification could be non-uniformity of user traffic. Due
to reduced coverage, the number of users in a small cell would
be typically not so large and the traffic load per small cell can
be highly time varying (e.g., due to user mobility). Therefore,
backhaul resource allocation solutions need to be developed
that can adapt to the traffic load conditions in the small cells.
9) Downlink (DL)/uplink (UL) traffic asymmetry: The mo-
bile traffic in the uplink and downlink can be highly asymmet-
ric. The ratio of downlink to uplink traffic varies in the range
from 4:1 to 8:1. The backhaul resource allocation solutions
should be able to exploit this traffic asymmetry to utilize the
backhaul resources efficiently.
III. OVERVIEW OF EXISTING APPROA CHES FOR WIRELESS
BACK HAU LI NG O F SMA LL CE LL S
The major deployment techniques that have been considered
recently to ensure reliable wireless backhauling include the
deployment of aggregator nodes [11], wireless backhaul hubs
with multiple antennas [12], deployment of Type-A relay
systems [13], etc. Several recent studies have focused on
developing efficient backhaul interference management and
delay minimization solutions or performance characterization
of integrated backhaul systems where wired and wireless
backhauls for SBSs can coexist. A qualitative overview of
some of the existing approaches proposed in the state-of-
the-art literature to tackle the key challenges as discussed in
Section II is given in Table II.
A. Deployment-Based Backhauling Solutions
In [11], a two-tier network is considered where MBSs are
connected to the core network and small cells can access
the MBSs wirelessly. Small cells that are unable to access
the MBSs using single-hop wireless links utilize aggregator
nodes (ANs) to ensure backhaul connectivity. A joint cost func-
tion of placing the aggregator nodes, power control, channel
scheduling and routing is formulated which is minimized to
optimize the locations of the ANs. Another two-tier network
underlaid with single-antenna small cells is considered in [12].
The small cells are connected to the core network via wireless
backhaul links. A wireless backhaul hub (WBH) with multiple
antennas is deployed to provide backhauling for the SBSs.
To maximize the deployment benefits, it is desirable for the
WBH to support as many SBSs as possible. As such, given
the service constraints at the SBSs and power constraints at
the WBH, the number of SBSs that can be admitted into the
network is optimized.
B. Flexible Wireless Backhaul (Challenges 8 and 9)
In [13], deployment of Type-A relay is proposed where
the MBS transports backhaul data wirelessly to SBSs. From
a functional perspective, a Type-A relay resembles a user
equipment with relaying capability. However, Type-A relays
4
TABLE II
QUAL ITATIV E OVE RVIE W OFEXISTING WIRELESS BACK HAUL SOLUTIONS FOR 5 G SM ALL CE LLS
Assumptions Benefits Limitations
Deployment-based
backhauling solutions
[11]
•Out-of-band
•µwave backhaul, sub-6 GHz
backhaul
•Uplink, single cell
•Single antenna AN and small
cells
•Two hop network
•Minimized network
operational cost
•Optimal placement of AN
•Computational complexity
•Co-tier and cross-tier interfer-
ences ignored
[12]
•Out-of-band
•Single antenna SBSs and mul-
tiple antenna MBSs
•Fast convergence
•Minimized cost of small cell
backhauls
•Small cell SINR and backhaul
power constraints
•CSI is required
•MBS intercell interference ig-
nored
•Computational complexity
Flexible backhauling
(Challenges 8 and 9)
[13]
•In-band,
•Sub 6 GHz backhaul
•Downlink
•Single antenna small cell and
relay
•Single antenna macrocell
•Low operational cost
•Reduced interference at user
•Improved backhaul capacity
•Optimized location of relays
•Increased interference at SBS
•Co-tier and cross-tier interfer-
ences ignored
•Effect of backhaul delay ig-
nored
Backhaul delay
management solutions
(Challenge 6)
[10]
•Both sub 6 GHz backhaul and
wired backhaul
•Downlink
•Single antenna MBS and SBSs
•Improved delay performance
•Flexible choice of SBS density
•Minimize expected delay and
deployment cost
•MBS intercell interference is
ignored
[14]
•Out-of-band
•Wired backhaul
•Two macrocell, each with sin-
gle antenna MBS
•Overcome the CSI delay
•Optimal beamforming scheme
•Perfect CSI is required
Interference
management
(Challenge 4)
[15]
•In-band
•sub-6 GHz backhaul
•Downlink
•Massive MIMO at MBS
•Single antenna SBSs
•Maximize network sum log-
rate under constraints:
•Optimized bandwidth alloca-
tion for backhauling
•Optimized user association
•Computational complexity
•MBS intercell interference is
ignored
[16]
•In-band
•sub-6 GHz backhaul
•Uplink
•Single antenna MBS and SBSs
•Slow convergence
•MBS inter-cell interference is
ignored
•Interference is managed under
backhaul constraint
•Optimum uplink macrouser
performance
Backhaul signaling
overhead
(Challenge 5)
[9]
•In-band
•sub-6 GHz backhaul
•Downlink
•Multicell
•multi-antenna BSs
•Reduced signaling overhead
•Minimize backhaul user data
•Given QoS and per-BS power
constraint
•Computational complexity
•Backhaul delay is ignored
differ in terms of deployment (e.g., Type-A relays are operator-
deployed), protocol stack, scheduling strategy, transmission
power, etc. A Type-A relay communicates simultaneously with
two BSs by leveraging the downlink resources of MBS as
well as the uplink resources of SBS to effectively increase the
spectral efficiency while addressing the issue of DL/UL traffic
asymmetry. The position of a Type-A relay is optimized to
maximize the backhaul capacity.
C. Backhaul Delay Management Solutions (Challenge 6)
In [10], a tractable analytical model is developed to char-
acterize the average network backhaul delay and the delay
experienced by a typical user in the downlink considering both
wired and wireless backhaul scenarios. The network delay
is further investigated for both the in-band and out-of-band
wireless backhaul scenarios. It is shown that the total aggregate
wired backhaul delay can be minimized for an optimal density
of small cells. On the other hand, for wireless backhauling, it
is not cost-effective to increase the density of SBSs beyond a
certain point. It is thus concluded that deploying dense small
cell networks may not be as effective without comparable
investment in the backhaul network.
The cooperation among small cells requires efficient ex-
change of channel state information (CSI). However, real-time
CSI sharing is a crucial challenge since a user feedbacks CSI
directly to its serving BS only, while any further inter-cell
CSI exchange takes place over backhaul links. In practice,
information exchange over the backhaul links introduces ad-
ditional delays which further degrades the CSI reliability. Any
CSI at transmitter pertaining to an interfering user is subject
5
to a larger delay than that of the served user. In this context,
[14] devise an efficient modified zero forcing beamforming
technique to overcome the CSI discrepancy created by the
backhaul delay.
D. Interference Management (Challenge 4)
In [15], a large-scale MIMO is considered at the MBS to
mitigate intra-cell and inter-cell interferences. On the other
hand, the single-antenna small cell tier relies on large-scale
MIMO links to the MBS for backhauling. A duplex and spec-
trum sharing scheme, which are based on co-channel reverse
time-division duplex (TDD) and dynamic soft frequency reuse
(SFR), are proposed for backhaul interference management. A
joint optimization problem is formulated to optimize backhaul
bandwidth allocation and user association such that the sum
log-rate of the network is maximized.
In [16], an interference management strategy is proposed for
self-organized small cells considering the wired and wireless
backhaul (termed as heterogeneous backhaul) constraints. The
SBSs operate like decode-and-forward relays for the macrocell
users and forward their uplink traffic to the MBS over hetero-
geneous backhauls. Specifically, the users split their uplink
traffic into two parts. The first part is the coarse message
which can only be decoded at the MBS and second part
is the fine message which can be decoded by neighboring
SBSs as well as the MBS. The users select the best SBS and
optimize their transmission strategy while accounting for the
underlying backhaul conditions at the same time. The problem
is formulated as a non-cooperative game and a reinforcement
learning approach is used to find an equilibrium. Using the
proposed approach, the users self-organize and implicitly
coordinate their transmission strategies in a fully distributed
manner while optimizing their utility function which captures
the trade-off between throughput and delay.
E. Millimeter Wave Backhauling (Challenge 1)
Self-backhauled mm-wave small cell network is considered
in [7] where a fraction of SBSs, referred as anchor BSs (A-
BSs), have wired backhaul and the rest of SBSs backhaul
wirelessly to A-BSs. The A-BSs serve the rest of the SBSs in
the network resulting in two-hop links to the users associated
with the SBSs. The uplink and downlink coverage and rate
distribution are characterized. Mm-wave networks in dense
urban scenarios employing high-gain narrow beam antennas
have been shown to be noise-limited for practical BS den-
sities. Consequently, densification of the network improves
the signal-to-interference-plus-noise ratio (SINR) coverage. It
is concluded that increasing the fraction of A-BSs improves
the peak rates in the network, whereas increasing the density
of BSs while keeping the density of A-BSs constant in the
network, leads to saturation of user rate coverage.
F. Backhaul Signaling Overhead (Challenge 5)
When the joint processing technique is applied in the
coordinated multi-point (CoMP) downlink transmissions, the
data for each user needs to be shared among multiple BSs. This
data exchange can lead to a tremendous backhaul signaling
overhead if the number of users is large. To address this
backhaul signaling overhead, multi-cell CoMP network with
multi-antenna BSs and single antenna users is assumed. The
objective is to distribute the user data only to the minimum
number of cooperating BSs, while satisfying the SINR con-
straint of each user. The problem of minimizing backhaul
user data transfer is formulated, which jointly determines the
optimal BS clustering and the transmit beamformers [9].
G. Use of TV White Spaces (Challenge 3)
In [6] use of TV white spaces is proposed to provide a
backhaul network for rural areas and areas with no pre-existing
wired infrastructure. To quantify white space availability in the
considered region, the area is divided into a grid of 5 mi ×
5 mi square cells. Each cell has a radio tower in (or near)
the middle. Each radio tower has 4 sector antennas covering
all directions instead of one isotropic antenna. The distance
between transmitter and receiver is 5 miles. This model
allows more concentrated line-of-sight (LOS) transmission and
less interference. Achievable capacity is derived using FCC
power limits and widely accepted propagation models. Traffic
demand per cell is derived using a Cisco data traffic survey.
In the rest of the article we will focus on RAN spectrum6
(sub 6 GHz)-based wireless backhauling for small cells, iden-
tify the limitation of traditional HD wireless backhauling, and
introduce the idea of wireless FD backhauling for small cells.
IV. DESIGN GUIDELINES TO OVE RC OM E TH E
LIM ITATIONS OF WIRELESS-BACK HAU LE D SMA LL CE LL S
In this section, considering traditional HD SBS, we first
theoretically characterize the cellular region boundary beyond
which the downlink transmission capacity of a user served
by a given small cell becomes limited due to the backhaul
link capacity. This region is referred to as “backhaul-limited”
region in which the transmission link capacity cannot be
improved any further. As the distance between MBS and SBS
increases, the received signal power at the SBS decreases due
to path-loss which limits the backhaul link capacity and in
turn the transmission capacity for a user in the downlink.
For a clear exposition, we do not consider the shadowing
and fading effects in the propagation model. Note that the
distance between the MBS and the SBS is the root cause
of this backhaul limitation. We demonstrate the usefulness of
FD transmission in reducing the backhaul-limited areas under
certain interference conditions (as shown in Fig. 2(a)). We
then quantitatively analyze the benefits of FD transmission
and deployment of anchor base-stations (A-BSs) to enhance
the backhaul experience of a typical HD small cell. A-BSs
are connected to a MBS through wired connection while
providing wireless backhauling for the HD or FD SBSs. The
transmission power of an A-BS is considered to be same as
that of an SBS, i.e., Pa=Ps.
6The spectrum for cellular RANs is a part of the sub 6 GHz band. For
instance, bands including 450-470 MHz, 698-960 MHz, 1.710-2.025 GHz
have been identified for International Mobile Telecommunications-Advanced
(IMT-Advanced) system in the Radio Regulations (RR) 2008 [17].
6
2
R
m
R
uSI IR >
uSI IR £
(a)
1
R
2
R
1
R
b
hah
CC ,
,
£
b
faf
CC ,
,
£
b
hah
CC
,
,
£
bfaf CC ,, £
o
KR
b
1
1
-
=
o
u
SI
I
R
KR
b
s
s
1
2
2
2
-
÷
÷
ø
ö
ç
ç
è
æ
+
+
=
FD backhaul -limited region
HD backhau l-limited region FD backhau l-limited region
FTTC
SBS
Wireless ba ckhaul link between
MBS and SBS
Core Network
HD access lin k capacity
:
,
ah
C
HD backh aul capacity
:
,bh
C
FD access lin k capacity
:
,af
C
FD backh aul capac ity
:
,
bf
C
Wireless ba ckhaul link between
A-BS and SBS
(b)
A-BS
MBS
m
R
HD backhau l-limited region
Fig. 2. Graphical illustration of the backhaul-limited regions with HD and FD small cells. A demonstration of different backhauling options for SBSs deployed
in the backhaul-limited region, i.e., direct sub 6 GHz-backhauling with MBS and sub 6 GHz-backhauling via A-BS. MBSs are connected to the core network
via fiber-to-the-curb (FTTC).
A. Backhaul-Limited Region for HD SBS
Let us define the HD backhaul and access link capac-
ities as Ch,b =αlog21 + D−βoPm
σ2and Ch,a = (1 −
α)log21 + d−βiPs
σ2, respectively, where Pmrepresents the
transmit power of MBS, βoand βiare path-loss exponents
corresponding to macro cell and small cell propagation envi-
ronments, respectively, Drepresents the distance between the
MBS and the SBS, α= 0.5represents the fraction of time
allocated for backhaul transmission, dis the distance between
SBS and its user, Psis the transmit power of SBS, and σ2is
the noise power.
The achieved downlink transmission capacity of a small
cell user can then be given as Ch,u = min{O × (1 −
α)log21 + d−βiPs
σ2,O × αlog21 + D−βoPm
σ2}, where O
represents the overhead of the backhaul traffic which is typi-
cally generated at S1 interface. Given the definitions of back-
haul and access link capacities, the distance boundary R1at
which Ch,a =Ch,b can be derived as R1= (d−βiPs/Pm)−1
βo.
Beyond R1, the user capacity becomes limited with the
backhaul capacity.
B. Backhaul-Limited Region for FD SBS
Similarly, for FD SBS, the backhaul and access link capac-
ities can be defined as Cf,b = log2(1 + D−βoPm/(RSI +σ2))
and Cf,a = log21 + d−βiPs
Iu+σ2, respectively, where RSI =
Ps/CSI [18], CSI represents the self-interference cancellation
value, and Iuis the backhaul interference received at a user
from the MBS.
The achieved capacity of a small cell user can then be
given as Cf,u = min{O × log21 + d−βiPs
Iu+σ2,O × log2(1 +
D−βoPm/(RSI +σ2))}. Given the definitions of backhaul and
access link capacities, the distance R2at which Cf,a =Cf,b
can be derived as R2=d−βiPsA/Pm−1
βo, where A=
(RSI +σ2)/(Iu+σ2). Beyond R2, the user capacity becomes
limited by the backhaul link capacity.
Remark: It can be observed that the boundary point R2
depends on the value of A, i.e., if A≤1(self-interference
is less than the backhaul interference Iu, i.e., RSI ≤Iu) then
R2< R1, otherwise R2≥R1as illustrated in Fig. 2.
C. Quantitative Analysis
1) HD vs. FD SBS: Fig. 3 demonstrates the backhaul-
limited regions for HD and FD SBSs for a scenario when
A≤1. As expected, due to increase in path-loss, the capacity
of backhaul link decreases with increasing distance Dbetween
the MBS and the SBS. This trend remains valid for both FD
and HD SBS. However, the backhaul link capacity of HD
SBS turns out to be relatively limited compared to FD SBS
due to the orthogonal phases for backhauling and information
transfer.
Conversely, in FD SBS, the attained user capacity is signifi-
cantly reduced compared to backhaul capacity at small values
of D. This is due to the backhaul interference which does not
allow a user to enjoy higher backhaul capacity at small values
of D. Note that, in case of HD SBS, the attained user capacity
is quite close to HD backhaul capacity and it monotonically
decreases with increasing Dwhich is the converse of FD
SBS. The reason is the absence of backhaul interference and
7
50 100 150 200 250 300 350 400 450 500
0.5
1
1.5
2
2.5
3
Distance between MBS and SBS (D) [m]
Downelink User Capacity [bps/Hz]
FD User Capacity
HD User Capacity
FD Backhaul Capacity
HD Backhaul Capacity
R2
R1
FD backhaul− limited
region
HD backhaul−limited
region
Fig. 3. Graphical illustration of the backhaul-limited regions and user capacity
as a function of D(for Rm= 500m, Rs= 40m, βi= 2,βo= 3,
Pm= 5W, Ps= 2W, σ2= 1 ×10−12W/Hz), d= 30m, O= 0.14).
the dominating effect of path-loss. Interestingly, for small
values of D, the user capacity in HD SBS turns out to be
higher than that in FD SBS. However, as Dincreases the
user capacity with FD SBS tends to increase due to reduced
backhaul interference and becomes limited by the backhaul
capacity at a certain point. This is the point beyond which an
increase in the access link capacity will not bring any further
benefits to user capacity.
Remark: These facts motivate the need of adaptive FD in
sub 6 GHz-backhauled small cells that allow SBSs to decide
their mode of operation in an opportunistic manner. Moreover,
the need of other assisting deployments (e.g., anchor SBSs,
relays) or backhaul interference management solutions (e.g.,
power control) becomes evident. As such, we now numerically
investigate the performance gains in the backhaul-limited
regions by employing A-BSs.
2) Deployment of anchor-base stations (A-BSs): The de-
ployment of A-BSs can potentially reduce the backhaul-
limited regions. Fig. 4 demonstrates the impact of deploying
A-BSs on the user capacity considering that A-BSs are placed
around the MBS at a fixed distance ρ. A given SBS selects
the nearest A-BS or MBS for backhauling. The gains of HD
SBS with A-BSs in the backhaul-limited region, especially
in the vicinity of A-BSs, are observed to be significantly
high compared to all other schemes. This is due to the
strong received signal power at HD SBS from A-BS due to
short distance and absence of backhaul interference due to
orthogonal backhauling and transmission time slots. Hence,
the optimal capacity can be achieved at the point ρ, i.e., the
point where the A-BSs are deployed. However, as the distance
between HD SBS and A-BSs increases, the user capacity
degrades.
Conversely, the gains of FD SBS with A-BSs in the
backhaul-limited region, especially in the area far-away from
A-BSs, are observed to be high compared to all other schemes.
The worst-case capacity can be achieved at ρwhich is due
to backhaul interference from the nearest A-BS. In this case,
50 100 150 200 250 300 350 400 450 500
0.5
1
1.5
2
Distance between MBS and SBS (D) [m]
Downlink User Capacity [bps/Hz]
50 100 150 200 250 300 350 400 450 500
0.5
1
1.5
2
FD without A−BS
FD with A−BS
HD without A−BS
HD with A−BS
ρ = 200 m
ρ = 200 m
Fig. 4. Downlink user capacity as a function of distance between MBS and
SBS considering symmetric deployment of A-BSs (for Rm= 500m, Rs=
40m, βi= 2,βo= 3,Pm= 5W, Ps= 2W, Pa= 2W, σ2= 1 ×
10−12W/Hz), d= 30m, O= 0.14).
backhauling through MBS is more feasible. However, as the
distance between FD SBS and A-BSs starts increasing, the
user capacity tends to increase. This is due to reduction in
backhaul interference and gain due to simultaneous backhaul
and information transfer.
It has been observed that the optimal user capacity gains
with HD SBS depend directly on the location of A-BSs. On the
other hand, FD gains are mostly achieved at far-away locations
from A-BSs. It can thus be concluded that deploying A-BSs
help improving the user capacity with HD SBS in the same
region and enhances the user capacity with FD SBS in the far-
away areas. Thus, the use of HD-mode can be recommended
for the SBSs located nearby A-BSs and the FD-mode can be
recommended for the SBSs located far-away from the A-BSs.
D. Other Design Considerations for Wireless Backhauling
1) User association schemes: With the emerging non-
ideal wireless backhaul solutions, the existing user association
schemes such as channel aware, traffic-load aware, channel
access aware schemes may be highly sub-optimal. New user
association criteria are therefore required that perform cell
selection based on the backhaul capacity limitation, backhaul
delays, and backhaul interference in addition to traffic load
and channel conditions. Note that the traffic load conditions
need to be considered now for both the transmission link of
SBS as well as its backhaul link.
2) Resource allocation: Efficient solutions for backhaul
resource allocation need be developed that can adapt accord-
ing to the locations, channel conditions, and traffic load of
different SBSs. Moreover, the backhaul transmissions from
SBSs or MBS should be power adaptive depending on the
required backhaul capacity per small cell. This will minimize
interference while achieving the required backhaul capacity.
In dynamic TDD systems, the spectral efficiency of the small
cell user can be maximized by optimizing the time allocated
for backhaul and transmission given a total time constraint.
3) Massive MIMO for wireless backhauling: To serve mul-
tiple SBSs at a time in the downlink backhauling, the use
of multiple antennas or multiple channels is inevitable. From
operators’ perspective, deploying large antenna arrays at the
8
MBS to serve massive small cell deployments using the same
time-frequency resource could be more attractive than using
multiple set of channels. With such a deployment, the use
of efficient beam-forming techniques can completely cancel
the intra-cell backhaul interference. However, the limitations
of pilot training sequences per coherent time interval restricts
the total number of SBSs served per time-frequency resource.
4) FD SBS with satellite backhaul: The satellite backhaul
is a competitive solution to bring small cell services into
remote and rural areas. FD backhauling can be implemented
at satellite bands by an SBS to serve simultaneously small-
cell users and backhaul data to/from core network via satellite
gateways and in turn satellites. Specifically, in the uplink,
very small transmit power of the user would not have any
impact on the performance of satellite gateway receiver. In
the downlink, the directive feeder antennas at satellite gateway
with possibly additional isolation and processing would cause
negligible interference to the small cell user.
V. CONCLUSION
We have highlighted the primary challenges of wireless
backhauling of small cells in a multi-tier cellular network
where several types of backhaul solutions can coexist. Differ-
ent wireless backhauling options have been compared qualita-
tively. To this end, for a two-tier macrocell-small cell network,
we have characterized the backhaul-limited regions where the
downlink transmission capacity of a small cell user is con-
strained by the transmission capacity of half-duplex backhaul
link between the small cell base station and the macrocell base
station. Solution approaches such as full-duplex transmission
in the backhaul link and anchor-BS deployment have then been
considered and their performance gains have been illustrated
quantitatively. Finally, other possible solution techniques have
been discussed that can potentially improve the performance
of wireless backhauling of small cells.
REFERENCES
[1] “Small cell backhaul requirements,” NGMN Alliance White Paper, June
2012.
[2] “The Benefits of Cloud-RAN Architecture in Mobile Network Expan-
sion,” FUJITSU Network Communications, Inc. White Paper, Aug. 2014.
[3] P. Mugen, W. Chonggang, V. Lau, and H. V. Poor, “Fronthaul-
constrained cloud radio access networks: insights and challenges,” IEEE
Wireless Communications, vol. 22, no. 2, pp. 152–160, April 2015.
[4] “Backhaul Technologies for Small Cells, use cases, requirements and
solutions,” Small Cell Forum, White Paper, Feb. 2013.
[5] “Spectrum and technology issues for microwave backhaul in Europe,”
Innovation Observatory, White Paper, Nov. 2010.
[6] C. Gerami, N. Mandayam, and L. Greenstein, “Backhauling in tv white
spaces,” IEEE Global Telecommunications Conference (GLOBECOM),
2010, pp. 1–6, Dec. 2010.
[7] S. Singh, M. N. Kulkarni, A. Ghosh, and J. G. Andrews, “Tractable
Model for Rate in Self-Backhauled Millimeter Wave Cellular Networks,”
arXiv:1407.5537v1, July 2014.
[8] 3GPP TR 36.842 (V0.2.0), “Study on small cell enhancements for
EUTRA and E-UTRAN - Higher layer aspects (Release 12),” 2014.
[9] J. Zhao, and T. Q. S. Quek, and Z. Lei, “Coordinated Multipoint
Transmission with Limited Backhaul Data Transfer,” IEEE Transactions
on Communications, vol. 12, no. 6, pp. 2762–2775, June 2013.
[10] D. Chen, T. Q. S. Quek, and M. Kountouris, “Backhauling in Heteroge-
neous Cellular Networks Modeling and Tradeoffs,” submitted in IEEE
Transactions on Communications, Jan. 2015.
[11] M. N. Islam, A. Sampath, A. Maharshi, O. Koymen, and N. B. Man-
dayam, “Wireless backhaul node placement for small cell networks,”
48th Annual Conference on Information Sciences and Systems (CISS),
2014, pp. 1–6, Mar. 2014.
[12] J. Zhao, T. Q. S. Quek, and Z. Lei, “Heterogeneous Cellular Networks
Using Wireless Backhaul: Fast Admission Control and Large System
Analysis,” arXiv:1501.06988v1, Jan. 2015.
[13] Y. Shi, M. Li, X. Xiong, and G. Han, “A flexible wireless backhaul solu-
tion for emerging small cells networks,” IEEE International Conference
on Signal Processing, Communications and Computing (ICSPCC), pp.
591–596, Aug. 2014.
[14] X. Yi and D. Kerret, and D. Gesbert, “The DoF of network MIMO with
backhaul delays,” IEEE International Conference on Communications
(ICC), pp. 3318–3322, June 2013.
[15] X. Ge, H. Cheng, M. Guizani, and T. Han, “5G wireless backhaul
networks: challenges and research advances,” IEEE Network, vol. 28,
no. 6, pp. 6–11, Nov. 2014.
[16] S. Samarakoon, M. Bennis, W. Saad, and M. Latva-aho, “Backhaul-
Aware Interference Management in the Uplink of Wireless Small Cell
Networks,” IEEE Transactions on Communications, vol. 12, no. 11, pp.
5813–5825, Nov. 2013.
[17] “Frequency arrangements for implementation of the terrestrial compo-
nent of International Mobile Telecommunications (IMT) in the bands
identified for IMT in the Radio Regulations (RR),” Recommendation
ITU-R M.1036-4, Mar. 2012.
[18] S. Goyal, L. Pei, S. Panwar, R. A. Difazio, Y. Rui, L. Jialing, and E. Bala,
“Improving small cell capacity with common-carrier full duplex radios,”
IEEE International Conference on Communications (ICC), 2014, pp.
4987–4993, June 2014.
9
