Content uploaded by Aysha Alsayed
Author content
All content in this area was uploaded by Aysha Alsayed on Jan 14, 2019
Content may be subject to copyright.
IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 65, NO. 1, JANUARY 2017 257
Interference and Resource Management Through
Sleep Mode Selection in Heterogeneous Networks
Aysha Ebrahim, Student Member, IEEE, and Emad Alsusa, Senior Member, IEEE
Abstract— This paper investigates the capacity and energy
consumption metrics of small-cell networks that are enabled with
sleep mode (SL) functionality. A novel method is introduced to
systematically and accurately identify the potential SL cells that
can maximize the spectrum reuse efficiency without the need for
an exhaustive search. The performance of the proposed technique
is assessed and compared with the always-on approach and an
optimal benchmark. The results show that the proposed method
significantly outperforms the always-on system and approaches
the performance of the optimal benchmark with notably reduced
computational burden.
Index Terms—Femtocell, heterogeneous network, interference
avoidance, LTE, OFDMA, resource management, sleep mode.
I. INTRODUCTION
THE VAST requirement for wireless data and voice has
increased the pressure on the radio access links and back-
haul infrastructure of cellular networks. This poses a major
challenge for the operators who serve a growing number
of subscribers. While Long-Term Evolution (LTE) systems
can provide higher data rates than 3G, the expected future
capacity requirements cannot be satisfied even though LTE
enjoys a wider bandwidth. Several solutions were introduced
to enhance the capacity of such networks including adding
more macrocells and cell splitting. However, those solutions
are not cost effective and require high Capital Costs (CAPEX)
and Operational Costs (OPEX). An alternative approach is to
apply cell-densification within heterogeneous networks, where
the number of low-cost cells, such as pico- and femto-cells,
deployed within the coverage area of macrocell networks is
increased to enhance capacity and coverage [1].
However, the unplanned distribution of small cells poses
many threats for the network operators. In particular, cross-tier
and co-tier interference is a major challenge that results from
the unplanned deployment [2]. Co-tier interference results
from the densification of small cells which generates a large
number of overlapped cell boundaries. On the other hand,
excessive cross-tier interference caused by the macrocell can
Manuscript received November 23, 2015; revised April 6, 2016,
August 6, 2016, and October 16, 2016; accepted October 20, 2016. Date
of publication November 1, 2016; date of current version January 13, 2017.
The associate editor coordinating the review of this paper and approving it
for publication was M. C. Gursoy.
The authors are with the Department of EEE, The Uni-
versity of Manchester, Manchester, M13 9PL, U.K. (e-mail:
aysha.ebrahim@postgrad.manchester.ac.uk; e.alsusa@manchester.ac.uk).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TCOMM.2016.2623614
seriously degrade the performance of the low power cells.
To overcome such issues, network operators must utilize
efficient radio resource management (RRM) techniques. Inter-
cell interference coordination is one form of RRM that is used
to control interference by allocating users orthogonally in the
time and frequency domains such that potential interference
is prevented [3]. As interference may degrade the overall
performance of heterogeneous networks, special care should
be given to cross-tier interference mitigation. Fractional fre-
quency reuse (FFR) is one of the popular inter-cell interference
coordination (ICIC) techniques in heterogeneous networks due
to its low complexity and limited cooperation required between
BSs [4].
To achieve further performance gains, new strategies need
to be utilized in combination with RRM techniques. Sleep
mode is a promising solution for improving energy efficiency
and capacity in cellular networks as studies show that around
10-60% energy saving can be achieved with such tech-
niques [5]. A small cell can be set to operate in ready
mode (RE) or sleep mode (SL) [5]. In RE, all the hardware
parts of a small cell are switched on whilst in SL, part of the
hardware components can operate in low power while other
parts are fully switched off. The decision of which parts to be
switched off is determined by the energy saving algorithm [5].
The transition between SL and RE states can be controlled
either by the small cell, the user, or the core network. In small
cell controlled SL mode, the small cell can turn off the pilot
transmission when it senses no active mobile users using a
low-power sniffing capability. In core network controlled SL,
the transition between states is carried out by the core network
via the back-haul. In the case of the user controlled SL a
user can wake-up nearby small cells by broadcasting wake-up
signals.
Recently, SL has been proposed for optimizing various
aspects in cellular networks. In [6], the authors provided a
survey on energy efficient techniques and power consump-
tion models for enhancing the energy efficiency in femtocell
networks. The authors in [7] presented a method for minimiz-
ing energy in heterogeneous networks by proposing a joint
resource partitioning and user association method under the
assumption that both tiers of cells can be switched into SL
for a certain fraction of time/frequency. In [8], an SL based
mechanism for minimizing energy consumption and main-
taining QoS is proposed for heterogeneous networks. In this
technique, macrocells are put into SL and users are offloaded
to neighboring macrocells or small cells. The authors in [9]
1536-1276 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
258 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 65, NO. 1, JANUARY 2017
proposed a method for minimizing total power consumption
in heterogeneous networks by adaptively switching BSs on/off
taking into consideration parameters such as network coverage
and the user traffic rates. The work in [10] introduced an SL
method for studying the impact of switching off BSs on the
capacity and energy efficiency in a homogeneous network of
femtocells. An SL based technique was investigated in [11] to
assess the energy efficiency and the capacity of the system.
Different elements were taken into consideration to test the
QoS including the transmission power of femtocells, the
separation between BSs and the total quantity of femtocells.
The authors in [12] presented a collaborative SL scheme
that focuses on enhancing the energy saving of femtocell
assisted networks by only activating the necessary femtocells
to serve the users requirements. In [13], the authors introduced
a cell selection scheme for scaling down the number of active
femtocells while considering the QoS requirements of users.
A heuristic algorithm is presented in [14] to reduce the power
consumption by setting as many BSs as possible to SL to serve
the users rate requirements. To ensure the minimum number
of cells are active, the BSs offering acceptable service to the
largest number of users are activated first. In [15], the authors
proposed a SL activation method for two-tier cellular networks
to reduce the interference to macrocell users.
The majority of the techniques that employ SL strategies
aim to minimize energy consumption by exploiting the idle
times, when the small cells have no active users, and setting
them to SL mode to save power. In addition to minimizing
power consumption, this paper proposes a novel SL-based
RRM to leverage the capacity gain of cellular heterogeneous
networks employing FFR. The motivation therefore is to
exploit the ad-hoc nature of femtocell deployment and the
limited percentage of associated users, to examine the pos-
sibility of temporary cell switch off as an additional degree of
freedom in the optimization of the available resources. This
is achieved by reducing the interference that is produced by
poorly positioned small cells to allow more effective resource
utilization in that part of the network. A resource partitioning
scheme is incorporated in our work to control the inter-cell
interference by orthogonalizing the interfering cells/users. To
compensate for the reuse reduction that results from orthogo-
nalization, we introduce a small cell selection algorithm which
uses a set of criteria to identify the BSs that are causing the
most reuse reduction and feeds this information to the resource
partitioning scheme to determine whether or not if deactivation
can improve the network resource utilization. It is shown that
the proposed method improves the overall throughput as well
as the energy efficiency performance of the system at low
complexity and allows fast deactivation decisions using limited
information exchanged between the central unit and the BSs.
The performance of the proposed method is analyzed using a
mathematical model which reveals an interesting insight into
the superior performance of such systems and is verified by
a good match between the analytical and simulation results.
The main contributions in this work are:
•Formulation of a novel criterion for identifying interfer-
ence scenarios in which switching off certain small cells
can enhance the network capacity.
•Devised an effective SL-mode based resource allo-
cation scheme with enhanced resource re-utilization
gain.
•Presented a mathematical model that accurately captures
the received interference and derived the upper bound
performance that verify the anticipated gains.
The remaining part of the paper is structured according to
the following. Section II provides a description of the system
model and assumptions. Section III and IV illustrates detailed
discussion of the proposed method. Section V presents a
performance analysis of our proposed method. The simulation
and result discussion are discussed in section VI. Lastly;
concluding remarks are presented in section VII.
II. SYSTEM MODEL AND ASSUMPTIONS
A Downlink OFDMA system is utilized with a heteroge-
neous network with a macrocell and a group of small cells.
Assume Band Kdenote the sets of deployed small cells and
UEs in the network in which B={1,2,..., i,..., j,..., B}
and K={1,2,..., k,..., K}.Leti∈Band k∈Krepresent
any random BS and UE from the sets Band Kand jdenote a
neighboring cell of BS i, where a neighboring BS is a nearby
cell in which a UE kserved by BS ifalls within its coverage
range according to a predefined threshold. Furthermore, let Si
refer to the group of UEs served by base station i. The system
bandwidth is split into Nphysical resource blocks (PRB)
which are assumed to be orthogonal and represented by
N={1,2,..., n,..., N}. Both the macrocell and small
cells utilize a reuse-1 system where all tiers of cells coexist in
the same frequency band. Small cells are assumed to use an
open access policy and are transmitting at a fixed power level.
Moreover, users are associated with cells based on the max-
imum reference signal received power (RSRP) strategy [16].
The macrocell and picocells are connected to the core network
through the radio network controller (RNC) and the femtocells
are linked with the femtocell gateway through the internet. The
Macro- and Pico- cells are managed by the operation and man-
agement unit (OAM) and the femtocell gateway is responsible
for connecting the femtocells to the femtocell management
system (FMS) which also uses the OAM unit to manage
all the femtocells in the system. Therefore, the centralized
control required for all BSs in the heterogeneous network is
provided by both the OMS and FMS [17], [18]. Small cells
can operate either in RE state or SL state where the transition
between states is controlled by the core network using a wake-
up control message via the back-haul [5]. Additionally, users
send Channel Quality Indicator (CQI) reports to their serving
BSs to indicate the quality of the communication channel. The
signal-to-interference-plus-noiseratio (SINR) of user kat PRB
nis determined by [19]
ϒk,n=ρi,k,n.¯
hi,k,n
η0+B
j=1ρj,k,n.¯
hj,k,n
,j= i.(1)
where ρi,k,nand ¯
hi,k,ndenote the received signal power and
the channel gain from BS ito the served user kon PRB n
respectively. ρj,k,nis the interference power received from a
neighboring cell jand η0refers to the AWGN noise.
EBRAHIM AND ALSUSA: INTERFERENCE AND RESOURCE MANAGEMENT THROUGH SM SELECTION 259
III. INTERFERENCE MAP APPROACH
To identify possible conflicts in the system, the interference
map approach in [19] is employed. To build the interference
map, BSs use measurement reports (MRs) acquired from their
UEs and transmit radio resource control (RRC) messages to
the UEs to trigger the MRs. UEs then scan the neighborhoodto
determine the physical cell identity (PCI) of neighboring BSs
as well as their corresponding RSRP, which are reported back
to their serving BSs [16]. Given this information, the local
interference map, ζi„ofBSiis produced as follows [19]
ζi(k,j)=0i,k
j,k>β
1i,k
j,k<β, (2)
where kdenotes a user served by BS iand jis a neighboring
BS of i.βrefers to the signal-to-interference (SIR) threshold,
i,kdenotes the received power from BS ito its connected
UE k,j,krepresents the interference power received by UE
kfrom BS j. The value of the threshold can be selected
depending on the required optimization objective. Generally,
higher threshold values are used for maximizing the QoS while
lower values are used for maximizing the throughput. For
further discussion on this please refer to [19, Sec. V].
When a BS updates its local interference map, the current
measurement, ζi, is sent to the FMS only if the previous
record, ζ
i, is different from the current record [19]
Si
k=1
B
i=1|ζi(k,i)−ζ
i(k,i)|>0,(3)
The universal interference map of the network is managed
by the FMS, where ζ∈RK×Bis expressed as [19]
ζ(k,i)=w,when i is the serving BS of k
ζ(k,j)=1,when j i nter f eres k
ζ(k,j)=0,otherwise.(4)
where wrepresent any positive integer number exclud-
ing0and1.
IV. THE PROPOSED METHOD
The main concept of the proposed scheme is to identify
the scenarios in which deactivating certain small cells can
be beneficial from a network-wide perspective. Towards this
end, an algorithm is proposed to nominate the potential BSs,
(e.g., BS f1from fig. 1), to the centralized resource allocation
scheme to determine if deactivating these cells can result in
improving the network-wide frequency reuse and hence capac-
ity. If some cells are selected for deactivation, the resource
allocation scheme will re-associate their users to neighboring
cells and resolve any remaining co-tier interference instances
through orthogonal resource partitioning. Due to the unplanned
deployment of femtocells, the FMS is used in this algorithm
as a centralized controller.
A. SL for Capacity Maximization
Our objective is to find the set of BSs which, if deac-
tivated, could maximize the total capacity of the network.
Fig. 1. Illustration of cell switch-off scenario.
Therefore, the capacity maximization problem is formulated
in this section. Generally speaking, we presume a BS operates
either in active or sleep mode and the state of a BS is denoted
by the indicator variable φi∈{0,1},whereφi=1 indicates
active state and φi=0 indicates sleep state. The achievable
capacity, ck,nof UE kat PRB nis given by
ck,n=W.log 1+ϒk,n(5)
where Wrefers to the bandwidth of PRB nwhich is
identical for all PRBs. The capacity maximization problem
can be formulated as follows
max
φ,χ
B
i=1
K
k=1
N
n=1
φi.ck,n.χk,n(6)
s.t.
K
k=1
χk,n≤1,∀i,n(7)
N
n=1
χk,n=k,∀k.(8)
where χk,nis an indicator variable that is equal to 1 when
UE kis allocated in PRB nand 0 otherwise. The notation k
refers to the amount of resources assigned to user kwhich is
determined based on sec. IV-C.(7) ensures that each PRB is
allocated to one UE only in the same cell. (8) guarantees that
UE is allocated kPRBs. The problem in (6) is formulated
as an integer linear programming (ILP) problem which is
normally solved by ILP solving methods. Since the running
time of ILP solvers is uncertain, more efficient techniques
are needed to reduce the complexity. We propose a unique
solution that divides the problem into two parts. The first part
is solved by the FMS which uses a set of criteria, to be defined
later, to determine the potential BSs to be deactivated. This
is assisted by a resource partitioning scheme that is used to
determine whether or not switching off the selected BS’s can
lead to increasing the resource utilization. The second part
is the resource allocation within each BS, which is solved
independently by each BS using the CQI reports from users
and the output of the first part including the set of active
260 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 65, NO. 1, JANUARY 2017
BSs and the advised allocated PRBs per user. It is worth
highlighting that resource partitioning involves determining
the amount of resources to allocate to each user to ensure
orthogonality while the resource allocation is performed by the
BSs to allocate their users in the preferred resources according
to the wireless channel information.
B. Small Cell Deactivation Criteria
This section defines the small cell deactivation criteria
that will be used in sec. IV-D to select the candidate BSs
for switch off. The key deactivation tests can be listed as
follows:
Test 1:The number of conflicts (NoC) of a BS, which
is given by the total amount of interfered UEs by BS i,
in addition to the number of i’s connected UEs that are
interfered by nearby BSs. The FMS measures the NoC
associated with each BS based on the interference map ζ,
as follows
NoCi=
K
k=1{ζ(k,i)=1}+
B
j=1
j= i
Si
k=1{ζ(k,j)=1},(9)
where x={a=b}means that the value of xis equal to 1
if a=band is equal to 0 otherwise.
Test 2: The proportion of UEs positioned in the handover
region to the total amount of UEs belonging to the BS, which
is expressed as
Hi=Si
k=1Li
Si,(10)
Where Hirefers to the handover ratio of BS i,Siis the
number of UEs connected to iand Lidenotes the number of
UEs served by iand happen to be located within the region
overlapped with neighboring cells
Li(k,j)=1i,k
j,k<β
0Otherwise,(11)
Test 3:The total number of UEs connected to BS i,Si
Si=
K
k=1{ζ(k,i)=w},(12)
Test 4:The total sum of received signal power of all UEs
connected to BS i, which is given by Pi
Pi=
Si
k=1i,k,(13)
Test 5:The sum of interference power received by the i’s
user from neighboring BSs, Ii
Ii=
Si
k=1
B
j=1
j= ij,k.(14)
C. Resource Partitioning
The resource partitioning scheme is responsible for manag-
ing the inter-cell interference by orthogonalizing the resources
between the conflicting BSs/users [19]. Given the interference
map, ζ, the maximum number of resources that BS ioffers
each of its connected UEs, i, is measured by dividing the
available resources, N,by the total number of detected users.
i=N/K
k=1{ζ(k,i)=1}+
K
k=1{ζ(k,i)=w}.(15)
where the number of detected users include the UEs served
by BS iin addition to the interfered users that are located
in the coverage range of BS iwhich are determined based
on the SIR threshold βas discussed in sec. III. This value is
measured and stored in the vector B∈R1×Bwhere B=
1,...,
i,...,j,...,B.Agivenuserkconnected to
BS imay not necessarily be allocated the maximum number
of resources i. For instance, if jof a neighboring BS
jis less than i, the number of allocated resource to user
kis decreased to reduce the interference to the neighboring
users close from BS i. Therefore, the notation kis used to
denote the number of resources allocated to user kwhich is
determined as follows: (a) kis set to iif jis not less
than i,(b)kis set to jif jis less than ito allow BS
ja wider range of orthogonal resource for its own users where
the vector K=[1,... ,
k···K] stores the number of
resources dedicated for each user.
D. Small Cell Deactivation Algorithm
The small cell deactivation algorithm uses the deactivation
criteria and incorporates the resource partitioning scheme to
switch off the unwanted BSs that when switched off, the
interference reduces, the resource re-utilization improves and
the capacity is increased. This objective is pursued while
taking measures to prevent compromising the outage prob-
ability of the users that are connected to the switched off
cells. The small cell deactivation algorithm is performed as
follows
•Before beginning to execute the deactivation tests, the
initial sum of PRBs allocated to all the users in the system
is determined according to sec. IV-C, to be later compared
with the updated value after the deactivation
tot =
B
i=1
Si
k=1
N
n=1Ak,n.χk,n(16)
The notations tot and
tot are used to denote the current
and previous total utilized resources respectively.
•Since the NoC is one of the parameters that significantly
affects the reuse efficiency as will be shown later in sec.
V, Test 1 is applied first to determine the BS with the
maximum NoC
℘=max(NoC), (17)
where, ℘refers to the selected BS.
EBRAHIM AND ALSUSA: INTERFERENCE AND RESOURCE MANAGEMENT THROUGH SM SELECTION 261
If only one BS is found, this BS is selected for switch
off. Otherwise, in case there are multiple candidates, Test 2
is applied to identify the BS with the maximum handover
ratio
℘=max(H), (18)
•In case none of the UEs of the nominated BS are located
in the handover region of nearby cells, the BS to be
deactivated is selected based on Test 4 in case all BSs
have equal number of users
℘=min(P), (19)
The reason for applying (19) is to ensure that the selected
BS is the one that offers its users less received signal power
than other candidates. This is done to enhance the overall
capacity. Otherwise, Test 3 is applied and the BS with the
minimum connected users is selected to reduce the probability
of dropping UEs
℘=min(S), (20)
•If the nominated BSs have UEs in the handover region
of other cells, the BS that has the largest handover ratio
is selected to minimize the number of dropped UEs. In
case more than one BS have equal handover ratio H,
and all their connected users happen to be located in the
coverage area of nearby cells, the selection is performed
based on Test 5 as follows
℘=max(I). (21)
The reason for executing the test in (21)is to select the
BS with the users that are receiving the maximum inter-
ference from nearby cells, since such UEs are more likely
to achieve better signal quality when handed off to these
cells. Otherwise, the selection decision is performed based
on (19).
•Finally, the selected BS is deactivated and the discon-
nected users are handed off to nearby active BSs with
the maximum received signal power. However, in case a
user is not settled within the coverage range of any cell,
this user will be dropped.
•The interference map, ζis then updated and tot,is
recalculated.
•The updated tot is compared with the previous
tot and
the selected BS ℘is set to SL mode if tot is greater
than
tot. The process then repeats to search for other
BSs for deactivation.
These steps are formulated as a pseudo code structure in
Algorithm1.
E. Resource Allocation
The frequency resource allocation algorithm is implemented
independently by each BS after measuring the amount of
available resources per user as discussed in the previous part.
BSs use the CQI reports from their UEs in order to select the
preferred PRBs for each user [20]. Once the preferred PRBs
are determined, the allocation map, A, that is managed by the
BSs is updated.
Algorithm 1 Small Cell Deactivation Algorithm
//Initialize
1: Calculate: tot based on (16)
2: do {
3: Set
tot =tot
4: // Apply Test 1
a=max(NoC)
5: if only 1 BS found in athen
6: ℘=a
7: else
8: // Apply Test 2
9: if H=0 then
10: if candidate cells have different Sthen
11: //Apply Test 3
12: ℘=min (S)
13: else
14: //Apply Test 4
15: ℘=min (P)
16: end if
17: else
18: // Apply Test 2
19: From list of BSs in a, calculate
b=max(H)
20: if only 1 BS have max Hthen
21: ℘=b
22: else
23: if all BSs from bhave equal Hthen
24: // Apply Test 5
25: ℘=max(I)
26: else
27: // Apply Test 4
28: ℘=min (P)
29: end if
30: end if
31: end if
32: end if
33: if tot >
tot then
34: Deactivate ℘
35: Update ζand Cellular Association
36: Recalculate tot
37: end
38: } while tot >
tot
V. PERFORMANCE ANALYSIS
In this section, the Probability Density Function (PDF) of
the Downlink Signal-to-Interference Ratio (SIR) is derived
conditionally on the user location. This PDF is used to study
the effect of the SIR threshold on some key parameters in
our system and to verify to our simulation results. While the
model was simplified to a grid model with stationary BSs
positions it also considers the randomness of the user location
to accurately model the received interference.
A. Signal-to-Interference Model
The received signal power from BS, i, to its served UE
kis given by Ptd−ϕ
i,k, where the transmission power of iis
262 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 65, NO. 1, JANUARY 2017
represented by Pt, the distance between iand kis denoted
by di,kand the pathloss exponent is referred to as ϕand
the interference power received from a nearby BS, j, can be
expressed as Ptd−ϕ
j,k. If we assume Ptis identical for all BSs,
user kis not considered interfered by BS jwhen the recieved
interference from this BS is lower than the threshold
Ptd−ϕ
j,k<Ptd−ϕ
i,k
β,(22)
Which can be simplified and rewritten as
di,k
dj,k<1
β1
ϕ
,(23)
Where βdenotes the SIR threshold, di,k/dj,kis the distance
ratio between k,iand jand is denoted by D. Therefore, user
kis detected in the coverage of BS jif 1
β1
ϕ≤D≤1and
is not detected when 0 <D<1
β1
ϕ.
B. Distance Ratio Analysis
Consider a grid model where two BSs denoted by P1and P2
are located at the center of each square. Assuming P1and P2
are positioned on the x,y-plane with P1located at the origin,
and P2located at a fixed distance from P1denoted by ,
at position (xp2,yp2). Assume a user m, is randomly and
uniformly distributed in the circular area of P1whose location
is given by (r,θ). Assuming rand θare both random variables
following a uniform distribution, the pdf of r,fR(r),is
given by
fR(r)=1
R,0<r≤R
0,elsewhere (24)
And the pdf of θ, fφ(θ ) is given by
fφ(θ) =1
2π,0≤θ≤2π
0,elsewhere (25)
Using trigonometric functions, rand θwhich represent
the polar coordinates are transformed to the corresponding
Cartesian coordinates xand yas follows
x=r.cos(θ ) (26)
y=r.sin(θ ) (27)
The random variable θ, can be transformed into the form
cos(θ ) using the following
fZ(z)=fθg−1(z).|dg−1(z)
dz |(28)
which gives
fZ(z)=2.1
2π.|−1
√1−z2|(29)
where the term in (29)is multiplied by 2 for symmetry.
Therefore, fZ(z)can be re-written as
fZ(z)=1
π√1−z2,−1<z<1
0,elsewhere (30)
Similarly, the probability density of the transformation of
sin(θ ),isgivenby
fV(v) =1
π√1−v2,−1<v<1
0,elsewhere (31)
The probability density, fX(x)as stated in (26)can be
expressed as
fX(x)=+∞
−∞
fZ/Xz
x.fZ(x).1
xdx (32)
where X=Zand Z
X=R.From(24)and (30),fX(x)is
given by
fX(x)=1
πR.1
x
R
1
x.1
√1−x2dx (33)
Therefore, the pdf of xcan be written as
fX(x)=⎧
⎪
⎪
⎨
⎪
⎪
⎩
1
πR.{log (R)−log (x),
+log 1+1−x2
R2}−R≤x≤R/{0}
0,elsewhere
(34)
The euclidean distance between P2and mis defined by
G=x2+y2(35)
Where x=x−xp2and y=y−yp2
And the distance ratio between P1,P2and mis given by
D=G/ (36)
Assuming xp2and yp2are equal to a constant value denoted
by a,thePDFofX,fX(x)can be obtained using (28)
and the density function of x,fx(x)can be written as
fX(x)=⎧
⎪
⎪
⎨
⎪
⎪
⎩
1
πR.{log (R)−log (x−a)
+log 1+1−(x−a)2
R2}a−R≤x≤a+R/{a}
0elsewhere
(37)
In a similar fashion, the density function of y,fy(y)is
given by
fY(y)=⎧
⎪
⎪
⎨
⎪
⎪
⎩
1
πR.{log (R)−log (y−a)
+log 1+1−(y−a)2
R2}a−R≤y≤a+R/{a}
0elsewhere
(38)
The joint probability density of xand ycan be
defined as
fX,Y(x,y)=fX(x). fY(y)(39)
The pdf of G,fG(g)as described in (35)can be expressed
as
fG(g)=+∞
−∞ |g
g2−x2|.fX,Yx,g2−x2(40)
EBRAHIM AND ALSUSA: INTERFERENCE AND RESOURCE MANAGEMENT THROUGH SM SELECTION 263
The expressions in (37)and (38)can be simplified into
quadratic form as follows
fX(x)=p1x2+p2x+p3(41)
fY(y)=p1g2−x2+p2g2−x2+p3(42)
Therefore, the pdf of G,fG(g)can be written as
fG(g)=⎧
⎪
⎪
⎨
⎪
⎪
⎩
fG1(g), √2l1<g<l2
1+l2
2
fG2(g), l2
1+l2
2<g<√2l2
0,elsewhere
(43)
Where l1=a−R,l2=a+Rand fG1(g),fG2(g)are
given by (44)and (45)
fG1(g)=2gp2p3g2−l2
1−l1+gp2
2
2g2−2l2
1
+gp1p3⎡
⎣g2arctan ⎛
⎝g2−l2
1
l1⎞
⎠−g2arctan ⎛
⎝
l1
g2−l2
1⎞
⎠⎤
⎦
+gp2
1
8[2l1g2−l2
1g2−2l2
1+g4arctan ⎛
⎝g2−l2
1
g⎞
⎠
−g4arctan ⎛
⎝
l1
g2−l2
1⎞
⎠](44)
fG2(g)=2gp2p3l2−g2−l2
2+gp2
2
22l2
2−g2
+gp1p3⎡
⎣g2arctan ⎛
⎝
l2
g2−l2
2⎞
⎠−g2arctan ⎛
⎝g2−l2
2
l2⎞
⎠⎤
⎦
+gp2
1
8[2l2g2−l2
2l2
2−g2+g4arctan ⎛
⎝
l2
g2−l2
2⎞
⎠
−g4arctan ⎛
⎝
l1
g2−l2
1⎞
⎠](45)
Finally, the expression in (43)can be transformed as
described in equation (36)using (28)in order to obtain the
pdf of the distance ratio, fD(d)as follows
fD(d)=⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩
fD1(d), √2l1/ < g<l2
1+l2
2/
fD2(d), l2
1+l2
2/ < g<√2l2/
0,elsewhere
(46)
Where fD1(d)and fD2(d)are defined as in (47)and (48),
as shown at the bottom of this page.
C. Number of Conflicts Analysis
As the NoC is one of the parameters that has a major
impact on the BS deactivation decision, the effect of the SIR
threshold, β, on the NoC in the system is investigated based
on the analysis in sec. V-A and using the PDF in (46).Ifwe
assume a grid scenario is used with FBSs and SfUEs
connected to BS f. Therefore, the number of interfered users
that are detected in the coverage area of BS fis given by
L1=Sf(F−1)"Pr 1
β1
ϕ
≤Di,j≤1# (49)
eq. (49)can be re-written as
L1=Sf(F−1)1−FDi,j1
β1
ϕ$ (50)
where FDi,jdi,jis the cumulative distribution func-
tion (CDF) of fDi,jdi,j.
The number of f’s users interfered by neighboring cells is
given by
L2=Sf1−FDi,j1
β1
ϕ$ (51)
Where NoCfis equal to L1added to L2. Fig. 2 illustrates
the SIR threshold effect on the NoC under various distances
between the cell of interest and the neighboring cells. The
graph shows a gradual rise in the NoC as the SIR threshold
increases. This is because increasing the threshold increases
fD1(d)=dp2
1
8[2l1(d)2−l2
1(d)2−2l2
1+(d)4arctan⎛
⎝(d)2−l2
1
(d) ⎞
⎠−(d)4arctan⎛
⎝
l1
(d)2−l2
1⎞
⎠]+dp2
2
2(d)2−2l2
1
×dp1p3⎡
⎣(d)2arctan ⎛
⎝(d)2−l2
1
l1⎞
⎠−(d)2arctan ⎛
⎝
l1
(d)2−l2
1⎞
⎠⎤
⎦+2dp2p3(d)2−l2
1−l1(47)
fD2(d)=dp2
1
8[2l2(d)2−l2
2l2
2−(d)2+(d)4arctan⎛
⎝
l2
(d)2−l2
2⎞
⎠−(d)4arctan⎛
⎝(d)2−l2
2
l2⎞
⎠]+dp2
2
22l2
2−(d)2
+dp1p3⎡
⎣(d)2arctan ⎛
⎝
l2
(d)2−l2
2⎞
⎠−(d)2arctan ⎛
⎝(d)2−l2
2
l2⎞
⎠⎤
⎦+2dp2p3l2−(d)2−l2
2(48)
264 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 65, NO. 1, JANUARY 2017
Fig. 2. Impact of SIR threshold on the NoC.
the number of detected interfered users which leads to higher
NoC. It can be also noticed that the NoC is higher when the
neighboring cells are closer to the cell of interest.
D. Reuse Efficiency Analysis
The reuse efficiency of the system is highly influenced by
the SIR threshold, therefore, the effect of the SIR threshold,
β, on the resource utilization of the system is analyzed in
this section. The PDF in (46)is used to study the resource
utilization of the system. Assuming a system with FBSs and
Sfuser per BS, the number of detected users by BS fis
equivalent to
Tf=Sf+Sf(F−1)%1−FDi,j1
βϕ& (52)
The amount of available resources for each user can be
expressed as k=N/Tf, therefore, the allocated resources
per BS is given by f=kSfwhich can be written as
f=N
1+(F−1)%1−FDi,j1
β1/ϕ& (53)
Fig. 3 illustrates the impact of the SIR threshold on the
reuse efficiency. The SIR threshold has major influence on the
NoC which is one of the key parameters that affect the reuse
efficiency of the system. It is shown that increasing the SIR
threshold lowers the reuse efficiency as a result of higher NoC
in the system as shown the previous section.
E. Signaling Overhead
The total amount of information exchange between the FMS
andtheBSsisgivenby
ds=dfi+dc(54)
where the total number of bits required to send the interfer-
ence map from BS fito the FMS is denoted by dfi. Assuming
BS fihas a total of Siusers and Jineighboring BSs, then dfi
is given by
dfi=Si.dm.Ji(55)
Fig. 3. Impact of SIR threshold on resource utilization.
where, dmdenotes the number of bits necessary to encode
the local interference map of each BS. The FMS needs the
local interference map feedback from each BS which can be
used by the system to extract some necessary information such
as the NoC, the handover ratio, H, and the number of served
users, S. In case more than one candidate is detected, the
received power of candidate BSs users needs to be fed back
to the FMS
dc=.Si.dt(56)
where dtdenotes the number of candidate BSs for deacti-
vation and is the total needed bits to send the CQI feedback
which is given by 4 bits. On the other hand, the FMS does
not need to send feedback to the BSs as the allocation process
is performed independently by each BS.
F. Computational Complexity
An exhaustive approach is used as an optimal benchmark
in which the impact of deactivating every BS on the sum of
the available resources is checked using an iterative process.
In contrast, the proposed SL method is capable of selecting
BSs to find the optimal outcome without the need to perform
an exhaustive search. To illustrate this, assume Mdenotes the
total number of BSs, which means that the loop runs for M
times as there are Mdifferent possibilities to test. This shows
that the optimal benchmark approach requires the number
of iterations to be O(M). On the other hand, the proposed
SL method does not need iterations to search for the most
appropriate small cells to be switched off.
VI. RESULTS AND PERFORMANCE ANALYSIS
A. Homogeneous Network Simulation
A 10×10 grid model is used in this simulation to represent
femtocells urban deployment where the size of each apartment
is 10 m×10 m [1]. It is assumed that one femtocell and one
user are randomly and uniformly dropped around the center
of each apartment. Moreover, a number of femtocells are
activated at random apartments following a uniform distrib-
ution. The bit rate is calculated using adaptive modulation
EBRAHIM AND ALSUSA: INTERFERENCE AND RESOURCE MANAGEMENT THROUGH SM SELECTION 265
TAB LE I
SIMULATION PARAMETERS
and coding (AMC) scheme [21] and the pathloss between
femtocellsandUEsisgivenby[19]
PL =PL(d0)+10ϕlog di,k/d0+X+WL (57)
where di,kdenotes the distance between BS iand UE
k,PL(d0)is given by 20 log10 (4πd0/λ)where d0is the
reference distance (1m) and λdenotes the signal wavelength.
A bandwidth of 5 MHz is utilized and the remaining simula-
tion parameters are shown in Table 1.
The energy efficiency of the proposed SL method is mea-
sured using the energy consumption ratio (ECR) model in [22].
To evaluate the ECR of BS i, the consumed power by BS i,
PDLiis divided by the mean data rate per user, Ri
ECRi=PDLi/Ri(58)
The power saving of the proposed SL method is evaluated
by measuring the total consumed power, P[5]
P=NActive.PRE +NSleep.PSL (59)
Where the number of active and sleep femtocells are
denoted by NActi veand NSleep. The powers required to operate
the femtocell in RE and SL are represented by PRE and PSL
respectively [5].
To confirm the validity of our simulation results, we have
evaluated the capacity upper bound of the full spectrum
reuse (FSR) - Always-ON system, which represents a generic
benchmark that does not consider cell switch off. The upper
bound on the capacity, Ccan be obtained using Jensen’s
inequality [23]
C≤W.log2(1+E(γ0)) (60)
where E(γ0)is given by
E(γ0)=
Er−ϕ
0
η0+B
b=1Er−ϕ
b(61)
Proof: See Appendix A.
The upper bound of Ccan be obtained by substituting
(61)in (60). The derived bound is numerically calculated and
validated through comparison with Monte-Carlo simulation
results. The upper bound for the capacity is illustrated in fig. 4
with the assumption that ϕisgivenby2.7,η0is 0.001 watts.
Fig. 4. Upper bound for the capacity of the FSR - always-on system with
a bandwidth of 5 MHz.
The efficiency of our proposed SL method is evaluated
compared to the Manchester technique, which is based on
the method in [19] in terms of using the interference map
which is also applied in our method to indicate the inter-
ference between BSs and users. It uses similar assumptions
as our system except that the proposed SL algorithm is not
applied. Therefore, we call this method the always-ON to
show the gain achieved by our proposed SL with respect to
the always-ON system. Our method is also compared with
an optimal approach which provides an upper bound for the
system performance. Since our objective is to maximize the
overall capacity, the optimal approach evaluates the effect of
switching off each BS in the system on the overall throughput
in each iteration and determines the BS that maximizes the
overall system capacity when switched off. Normally, the first
deactivated BS is the one that is causing the highest NoC in
the network and is more likely to make the maximum impact
when switched off. Therefore the performance of the proposed
method is tested by checking the effect of switching off the
first BS which is illustrated as a benchmark in the results. The
threshold βis assumed to be 20 dB. Because SL techniques
are very widely used for optimizing the power consumption in
cellular systems, the method in [14] discussed in the related
work will be used as a benchmark in our comparison.
The mean data rate and the ECR are illustrated in
figures 5 and 6 at various densities of femtocells. It can be seen
from Fig. 5 that the data rate is significantly improved in our
proposed SL method in comparison to the always-ON method.
This results from the improved reuse efficiency which is an
outcome of reducing the interference to the nearby cells by
minimizing the quantity of unnecessary active femtocells. The
gap in the performance between the proposed SL method and
the always-ON method is noticeably increased as the network
becomes denser with an improvement of 20%, 24.5% and 34%
at the densities of 7, 11 and 17 respectively. Fig. 6 shows that
the proposed method is capable of preserving the improvement
in data rate as it can be noticed that the ECR gap is increased
when comparing the proposed and the always-ON methods
even at higher densities. Since the objective of the benchmark
266 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 65, NO. 1, JANUARY 2017
Fig. 5. Data rate per user performance comparison with varying femtocell
densities.
Fig. 6. Energy efficiency performance comparison with varying femtocell
densities.
Fig. 7. Power saving performance comparison with varying femtocell
densities.
method in [14] is to minimizethe energy consumption, the data
rate performance appears to be lower than the other methods.
Fig. 7 illustrates a power consumption comparison between
the proposed SL method and the other methods. Clearly, the
Fig. 8. Femtocell network average throughput CDF with 7 deployed cells.
proposed method consumes less power than the always-ON as
the overall power consumption is reduced when some hard-
ware parts such as the power amplifier are switched off during
SL [5]. On the other hand, the benchmark method achieves
lower power consumption compared to the other methods
as it switches off more BSs. Furthermore, the performance
of the benchmark FSR - Always-On method is included in
fig. 5, which is seen to approach the performance of the
Always-On method as the density of cells increases. This is
because the FSR - Always-ON method allows BSs to utilize
the entire resources where the inner users that are not exposed
to interference can achieve very high data rate as opposed to
the Always-ON method, which takes the inter-cell interfer-
ence into account. Fig. 8 illustrates the average throughput
cumulative distribution function (CDF) assuming the number
of femtocells deployed in the system is 7 femtocells. It is
shown that applying the proposed SL method can remarkably
improve the throughout performance with approximately 30%
increase compared to the always-ON method. This results from
switching off the unnecessary femtocells that are causing high
amount of disturbance in the surrounding environment which
leads to significant reduction in the interference and allows
the neighboring BSs to utilize the frequency resources more
efficiently. Users are associated with BSs providing the highest
received signal strength as all femtocells are assumed to use
an open access policy. Fig. 9 shows a comparison between
the three methods in terms of the total number of deactivated
femtocells. Fig. 10 shows the ratio of dropped users vs. the
total number of UEs in the system where it is shown that the
ratio slightly increases at higher user densities as the proposed
SL method aims to minimize the number of dropped UEs.
Fig. 11 illustrates the average capacity per UE performance, it
is clear from this figure that the proposed technique provides
an average improvement between 13% - 20% depending on
the number of BSs.
B. Heterogeneous Network Simulation
In this part, we consider a suburban model with a macrocell
overlaid with 10 femtocells deployed in 10m×10mhouses
EBRAHIM AND ALSUSA: INTERFERENCE AND RESOURCE MANAGEMENT THROUGH SM SELECTION 267
Fig. 9. Number of deactivated femtocells Comparison.
Fig. 10. Dropped users ratio vs. the total number of users in the system.
Fig. 11. Average capacity performance with varying femtocell densities.
which are uniformly distributed within the macrocell coverage
area [1]. A total of 30 UEs are deployed with 2/3 of the
UEs generated in hotspots around femtocells within a radius
of 10 m and the remaining UEs are positioned within the
macrocell area in a random and a uniform manner [24].
A system bandwidth of 10 MHz is used in this simulation
Fig. 12. Heterogeneous network average throughput CDF.
with 50 PRBs available for transmission. The macrocell and
femtocells are equipped with two antennas space-time-block-
code (STBC) and the Stanford university interim (SUI) path
loss model for Terrain type C is used to model the pathloss
between the macrocell and all UEs [25]
PLk
M=A+10σlog dM,k
100 +Xf+Xh+s+WL (62)
where A=PL(d0=100),dM,kis the distance between the
macrocell and UE k,WL is the wall loss, σ=3.6−0.005hM+
20
hM+χa0.59, Xf=6logfc
2000 ,Xh=20 log10 2
hkand
s=χb(8.2+χx1.6)where χa,χband χcrepresent Gaussian
random variables. hMand hkrepresent the macrocell and the
receiver heights which are given by 30m and 2m respectively.
FFR is employed to mitigate interference based on [4] where
the coverage area of the macrocell is divided into inner and
outer regions based on the FFR threshold which is presumed to
be 20dB above the noise level. The resources allocated to the
inner and outer regions are N/2andN/2respectively where
denotes the FFR reuse factor which is assumed to be 3.
First, the SIRs to the macrocell and to the closest femtocell
are determined by the UE which are denoted by SIR1 and SIR2
respectively, in case SIR1 is higher than the FFR threshold,
the UE is considered inner and is connected to the macrocell if
SIR1>SIR2 and to the femtocell otherwise. On the other hand,
the UE is categorized as an outer UE if SIR1 is less than the
FFR threshold. The access threshold of the femtocells, μ,is
adjusted to offload the macrocell UEs to the femtocells where
μis set to 12dB. Fig. 12 illustrates the average throughput
CDF to show the performance of the proposed SL method
compared to the always-ON and the benchmark methods
which shows superior performance of the proposed method
as compared to the other techniques. It is noticed that the
performance improvement is less evident in heterogeneous
network compared to the homogeneous simulation. This is
due to the deployment of femtocells which are randomly
positioned in the coverage area of the macrocell which in
turn reduces the overlapped regions causing the number of
deactivated cells to diminish.
268 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 65, NO. 1, JANUARY 2017
VII. CONCLUSION
A novel sleep mode method is presented in this paper to
improve the capacity and the energy efficiency in hetero-
geneous networks. It was shown that performance improve-
ments can be achieved by identifying the small cells that
are positioned at undesirable interference spots and selecting
them for deactivation. The results show that, compared to the
always-ON system, the proposed SL method can increase the
gain by up to 34% and 15.6% in terms of data rate and
the energy efficiency performance, respectively. Furthermore,
the CDF result showed that the throughput is improved by
30% compared with the always-on method. Finally, it was
also shown that the proposed SL technique approaches the
performance of the optimal benchmark with considerably
lower computational complexity.
APPENDIX A
SHANNON CAPACITY UPPER BOUND
Consider a network with a reference BS, B0at the origin
(0,0)and 6 co-channel interferers B1,...BBlocated at fixed
distances from B0. A reference user is located at random
position (r0,θ
0)within the area of B0. The mean capacity
of the reference user is generally expressed as
C=EW.log2(1+γ0)(63)
where the γ0is the SINR which is defined by
γ0=P0.¯
h.r−ϕ
0
η0+B
b=1Pb.¯
h.r−ϕ
b
(64)
where, P0denotes the required transmission power and Pb
is the transmission power of the bth co-channel interferer
whichareassumedtobeidentical.¯
hrefers to Rayleigh fading
channel which is exponentially distributed. r0denotes the
distance between the reference BS and the reference user and
rbis the distance between the reference user and the bth
co-channel interferer. The upper bound on the Shannon capac-
ity, Ccan be obtained using Jensen’s inequality [23]
CW.log2(1+E(γ0)) (65)
Evaluating the expression in (65) requires measuring the
average of E(γ0). From [26], the distribution of the reference
user in the cell-edge region is given by
fr0(r0)=⎧
⎨
⎩
2r0
R2−R2
0,R0<r0<R
2r0
R2
0,0<r0<R0
(66)
where, Rdenotes the cell radius and R0is the radius
of the inner are part of the cell which are given by 10m
and 5m, respectively. Using (66), the average of r−ϕ
0can be
expressed as
Er−ϕ
0=R
R0
r−ϕ
0.2r0
R2−R2
0
.dr0(67)
that gives
Er−ϕ
0=2−R2−ϕ+R2−ϕ
0/R2−R2
0.(ϕ−2)
(68)
The distance between the bth co-channel interferer and the
reference user is given by [26]
rb=D2+r2
0−2.r0.D.cos (θb+φb)(69)
where Dand φbrepresent the distance and the angle
between the B0and the bth interferer respectively, (rb,θ
b)
are the polar coordinates of the bth interferer related to the
reference user. Based on (69), the average of r−ϕ
bcan be
obtained as follows
Er−ϕ
b=ED2+r2
0−2.r0.D.cos (θb+φb)−ϕ
(70)
which can be written as
Er−ϕ
b=R
R02π
0D2+r2
0−2.r0.D.cos (θb+φb)−ϕ
·fr0(r0).f(θb).dr0.dθb(71)
The distribution of fθb(θb)=1
2π,where0≤θb≤2π.
The expression in (71) does not have a closed-form solution
and therefore the average can be obtained using numerical
integration methods. The average of E(γ0)can be obtained
using (61).
REFERENCES
[1] X. Chu, D. Lopez-Perez, Y. Yang, and F. Gunnarsson,
Heterogeneous Cellular Networks: Theory, Simulation and Deployment,
2nd ed. Cambridge, U.K.: Cambridge Univ. Press, 2008.
[2] D. Lopez-Perez, A. Valcarce, G. de la Roche, and J. Zhang, “OFDMA
femtocells: A roadmap on interference avoidance,” IEEE Commun.
Mag., vol. 47, no. 9, pp. 41–48, Sep. 2009.
[3] A. S. Hamza, S. S. Khalifa, H. S. Hamza, and K. Elsayed, “A survey on
inter-cell interference coordination techniques in OFDMA-based cellular
networks,” IEEE Commun. Surveys Tuts., vol. 15, no. 4, pp. 1642–1670,
4th Quart., 2013.
[4] T. D. Novlan, R. K. Ganti, A. Ghosh, and J. G. Andrews, “Analytical
evaluation of fractional frequency reuse for heterogeneous cellular net-
works,” IEEE Trans. Commun., vol. 60, no. 7, pp. 2029–2039, Jul. 2012.
[5] I. Ashraf, F. Boccardi, and L. Ho, “SLEEP mode techniques for small
cell deployments,” IEEE Commun. Mag., vol. 49, no. 8, pp. 72–79,
Aug. 2011.
[6] Y. Li, H. Celebi, M. Daneshmand, C. Wang, and W. Zhao,
“Energy-efficient femtocell networks: Challenges and opportunities,”
IEEE Wireless Commun., vol. 20, no. 6, pp. 99–105, Dec. 2013.
[7] C. Jia and T. J. Lim, “Resource partitioning and user association
with sleep-mode base stations in heterogeneous cellular networks,”
IEEE Trans. Wireless Commun., vol. 14, no. 7, pp. 3780–3793,
Jul. 2015.
[8] P. Ren and M. Tao, “A decentralized sleep mechanism in heterogeneous
cellular networks with QoS constraints,” IEEE Wireless Commun. Lett.,
vol. 3, no. 5, pp. 509–512, Oct. 2014.
[9] C.-Y. Chang, W. Liao, H.-Y. Hsieh, and D.-S. Shiu, “On optimal cell
activation for coverage preservation in green cellular networks,” IEEE
Trans. Mobile Comput., vol. 13, no. 11, pp. 2580–2591, Nov. 2014.
[10] A. Ebrahim and E. Alsusa, “Interference minimization through sleep
mode based resource allocation for future femtocell networks,” in Proc.
IEEE Int. Conf. Commun. (ICC), Jun. 2015, pp. 1679–1684.
[11] M. Wang, X. Zhu, Z. Zeng, S. Wan, and W. Li, “System performance
analysis of OFDMA-based femtocell networks,” in Proc. IET Int. Conf.
Commun. Technol. Appl. (ICCTA), Oct. 2011, pp. 405–410.
[12] Y. Li, Y. Jia, Y. Wang, and Q. Liu, “Collaborative sleeping scheme
for femtocell networks,” in Proc. IEEE Green Comput. Commun.
(GreenCom), Aug. 2013, pp. 142–147.
[13] A. De Domenico, E. Strinati, and A. Duda, “An energy efficient
cell selection scheme for open access femtocell networks,” in Proc.
IEEE 23rd Int. Symp. Pers. Indoor Mobile Radio Commun. (PIMRC),
Sep. 2012, pp. 436–441.
EBRAHIM AND ALSUSA: INTERFERENCE AND RESOURCE MANAGEMENT THROUGH SM SELECTION 269
[14] W. Vereecken et al., “Evaluation of the potential for energy saving in
macrocell and femtocell networks using a heuristic introducing sleep
modes in base stations,” EURASIP J. Wireless Commun. Netw., p. 170,
Dec. 2012.
[15] S. Ali, M. Ismail, and R. Nordin, “Femtocell sleep mode activation
based interference mitigation in two-tier networks,” in Proc. 4th Int.
Conf. Elect. Eng. Informat., (ICEEI), Sep. 2013, pp. 1088–1095.
[16] S.-F. Chou, H.-L. Chao, and C.-L. Liu, “An efficient measurement report
mechanism for long term evolution networks,” in Proc. IEEE 22nd
Int. Symp. Pers. Indoor Mobile Radio Commun. (PIMRC), Sep. 2011,
pp. 197–201.
[17] S. Saunders and et al.,Femtocells: Opportunities and Challenges for
Business and Technology. Hoboken, NJ, USA: Wiley, 2009.
[18] M. Ruggiero and J. Boccuzzi, Femtocells Design Application.NewYork,
NY, USA: McGraw-Hill, 2011.
[19] W. Pramudito and E. Alsusa, “A hybrid resource management tech-
nique for energy and QoS optimization in fractional frequency reuse
based cellular networks,” IEEE Trans. Commun., vol. 61, no. 12,
pp. 4948–4960, Dec. 2013.
[20] N. Kolehmainen, J. Puttonen, P. Kela, T. Ristaniemi, T. Henttonen, and
M. Moisio, “Channel quality indication reporting schemes for UTRAN
long term evolution downlink,” in Proc. IEEE Veh. Technol. Conf. (VTC),
May 2008, pp. 2522–2526.
[21] Technical Specification Group Radio Access Network; Physical Layer
Aspects of UTRA High Speed Downlink Packet Access, document
TR 25.848, 3rd Generation Partnership Project (3GPP), Mar. 2001.
[22] O. Arnold, F. Richter, G. Fettweis, and O. Blume, “Power consumption
modeling of different base station types in heterogeneous cellular
networks,” in Proc. Future Netw. Mobile Summit, Jun. 2010, pp. 1–8.
[23] A. Goldsmith, Wireless Communications. Cambridge, U.K.: Cambridge
Univ. Press, 2005.
[24] Evolved Universal Terrestrial Radio Access (E-UTRA); Further
Advancements for E-UTRA Physical Layer Aspects, document
TS 36.814, 3rd Generation Partnership Project (3GPP), Mar. 2010.
[25] S.-S. Jeng, J.-M. Chen, C.-W. Tsung, and Y.-F. Lu, “Coverage proba-
bility analysis of IEEE 802.16 system with smart antenna system over
stanford university interim fading channels,” IET Commun., vol. 4, no. 1,
pp. 91–101, Jan. 2010.
[26] A. Mahmud and K. A. Hamdi, “A unified framework for the analysis of
fractional frequency reuse techniques,” IEEE Trans. Commun., vol. 62,
no. 10, pp. 3692–3705, Oct. 2014.
Ays ha Ebrahi m (S’12) received the B.Sc. degree
in computer engineering from the University of
Bahrain in 2009. In the same year, she worked as
a Graduate Assistant in the Computer Engineering
Department in the University of Bahrain, where she
participated as a Lecturer. She received the M.Sc.
degree in electronic engineering from the University
of York in 2011. Since September 2012, she has been
pursuing the Ph.D. degree in electrical and elec-
tronic engineering at The University of Manchester.
Her research focuses on radio resource management
techniques for heterogeneous wireless networks.
Emad Alsusa (M’99–SM’02) received the
Ph.D. degree in telecommunications from the
University of Bath in 2000. In 2000, he became
a Post-Doctoral Research Associate with The
University of Edinburgh. He joined The University
of Manchester in 2003 as a full faculty member.
His research interest lies in the area of signal
processing and communication theory with
a focus on the physical and MAC layers of
wireless communication networks for which he
designs advanced algorithms for energy and
spectrum optimization in the presence of interference and various channel
nonlinearities. His research work has so far resulted in well over a 150 journal
and refereed conference publications mainly in the top IEEE venues. He is a
fellow of the U.K. Higher Academy of Education, and has volunteered for
various IEEE activities, including a VTC TPC Co-Chair of the Greencom
Track in VTC 2016 and a General Co-Chair of the Online Green Com
Conference in 2016. He is the co-recipient of the Best Paper Award in the
International Symposium on Power Line Communications in 2014.
