QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro Cellular Networks

Abstract

The hierarchical coexistence of femtocells and macrocells is a promising approach for heterogeneous networks (HetNets), dealing mainly with indoor coverage issues and providing high data rates in cellular networks. As for any HetNet, mobility management with handoff issues on top, should be studied prior to hindering the successful deployment of these networks. The current study introduces a new handoff algorithm in hierarchical macro/femtocell HetNets based on the combination of quality of service metrics for efficient network selection including: received signal strength, co-channel interference level, and outage probability of each of femtocell and macrocell networks. The proposed algorithm first collects the measured three mentioned metrics based on mobile station (MS) location, then applies a dynamic weighting system to three-metric sets according to the significance of each metric to obtain one utility for each of femtocell and macrocell networks. The obtained utility is then used as a measure for determining handoff necessity. In order to evaluate the performance of the proposed approach, the paper then introduces analytical model of cell assignment probability for an MS moving from the serving macrocell base station to the target femtocell base station in a two-tier cellular network. The analytical and simulation results indicate the efficiency of the proposed handoff algorithm in comparison with the existing algorithms in terms of cell assignment probability, throughput, number of handoffs, and ping-pong rate.

Full text

QoS-Based Multi-criteria Handoff Algorithm for
Femto-Macro Cellular Networks
Hashem Kalbkhani
1
•Sahar Jafarpour-Alamdari
1
•
Mahrokh G. Shayesteh
1,3
•Vahid Solouk
2
Springer Science+Business Media, LLC 2017
Abstract The hierarchical coexistence of femtocells and macrocells is a promising
approach for heterogeneous networks (HetNets), dealing mainly with indoor coverage
issues and providing high data rates in cellular networks. As for any HetNet, mobility
management with handoff issues on top, should be studied prior to hindering the successful
deployment of these networks. The current study introduces a new handoff algorithm in
hierarchical macro/femtocell HetNets based on the combination of quality of service
metrics for efficient network selection including: received signal strength, co-channel
interference level, and outage probability of each of femtocell and macrocell networks. The
proposed algorithm first collects the measured three mentioned metrics based on mobile
station (MS) location, then applies a dynamic weighting system to three-metric sets
according to the significance of each metric to obtain one utility for each of femtocell and
macrocell networks. The obtained utility is then used as a measure for determining handoff
necessity. In order to evaluate the performance of the proposed approach, the paper then
introduces analytical model of cell assignment probability for an MS moving from the
serving macrocell base station to the target femtocell base station in a two-tier cellular
network. The analytical and simulation results indicate the efficiency of the proposed
&Hashem Kalbkhani
h.kalbkhani@urmia.ac.ir
Sahar Jafarpour-Alamdari
st_s.jafarpour@urmia.ac.ir
Mahrokh G. Shayesteh
m.shayesteh@urmia.ac.ir
Vahid Solouk
v.solouk@it.uut.ac.ir
1
Department of Electrical Engineering, Urmia University, Urmia, Iran
2
Department of IT and Computer Engineering, Urmia University of Technology, Urmia, Iran
3
Electrical Engineering Department, Wireless Research Lab, ACRI, Sharif University of
Technology, Tehran, Iran
123
Wireless Pers Commun
DOI 10.1007/s11277-017-4925-5
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

handoff algorithm in comparison with the existing algorithms in terms of cell assignment
probability, throughput, number of handoffs, and ping-pong rate.
Keywords Femtocell Handoff Received signal strength (RSS) Interference Outage
probability
Abbreviations
AHP Analytical hierarchy process
BS Base station
FBS Femtocell base station
FFR Fractional frequency reuse
FUE Femtocell user equipment
HetNet Heterogeneous network
LTE Long-term evolution
MADM Multi-attribute decision making
MBS Macrocell base station
MS Mobile station
MUE Macrocell user equipment
OP Outage probability
PDF Probability density function
PRWMM Probabilistic random walk mobility model
QoS Quality of service
RSS Received signal strength
SINR Signal-to-interference-plus-noise ratio
SIR Signal-to-interference ratio
1 Introduction
1.1 Motivations
The emerge of wireless cellular networks and their strive on providing high quality and
high data rate services have led the researches toward introducing femtocell as one of the
leading technologies of LTE that deals with coverage issues in indoor environments [1].
Femtocell is a small, low power, and cost effective home base station that is installed by
the consumer to connect to the network through broadband connections such as digital
subscriber line or cable modem [2]. Also, system capacity and QoS enhancements are other
reasons to deploy the femtocells over the macrocell layout in the next generation of mobile
networks. Such HetNets are eventually prone to mobility management issues and handoff
as potential challenges that need to be resolved.
The process of handoff occurs based on MS mobility in both open and hybrid access
femtocells in two-tier femto-macro cellular networks [3]. As a result of maintaining a
standard MS mobility, an ongoing communication may continuously switch between the
macrocell and potential femtocells. However, the user is ultimately expected to attach to a
network with the highest RSS as well as the best QoS. As another boundary, increasing
number of handoffs results in ping-pong effect that in turn leads to increase in delay and
signaling load as an undesired impact on the network. Hence, while an efficient handoff
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

method offers the best available target BS for a requesting user, the method is also
expected to avoid unnecessary handoffs.
The handoffs that mainly occur within a femto-macro cellular network are referred to as
inter-tier handoff and are performed in two modes based on the direction of user move-
ment; either from femtocell to macrocell or vice versa. Due to the distinguishing variations
of RSS, handoffs from femtocell to macrocell can follow traditional algorithms and be
utilized upon and thereby, impose rather less complication on the process. On the contrary,
handoffs from macrocell to femtocell are more challenging because of the small coverage
area of femtocell, low transmission power of FBS in comparison with MBS, and lower
threshold for RSS from FBS (-72 dBm). The concerns on these challenges over the recent
era have led to several investigations for introducing effective handoff decision algorithms.
1.2 Related Works
Due to the significance of the preparation phase of handoff process, majority of related
studies concentrated on handoff decision. RSS-alpha, a handoff algorithm for hierarchical
femto-macro HetNet was proposed in [4]. The study introduces an adaptive factor, a,
determined based on the FBS–MBS separation distance and is used to unify the RSS values
from the both FBS and MBS. The proposed algorithm initially intends to resolve the
problem of connecting to an FBS in the vicinity of MBS. While the study provides a fairly
solid analysis for determining a, the algorithm relies solely on RSS which could barely
satisfy as a distinguishing factor for handoff. Another handoff algorithm called RWTL was
introduced in [5], which takes the RSS and wireless transmission loss parameters as
decision factors to reduce handoff complexity and unnecessary handoffs. In this algorithm,
when the RSS from FBS exceeds the RSS from MBS plus a hysteresis margin, the MS
connects to the FBS. Otherwise, transmission loss is used as handoff parameter and MS
connects to the BS with the minimum transmission loss. The algorithm was compared with
some conventional methods in terms of computational complexity and the number of
handoffs [4]. Although the method of [5] outperforms the algorithm in [4], it suffers from
the increasing number of handoffs when FBS is located in macrocell edge area. The MS
velocity is another parameter that alongside the QoS was considered for handoff decision
in [6]. The evaluations of the performance indicated that the algorithm introduced in [6]
has led to lower rates of unnecessary handoffs as well as lower average number of
handoffs, especially in medium and high velocities of MS, for a small penalty of signaling
overhead. In [7] handoff process between FBSs and the existing MBS was investigated.
The authors proposed a new handoff algorithm based on the signal flow and SIR to reduce
the unnecessary handoffs.
In [8], the problem of joint handoff decision and sub-channel allocation was formulated
as a Markov decision problem. This formulation takes various system parameters into
account such as the movement velocity of MS, required buffer size, cell switching cost,
and call dropping penalty. The optimal solution for cell handoff and channel allocation is
obtained using Q-learning techniques. In [9], a two-fold algorithm for handoff was pro-
posed based on mobility prediction and femtocell capacity estimation. The reported results
indicate that mobility prediction and capacity estimation lead to lower number of handoffs
and lower data delay, respectively. However, this algorithm does not consider the QoS
parameters, therefore, target BS does not necessarily provide the minimum throughput
requirement of user. As part of handoff process, selection of optimal target cell was
modeled as three-dimensional Markov chain in [10]. The algorithm considers some
parameters such as RSS and available bandwidth to reduce the number of handoff failures.
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

Distance-based handoff algorithm was presented in [11] that tries to reduce the handoff
failure and unnecessary handoffs. Based on the reported study, as user enters the overlaid
area, the current distance is estimated using RSS and user velocity. To make handoff
decision, moving distance is compared with pre-defined threshold. The algorithm does not
take into account any QoS factor and cannot guarantee the minimum QoS requirement in
the target cell.
Furthermore, a hybrid algorithm was proposed in [12] to obtain the optimized unique
hysteresis for MS. The study introduces a centralized function to evaluate the overall
handoff performance. The study also proposes a distributed function in user equipment to
determine a unique hysteresis based on the received reference SINR. In a similar study, the
ratio of handoff processing time to the expected resident time in the cell was considered as
the criterion for user handoff among small cells and femtocells [13]. The consideration of
user velocity in the proposed method has led to lower probability of inappropriate hand-
offs. This method does not consider QoS parameters such as interference or SIR and only
considers the velocity. Context parameters such as user velocity, picocell/femtocell radius,
transmit power of main and subsidiary base stations, were the decision criteria to establish
context-aware handoff policy as reported in [14]. Moreover, in the same context, a
probabilistic handoff method was presented in [15] that reduces the number of unneces-
sary handoffs. The presented method estimates the path in a femtocell and makes
a handoff decision based on the available data capacity of a mobile station on the estimated
path. The algorithm does not take into account the minimum acceptable RSS from fem-
tocell. Also, estimating probabilistic path is time-consuming. In [16], the authors proposed
to use the trigger time as a parameter along with the conventional hysteresis to improve the
performance of handoff between FBS and MBS. Their method does not consider any QoS
policy which can degrade the connection in target BS. In [17], the closest FBS to the user is
utilized as the target FBS. User measures the RSS from the surrounding FBSs and if there
is a femtocell with higher RSS than the RSS from the serving cell, user calculates the
distance from femtocell. For shorter distances than the coverage area of the femtocell, the
user chooses the femtocell as the target. The algorithm only considers RSS, and inter-
ference and QoS factors were not taken into account. In [18], the RSS from MBS, the RSS
from FBS, and users’ velocity were employed as inputs of a fuzzy neural network to make
a decision about the target cell. However, QoS factors were not used in the algorithm. Also,
the performance of the algorithm strongly depends on the training of neural network which
requires huge number of data and is time consuming.
1.3 Contributions
As mentioned, in the most previous works, the minimum QoS requirements of user were
not considered in cell selection and handoff process. Also, the problem of user connection
to FBS when the RSS from MBS is higher than that of FBS was not discussed in efficient
manner. Our aim in this study is to increase femtocell assignment probability as well as
satisfying users’ QoS requirements using decision algorithm for femto-macro HetNet
based on a MADM method with the input criteria as OP, co-channel interference, and RSS
that can provide an acceptable determination of the network QoS. These criteria are highly
dependent upon user location so that their impacts on the handoff process vary based on
user mobility. Therefore, we use AHP method [19] to assign appropriate weights to each
criterion and then acquire a handoff utility for each network by applying a weighted sum
method. The target of the optimization method in weighting is to achieve maximum QoS
based on user location. We provide analytical model based on cell assignment probability
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

for a MS moving through straight line from the MBS to the target FBS. We then evaluate
the performance of the proposed algorithm through the analytical model and simulation
results. Also, we assess the performance of the introduced method when MS has random
mobility according to the PRWMM in the presence of several FBSs. The performance
measures used to evaluate the proposed method are cell assignment probability, number of
handoffs, and ping-pong rate. The results show that the proposed method outperforms the
SIR metric and the previous schemes in terms of the mentioned measures.
The rest of the paper is organized as follows. The system model used in this study is
described in Sect. 2. In Sect. 3, the proposed handoff algorithm is presented and analytical
expressions for cell assignment probability are derived. Analytical and simulation results
are provided in Sect. 4. Finally, Sect. 5concludes the paper.
2 System Model
The layout of the macrocell network used in this work is depicted in Fig. 1. The perfor-
mance of this layout has been evaluated for two-tier femto-macro cellular networks and
was reported in [20]. The model is built upon the emerging macrocell architecture known
as FFR [21]. Considering hexagonal macrocells with radius R
m
, each cell is divided into
the inner area with radius Rth and outer area, also divided into six sub-areas. An omni-
directional antenna covers the inner area, while six 60directional antennas (sectors) cover
the sub-areas in outer area. As shown in Fig. 1a, assuming two layers around the central
macrocell (M
0
), the cellular system comprises a total of 19 macrocells in which 6 and 12
macrocells belong to the first and the second layers, respectively.
According to Fig. 1b, the available spectrum is partitioned into seven non-overlapping
sub-bands. Sub-band S
0
is allocated to MUE in the inner area and sub-bands S
1
–S
6
are used
to serve MUE in each sub-area of outer area. To avoid severe cross-tier interference
between macrocell and femtocell networks, the use of different sub-bands for MUE and
FUE is proposed. In this way, each MBS uses whole spectrum to serve its MUE and each
FBS can use portion of spectrum to serve its associated users. The resource allocation
method is described in detail in [20].
The target MBS (the BS pertaining to M
0
) is located in center of the macrocell network,
while FBSs are distributed in the macrocell area within the range p=6hp=6,
henceforth called first sector. The MUE in the first sector is assigned the sub-band S
1
as
depicted in Fig. 1. Therefore, FUE in this region can use the sub-bands S
2
–S
6
to avoid
cross-tier interference. Because of the symmetry in the macrocell layout, the model is
described based on the first sector, so that the results can be generalized to the other
sectors.
The handoff scenario is defined as follows: using a predefined trajectory, a MS passes
through a femtocell coverage area while maintaining an ongoing connection to a macro-
cell, i.e., MS during movement experiences the benefit of femtocell coverage until it leaves
the femtocell and loses the signal. For the sake of initial design, we assume the trajectory
as straight line and the velocity as constant.
Let PT
Mand PT
Fdenote the transmission powers of MBS and FBS, respectively, and PR
M
and PR
Frepresent the RSSs from MBS and FBS, respectively. Then, PR
Mand PR
Fare
obtained as follows:
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

PR
M¼PT
MLMUMð1Þ
PR
F¼PT
FLFUFð2Þ
where L
M
and L
F
are the path losses from the MBS and FBS, respectively, and U
M
and U
F
denote the shadowing coefficients from MBS and FBS, respectively.
The shadowing phenomenon occurs when obstacles obscure the propagation between
BS and MS and is characterized by a lognormal distribution around the local mean with a
typical standard deviation in the range 4-8 dB. The independent variables U
M
and U
F
have
lognormal distribution, therefore, their logarithms are Gaussian with zero mean and
variances d
M
and d
F
, respectively. Assuming ds¼d1d2
jjas the distance step between
M
0
M1,2
M
1,3
M1,4
M
1,5
M
1,6
M1,1
M
2,1
M
2,2
M2,3
M2,4
M
2,5
M
2,6
M2,7
M
2,8
M
2,9 M2,10 M2,11
M
2,12
(a)
S
0
S
1
S
2
S
3
S
4
S
5
S
6
(b)
Fig. 1 Macrocell network layout used in this study. aMacrocell layout pattern. bSpectrum partitioning
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

the two adjacent measurement locations, the autocorrelation of shadowing can be obtained
by using the expectation function as follows [22]:
EU
Md1
ðÞUMd2
ðÞ½¼d2
Mexp d1d2
jj=d0
ðÞ ð3Þ
EU
Fd1
ðÞUFd2
ðÞ½¼d2
Fexp d1d2
jj=d0
ðÞ ð4Þ
where d0denotes the correlation distance (or decay).
To eliminate the abrupt variations of RSS, MS applies exponential averaging window to
the received signal. The impulse response of the window can be expressed as follows [22]:
havg kds
ðÞ¼
ds
davg
exp kds
davg

k0ð5Þ
where kand d
avg
denote the sample number and effective length of the averaging window,
respectively, and d
s
is assumed to be equal to one in this research. Therefore, the smoothed
RSS is calculated as:

PR
Fk½¼PR
Fk½havg k½¼ 1
davg X
k
n¼0
PR
FknðÞexp n
davg
 ð6Þ

PR
Mk½¼PR
Mk½havg k½¼ 1
davg X
k
n¼0
PR
MknðÞexp n
davg
 ð7Þ
where * denotes convolution. The variance of the smoothed RSS can be obtained as [22]:
r2

PR
F¼r2
Fd0
d0þdavg
;r2

PR
M¼r2
Md0
d0þdavg
ð8Þ
The layered architecture of the network imposes considerable interference from the
neighbor MBSs which is dependent upon MS location. Therefore, the received interfer-
ences from other MBSs when MS is connected to central MBS, in the inner and outer
areas, are respectively calculated as:
Iin
M¼X
j¼6
j¼1
PR
Mj;l¼1ðÞþ
X
j¼12
j¼1
PR
Mj;l¼2ðÞ ð9Þ
Iout
M¼PR
Mj;l¼1ðÞþ
X
j2fBMg
PR
Mj;l¼2ðÞ ð10Þ
where PR
Mj;lðÞdenotes the received interference (or RSS) from the jth MBS in the lth layer
and BM
fg
represents the set of interfering MBSs in the second layer that interfer the MBS-
connected MS in the outer area. According to [23], when the macrocell radius is equal to or
higher than 500 m, the interference from the forth and upper layers can be ignored.
Accordingly, in this paper we neglected the interferences from these layers.
Since the transmission power of MBS is considerably higher than that of FBS and due to
the double wall penetration loss between indoor FBSs, we can conclude that the downlink
femto–femto interference is much smaller than the femto-macro interference [24]. Hence,
the downlink femto–femto interference is neglected in our analysis. Accordingly, the
interference from the co-channel MBSs to FUE can be calculated as
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

IF¼IMF ¼PR
Mj;l¼1ðÞþ
X
j2fBFg
PR
Mj;l¼2ðÞ ð11Þ
where BF
fg
denotes the set of the macrocells in the second layer that interfer the FUE.
Similar to the RSS, the received interferences are smoothed, that is:

IMk½¼IMk½havg k½¼ 1
davg X
k
n¼0
IMknðÞexp n
davg
 ð12Þ

IFk½¼IFk½havg k½¼ 1
davg X
k
n¼0
IFknðÞexp n
davg
 ð13Þ
Similar to (8), the variances of the smoothed interferences can be calculated as:
r2

IF¼r2
IFd0
d0þdavg
ð14Þ
r2

IM¼r2
IMd0
d0þdavg
ð15Þ
where r2
IFand r2
IMare the variances of the received interferences of FUE and MUE,
respectively. From (9)to(11), it can be concluded that the received interferences in
different networks are the sum of several log-normally distributed signals. The variance of
the sum of lognormal can be calculated using the moment generating function method [25].
3 Proposed Handoff Algorithm
3.1 Handoff Parameters
The heart of the proposed handoff algorithm is based on the QoS parameters as the metrics
of handoff decision. The introduced QoS factors as well as the method of determining
decision margin improve the handoff process in terms of the number of handoffs and cell
assignment probability that in turn, lead to higher throughput. The RSS is known as the
most common parameter used to make handoff decision and achieves performance
improvement of handoff process. Therefore, this factor is considered to play role in the
proposed algorithm. As extensively discussed in the previous section, interference is also
an important factor for making wise handoff decision. As another important factor, we
employ outage probability as a QoS guarantee of the network. Both RSS and interference
are user-centric factors, measured and determined by MS and depend on the MS location
and its distance from surrounding base stations. In contrast, outage probability is a system-
based factor and can be determined or estimated by system in different MS locations. In
order to exhibit a robust algorithm, we first analyze the impact of each factor as a decision
criterion in different locations of MS in the coverage of macrocell.
Figure 2shows the RSSs from FBS and MBS versus the distance between MS and
MBS. The minimum distance of MS and FBS from MBS which is indicated as the MBS–
FBS separation, is assumed to be 40 m in this study [26]. As Fig. 2a shows, employing
only RSS as decision factor can lead to unbalanced load distribution due to the hetero-
geneity of the BSs. This is specifically worse when MBS–FBS separation is low.
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

There are number of instances that an MS receives acceptable level of RSS from various
BSs. Thus, it is unlikely that RSS be the sole criterion of handoff decision. It was shown on
the other hand that interference can significantly affect the system performance. Hence, the
received interference when MS connects to the MBS and FBS, which is deployed in 40 m
far from MBS, has been calculated and is shown in Fig. 2b. As shown in Fig. 1, the
50 100 150 200 250 300 350 400
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
Distance between MS and MBS (m)
RSS (dBm)
RSS from FBS
RSS from MBS
(a)
50 100 150 200 250 300 350 400
Distance between MS and MBS (m)
-88
-86
-84
-82
-80
-78
-76
-74
-72
-70
Received interference (dBm)
Received interference when MS connects to MBS
Received interference when MS connects to FBS
(b)
Fig. 2 Illustration of RSS and interference in macro-femto HetNet. aRSS from MBS and FBS in different
distances between MS and MBS, bReceived co-channel interference when MS connects to different
networks
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

coverage of macrocell is divided into two regions; inner and outer. According to (9) and
(10), the interfering MBSs when MS connects to MBS in inner and outer regions are
different. When MS in outer region connects to MBS, noting (10) the number of interfering
MBSs reduces which results in interference reduction as depicted in Fig. 2b. In the inner
area, the received interference when MS connects to MBS is higher than the case that it
connects to FBS; while in the outer area, the received interference in the femtocell network
is higher.
As a system-based factor, outage probability is defined as the probability of SIR being
smaller than a pre-defined threshold value cT, i.e.,
OP ¼Pr c\cT
ðÞ¼Pr PR
I\cT
 ð16Þ
where c,P
R
, and Idenote the SIR, RSS, and co-channel interference, respectively. RSS and
interference in different cases are calculated using (1)–(2) and (9)–(11), respectively. OP is
considered as QoS requirement in our algorithm and if we need high QoS, we can assign
greater weight to OP. Using the combination of RSS, interference, and OP to introduce
new handoff metric, makes it similar to the SIR metric. However, the received power and
interference in SIR have equal weights in all regions, while we assign different weights to
them in our method. In Sect. 4, we will show that the proposed metric outperforms the SIR
metric in terms of femtocell assignment probability.
The origins of delay in handoff methods are initially represented as handoff process
latency (switching delay) and data transfer delay. While the former should majorly be
considered in proactive handoffs, the latter is known as delivery delay and affects the
whole process of handoff in both proactive and reactive handoffs. Our method covers the
handoff decision phase and although the decision phase affects the delivery delay of
overall handoff, the study of handoff delay is not feasible in handoff decision phase.
3.2 AHP
The main concept on using these three factors as handoff criteria arises from their non-
linear behavior. Hence, we divide the whole cell area into three non-overlapping regions as
two regions for inner area (first and second regions) and one region for outer area (third
region) as shown in Fig. 3, and determine handoff criteria in each region based on the
weights of each region.
First region
Second region
Third region
Fig. 3 Macrocell coverage
partitioning into three non-
overlapping regions
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

In order to determine the impact of each factor according to the region, we used AHP to
assign relative weights to each factor. AHP [19] is MADM method for determining the
impact of set of attributes, specifically when multiple criteria affect the decision [27].
Figure 4depicts the diagram of decision hierarchy for the proposed work based on AHP.
The scales for the multiple pair-wise comparisons of different criteria are shown in Table 1
[19]. Based on the importance and preference of the metrics in different regions, we
determined the weights through several experiments and the results showed that these
weights result in efficient handoff decision. Based on these scales, AHP is applied to each
of the factors RSS, interference, and OP.
3.2.1 RSS
The significance of RSS is determined based on each of three regions, as shown in the
matrix below. Earlier (Fig. 2) it was shown that in the first and the second regions,
acceptable level of RSS is achieved and hence, a handoff decision is less prone to be
influenced by RSS. However, due to the downfall power of macrocell in the third region,
RSS is expected to play more important role for handoff occurrence.
Weight
RSS First region Second region Third region
First Region
Second region 5/3
Third region 7/3
Sum of column
35 7
13/53/7
15/7
7/5 1
5315/7
ð17Þ
Interference
Second region
Third regi on
Goal
Fir st regi on
RSS
Seco nd reg ion
Third region
First region
OP
Seco nd reg ion
Third region
First regi on
Fig. 4 Decision hierarchy in
AHP method used in our study
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

The matrix is then normalized based on the column-sum which results in
RSS First region Second region Third region
First region 0.2 0.2 0.2
Second region 0.33 0.33 0.33
Third region 0.47 0.47 0.47
ð18Þ
In order to calculate the eigenvectors for the weight matrix, we deducted the following
equation from the formula introduced in [19]as
w¼w1¼1
NX
N
i¼1
M1;i
ðÞ
;...;wN¼1
NX
N
i¼1
MN;i
ðÞ
"#
T
ð19Þ
where Mk;i
ðÞ
denotes the element of the normalized matrix (obtained in the previous step)
in the kth row and ith column, Nis the number of alternatives, and [.]
T
indicates transpose
operation. Hence, the weight vector for RSS is obtained as
wRSS ¼
First region
Second region
Third region
0
@1
A¼
0:20
0:33
0:47
0
@1
Að20Þ
3.2.2 Interference
Similar to RSS, the weight matrix for interference is constructed as
Weight 3 5 7
Interference First region Second region Third region
First region 1 7/5 7/3
Second region 5/7 1 5/3
Third region 3/7 3/5 1
Sum of column 15/7 3 5
ð21Þ
Table 1 Pairwise comparison
matrix Scale Description
1 Very smaller importance of another element
2 Smaller importance of another element
3 Moderate smaller importance of another element
4 Equal Importance of both elements
5 Moderate importance of one element over another
6 Strong importance of one element over another
7 Very strong importance of one element over another
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

Then, it is normalized as
Interference First region Second region Third region
First region 0.47 0.47 0.47
Second region 0.33 0.33 0.33
Third region 0.2 0.2 0.2
ð22Þ
And finally, the weight vector is obtained as
wI¼
First region
Second region
Third region
0
@1
A¼
0:47
0:33
0:2
0
@1
Að23Þ
3.2.3 Outage Probability (OP)
Using the same steps as RSS and interference, the weight vector for OP is obtained as
wOP ¼
First region
Second region
Third region
0
@1
A¼
0:33
0:34
0:33
0
@1
Að24Þ
3.3 Decision Criterion
The decision criterion of the proposed handoff algorithm at the moment kis then deter-
mined by
Ck½¼CFk½CMk½ ð25Þ
where CFk½and CMk½are calculated as below
CFk½¼WRSS
^

PR
Fk½þWI1^

IFk½

þWOP 1OPFk½ðÞ ð26Þ
CMk½¼WRSS
^

PR
Mk½þWI1^

IMk½

þWOP 1OPMk½ðÞð27Þ
In the above equations W
RSS
,W
I
, and W
OP
denote the normalized weights, ^

PR
Fand ^

PR
M
represent the normalized smoothed received RSSs from FBS and MBS, respectively.
Similarly, ^

IMand ^

IFare the normalized smoothed received interferences when MS con-
nects to the FBS and MBS, respectively. Due to the negative impact of the interference and
OP, the values of the related factors are set accordingly. The RSS and interference have
normal distribution, hence, CFk
½
and CMk
½
are normally distributed. Therefore, the mean
and variance of CFk½and CMk½are calculated as
lFk½¼WRSSl^

PR
F
k½þWI1l^

IFk½

þWOP 1OPFk½ðÞð28Þ
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

r2
Fk½¼W2
RSSr2
^

PR
F
k½þW2
Ir2
^

IF
k½ ð29Þ
lMk½¼WRSSl^

PR
M
k½þWI1l^

IMk½

þWOP 1OPM
ðÞ ð30Þ
r2
Mk½¼W2
RSSr2
^

PR
M
k½þW2
Ir2
^

IM
k½ ð31Þ
The random variable C[k]in(25) has also normal distribution, with mean and variance
obtained as follows
lk½¼lFk½lMk½;r2k½¼r2
Fk½þr2
Mk½ ð32Þ
3.4 Performance Analysis: Cell Assignment Probability
In order to analyze the proposed algorithm, we use handoff probability as performance
metric, which is defined as the probability that a requesting MS connects to each network.
More precisely, it can be defined as the probability that the resource required by MS be
assigned by each network. Henceforth, we name this metric as cell assignment probability.
Accordingly, cell assignment probabilities for femtocell (Pfk½) and macrocell (Pmk½)at
the moment kcan be expressed as follows [22,28]:
Pmk½¼Pmk1½1Pfm
jk½

þPfk1½Pmf
jk½ ð33Þ
Pfk½¼Pmk1½Pfm
jk½þPfk1½1Pmf
jk½
 ð34Þ
where Pfm
jk½and Pmf
jk½represent the probability of asigning from MBS to FBS and vice
versa, respectively. Pfm
jk½can be expressed as
Pfmjk½¼PFk½ Mk1½
jfg
¼Pr Fk½ and Mk1½fg
Pmk1½ ð35Þ
where Fk½and Mk½refer to the assignment of MS to FBS and MBS at the kth moment,
respectively. Let Mk
½
¼M1[M2, where M1¼
PR
Fk
½
\72 dBm

,M2¼

PR
Fk½[72 dBm;Ck½\0

and assume M1\M2¼;. Therefore, we have
Pr Fk
½ and Mk1
½
fg¼Pr Fk
½
;M1k1
½
fgþPr Fk
½
;M2k1
½
fg
¼Pr 
PR
Fk½72 dBm , Ck½0;
PR
Fk1½\72 dBm

þPr 
PR
Fk½72 dBm , Ck½0;Ck1½\0

¼Pr 
PR
Fk½72 dBm , 
PR
Fk1½\72 dBm

Pr Ck½0
fg
þPr 
PR
Fk½72 dBm

Pr Ck½0;Ck1½\0
fg
ð36Þ
where Pr Ck½0;Ck1½\0
fg
and Pr 
PR
Fk½72 dBm , 
PR
Fk1½\72 dBm

can
be obtained using the conditional PDFs fCkCk1
jCkCk1
j
ðÞand f
PR
Fk

PR
Fk1

PR
Fk

PR
Fk1


as
follows [13]
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

fCkCk1
jCkCk1
j
ðÞ¼
1
2pr2
Ckk1
j

0:5exp 1
2r2
Ckk1
j
CkmCkk1j

2
!
ð37Þ
f
PR
Fk

PR
Fk1

PR
Fk

PR
Fk1


¼1
2pr2

PR
Fkk1j

0:5exp 1
2r2

PR
Fkk1
j

PR
Fkm
PR
Fkk1j

2
0
@1
Að38Þ
and
mCkk1j¼mCkþqC
rCk
rCk1
Ck1mCk1
ðÞ ð39Þ
r2
Ckk1j¼r2
Ck1q2
C
 ð40Þ
m
PR
Fkk1
j
¼m
PR
Fk
þq
PR
F
r
PR
Fk
r
PR
Fk1

PR
Fk1m
PR
Fk1
 ð41Þ
r2

PR
Fkk1j
¼r2

PR
Fk
1q2

PR
Fk
 ð42Þ
where qCis the correlation coefficient given by qC¼EC
kCk1
½
rckrck1
and mkand rkdenote the
mean and standard deviation, respectively. Now, we compute
Pr Ck½0;Ck1½\0
fg
¼Z0
1
fCk1cðÞQmCkk1j
rCkk1
j
!
dc ð43Þ
The term Pr Ck½\0;Ck1½0
fg
appeared in Pmf
jk½can be computed as
Pr Ck½\0;Ck1½0fg¼
Z1
0
fCk1cðÞ 1QmCkk1
j
rCkk1j
! !
dc ð44Þ
Hence, the handoff probability Pfm
jk½is calculated by substituting (43)in(36). The
analytical model introduced here is used to evaluate the performance of the algorithm, as
we describe in the next section.
3.5 Complexity Analysis
The main computational complexity of the proposed algorithm is related to AHP decision
process. According to [29], if there are klevels, the computational complexity of AHP
algorithm is of order O mkn4
ðÞ, where nand mdenote the number of elements and eval-
uators’ associated matrices, respectively.
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

4 Performance Evaluation
In this section we present the numerical and simulation results to assess the performance of
the proposed handoff algorithm. For the analysis, the macrocell and femtocell radiuses are
set to 500 and 30 m, respectively, and the inner region radius is set to 330 m. FBS and
MBS transmission powers for inner and outer regions are 20, 40, and 43 dBm, respec-
tively; and the threshold value for acceptable RSS from femtocell is considered -72 dBm.
According to (5), we set the effective length of filter to davg ¼20, the correlation distance
to d0¼30 m, and the sampling distance to ds¼1 m. Furthermore, the standard deviations
for shadowing of femtocell and macrocell tiers are considered 4 and 8 dB, respectively.
The path losses from the FBS to indoor MS (LF) and MBS to outdoor MS (LM) are
respectively given by [30]
LF¼38:5þ20 log10 dðÞ ð45Þ
LM¼28 þ35 log10 dðÞ ð46Þ
where ddenotes the transmitter-MS separation. To calculate the path loss from FBS to
outdoor MS and path loss from MBS to indoor MS, 25 dB wall loss is added to (45) and
(46). In the following, we evaluate the proposed algorithm for two different mobility
models.
4.1 Straight Line Mobility
In this mobility model, we assume that one MS leaves the MBS toward a FBS in a straight
line at the velocity 1 m/s.
4.1.1 Femtocell Assignment Probability
Figure 5shows the femtocell assignment probability versus the location of the MS in the
three regions. In order to evaluate the robustness of the proposed algorithm against the
location, each FBS is located in a presumed position within the three regions of macrocell.
The results are compared with three previously presented algorithms namely RWTL [5],
RSS-alpha [4], and the conventional that uses only one parameter, i.e., RSS, as decision
criterion.
As depicted in the figures, the proposed algorithm incurs higher femtocell assignment
probability in coverage of femtocell than the other methods. Also, in outside of the fem-
tocell coverage, the proposed algorithm has lower femtocell assignment probability in
comparison with other algorithms. In nearby locations to MBS, the RSS from MBS is
considerable in the coverage of FBS, which results in low femtocell assignment probability
for conventional algorithm. As distance between FBS and MBS increases, the RSS from
MBS reduces in the coverage of FBS, therefore, the femtocell assignment probability
increases. Also, the femtocell assignment probability of the proposed algorithm has very
low variations when the location of FBS varies in the coverage of macrocell, but these
variations are considerable in other algorithms. In the first and second regions of macrocell
area, our proposed algorithm has higher femtocell assignment probability than the other
algorithm inside the femtocell coverage area. In the third region, the proposed algorithm in
small areas close to femtocell radius has lower femtocell assignment probability than the
other algorithms, but, in the rest of femtocell coverage area has higher or equal assignment
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

probability compared to other algorithms. It is concluded that the overall performance of
the proposed algorithm is better than the other algorithms.
Figure 6compares the analytical and Monte Carlo simulation results when the FBS–
MBS separation is 120 m (first region), 250 m (second region), and 350 m (third region).
We observe that in the distances near to femtocell, there is a little difference between the
analytical and simulations results. This is due to the MFG approximation used to calculate
the sum of lognormals in the received interference. But in the distances far from the
femtocell coverage, the difference is more noticeable due to the approximations used in
calculation of Pr Fk½and Mk1½
fg
.
4.1.2 Number of Handoffs
In another experiment, we repeated the three algorithms for 10
5
iterations simultaneously
and counted the number of handoffs for different FBS locations. Figure 7demonstrates the
results. It is observed that the proposed algorithm performs better than the two algorithms.
Moreover, the proposed method can reach the best case in which there are just two
handoffs and non-redundant handoff in the third area (one from macro to femto when MS
arrives the femtocell and the other one when MS leaves the femtocell).
(a) (b)
(c)
Fig. 5 Femtocell assignment probability versus MS location for different distances between FBS and MBS.
a120 m (first region), b250 m (second region), c350 m (third region)
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

4.1.3 Throughput
We also obtain the throughput of different methods. Throughput is defined as
T¼log21þcðÞ;bits=s=Hz ð47Þ
The results are presented in Fig. 8in term of cumulative distribution of throughput. The
results indicate that the proposed method achieves higher throughput than the conventional
and RSS-alpha methods.
4.1.4 Effect of Velocity
In Sects. 4.1.1–4.1.3, it is assumed that MS leaves the MBS toward a FBS in a straight line
at the velocity 1 m/s. Here, we consider different velocities and obtain the throughput of
MS. The results are shown in Fig. 9in term of cumulative distribution of throughput. It is
observed that that as velocity increases, the throughput of MS reduces. As velocity
increases, the stay time of MS in femtocell coverage reduces, which results in lower
throughput values.
(a) (b)
(c)
Fig. 6 Simulation and analytical results for femtocell assignment probability for different distances
between FBS and MBS. a120 m (first region), b250 m (second region), c350 m (third region)
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

4.2 Analysis of Decision Criterion Against SIR
SIR, as a function of the two metrics, has been extensively used as the sole handoff
criterion in the literature. Since two of the metrics are utilized to determine the decision
criterion introduced in the current paper, we intend to present a comparison of the impacts
of both handoff criteria. SIR is determined as c¼PR=I, with identical weights for RSS and
interference throughout all regions. OP is the third metric to be incorporated and similar to
SIR, it is a function of RSS and interference. However, this criterion differs from SIR from
two perspectives; OP behaves totally differently with the variations of RSS and interfer-
ence, the equations and the corresponding values of OP are different in each region since
they are opted to enhance the femtocell assignment probability. Figure 10 depicts the
femtocell assignment probabilities pertaining to the proposed decision criterion and SIR
metric in different regions with 30 m of coverage for femtocell. As shown, the proposed
decision criterion incurs higher femtocell assignment probability than SIR. In the regions
close to MBS, since the received SIR from macrocell is higher than from femtocell, the
femtocell assignment probability experiences dramatic decrease. As shown earlier in
Fig. 2, the third region is realized with lower interference for macrocell rather than for
femtocell, whereas macrocell transmission power for this region is higher than the second
region. Consequently, the SIR of macrocell in the third region is higher than that in the
second region, which results in lower femtocell assignment probability compared to the
second region.
4.3 Probabilistic Random Walk Mobility Model (PRWMM)
For more practical investigation of the proposed algorithm, we consider the PRWMM for
MS mobility. PRWMM utilizes a probability matrix to determine the position of MS in the
next step [31]. State 0 represents the current position of MS, state 1 indicates the previous
Fig. 7 Average number of handoffs in three different algorithms
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

position, and state 2 denotes the next position. The probability of the next movement
direction is described by the matrix Pas follows:
P¼
P0;0ðÞP0;1ðÞP0;2ðÞ
P1;0ðÞP1;1ðÞP1;2ðÞ
P2;0ðÞP2;1ðÞP2;2ðÞ
2
43
5ð48Þ
where Pi;jðÞis the probability of transition from the state ito the state j. A typical matrix
used in this work is:
P¼
00:50:5
0:20:80
0:200:8
2
43
5ð49Þ
PRWMM uses the previous position to determine the next step and demonstrates that
the probability of continuing to move in the previous path is higher than the probability of
changing the path. For this experiment, we employed more number of FBSs, considered
MBS–FBS and FBS–MBS handoffs, and ignored the sector–sector and FBS–FBS handoffs.
Therefore, femtocells are distributed only in one sector of macrocell layout and are located
far from each other. The snapshot of this scenario is shown in Fig. 11.
(a) (b)
(c)
Fig. 8 Comparison between throughputs of different methods. aFBS location =120 m, bFBS
location =250 m, cFBS location =350 m
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

We calculated the number of handoffs, number of ping-pongs, and ping-pong rate to
evaluate the algorithm based on PRWMM. The number of iterations is 10
5
and the
handoffs with stability time below 5 s are considered as ping-pong. The start point is
selected randomly in the coverage area of the sector. Each of the three algorithms, i.e.,
conventional, RSS-alpha, and proposed were applied simultaneously to have fair com-
parison. The results are given in Table 2.
As mentioned, the convectional algorithm incurs essential problem when handoff is
performed toward femtocell, especially when the FBS is close to the MBS. Therefore, the
number of unnecessary handoffs in the proposed method is less than the two other
methods, while the ping-pong rate is also considerably small. It should be noted that
although RSS-alpha incurs less unnecessary handoffs, it causes high ping-pong rate. The
reason can be seen in the ping-pong rate listed in the table. Based on the table, majority of
handoffs established by RSS-alpha are unstable which cause the increase in the system
complexity as well as ping-pong rate. The results of Table 2indicate the success of the
proposed algorithm in performing necessary handoffs with the minimum ping-pong rate.
According to the graphs presented for femtocell assignment probability, we argue that
these handoffs are necessary handoffs and should occur to increase the efficiency of the
network.
(a) (b)
(c)
Fig. 9 The effect of MS’ velocity on the performance of the proposed method. aFBS location =120 m,
bFBS location =250 m, cFBS location =350 m
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

(a) (b)
(c)
Fig. 10 Performance comparison of the proposed decision criterion and SIR metric for different distances
between FBS and MBS. a120 m (first region), b250 m (second region), c350 m (third region)
150 200 250 300 350
-120
-100
-80
-60
-40
-20
0
20
40
60
80
X (m)
Y(m)
Mobility pattern
FBS
Fig. 11 FBS locations and MS
movement snapshot in PRWMM
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

5 Conclusion
The current paper proposed a new handoff decision algorithm to solve handoff problem in
two-tier femto-macro cellular networks. The proposed handoff decision algorithm uses
multiple factors as decision criteria including RSS, received interference, and outage
probability. We used an MADM method to prioritize each factor based on MS location and
achieved an adaptive handoff criterion. Femtocell assignment probability and number of
handoffs were calculated for performance evaluation of the proposed handoff decision
algorithm. Analytical and simulation results indicate the efficiency of the proposed algo-
rithm in comparison with other methods. The proposed algorithm attempts to cover the
issues of previous works in user handoff from macrocell to femtocell, especially in nearby
locations of the MBS, maintaining the effective ping-pong rate. Due to the minimum QoS
requirement, our algorithm reduces the unnecessary handoffs to avoid system efficiency
degradation. The proposed algorithm also increases the femtocell assignment probability
while decreasing the number of handoffs. Moreover, selecting the RSS threshold for FBS
as a criterion of user assignment to the femtocell, results in high accuracy in link switching
process and thereby, more handoff stability.
References
1. Andrews, J. G., Claussen, H., Dohler, M., Rangan, S., & Reed, M. C. (2012). Femtocells: Past, present,
and future. IEEE Journal on Selected Areas in Communications, 30, 497–508.
2. Chandrasekhar, V., Andrews, J., & Gatherer, A. (2008). Femtocell networks: A survey. IEEE Com-
munications Magazine, 46, 59–67.
3. David, L.-P., Alvaro, V., A
´kos, L., de la Guillaume, R., & Jie, Z. (2010). Intracell handover for
interference and handover mitigation in OFDMA two-tier macrocell–femtocell networks. EURASIP
Journal on Wireless Communications and Networking,2010, 1–15.
4. Moon, J.-M., & Cho, D.-H. (2009). Efficient handoff algorithm for inbound mobility in hierarchical
macro/femto cell networks. IEEE Communications Letters, 13, 755–757.
5. Xu, P., Fang, X., He, R., & Xiang, Z. (2013). An efficient handoff algorithm based on received signal
strength and wireless transmission loss in hierarchical cell networks. Telecommunication Systems, 52,
317–325.
6. Zhang, H., Wen, X., Wang, B., Zheng, W., & Sun, Y. (2010). A novel handover mechanism between
femtocell and macrocell for LTE based networks. In Second international conference on communica-
tion software and networks, 2010. ICCSN’10 (pp. 228–231).
7. Chowdhury, M. Z., Ryu, W., Rhee, E., & Jang, Y. M. (2009). Handover between macrocell and
femtocell for UMTS based networks. In 11th International conference on advanced communication
technology, 2009. ICACT 2009 (pp. 237–241).
Table 2 Number of handoffs and ping-pong rate in different algorithms
Algorithm Average number of ping-
pongs
Average number of
handoffs
Average ping-pong rate
(%)
Proposed 27.4 241 11.37
RWTL [5] 31.3 235 13.32
RSS-alpha [4] 36.5 231 15.80
Conventional
[4]
25 209 11.96
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

8. Chen, X., Wang, H., Xiang, X., & Gao, C. (2014). Joint handover decision and channel allocation for
LTE-A femtocell networks. In 2014 5th International conference on game theory for networks
(GAMENETS) (pp. 1–5).
9. Hoang, N.-D., Nguyen, N.-H., & Sripimanwat, K. (2014). Cell selection schemes for femtocell-to-
femtocell handover deploying mobility prediction and downlink capacity monitoring in cognitive
femtocell networks. In TENCON 2014–2014 IEEE Region 10 conference (pp. 1–5).
10. Roy, A., Shin, J., & Saxena, N. (2012). Multi-objective handover in LTE macro/femto-cell networks.
Journal of Communications and Networks, 14, 578–587.
11. Liu, C., Wei, J., Huang, S., & Cao, Y. (2012). A distance-based handover scheme for femtocell and
macrocell overlaid networks. In 2012 8th International conference on wireless communications, net-
working and mobile computing (WiCOM) (pp. 1–4).
12. Zhang, X., Xiao, Z., Mahato, S. B., Liu, E., Allen, B., & Maple, C. (2014). Dynamic user equipment-
based hysteresis-adjusting algorithm in LTE femtocell networks. IET Communications, 8, 3050–3060.
13. Chen, J., Sheng, C.-Y., & Hsiao, Y.-F. (2013). Handoff decision making in 3-tier heterogeneous cellular
networks. In 2013 International conference on computational problem-solving (ICCP) (pp. 187–190).
14. Guidolin, F., Pappalardo, I., Zanella, A., & Zorzi, M. (2014). Context-aware handover in HetNets. In
2014 European conference on networks and communications (EuCNC) (pp. 1–5).
15. Jae-Wook, L., & Sang-Jo, Y. (2014). Probabilistic handover decision for femtocell network. In 2014
International conference on information and communication technology convergence (ICTC) (pp.
331–334).
16. Abdullah, L. M., Baba, M. D., & Ali, S. G. A. (2014). Parameters optimization for handover between
femtocell and macrocell in LTE-based network. In 2014 IEEE international conference on control
system, computing and engineering (ICCSCE) (pp. 636–640).
17. Antoro, A. D., Mustika, I. W., & Wibowo, S. B. (2015). Downlink cross-tier interference mitigation for
macrocell user in open access femtocell using handover scenario. In 2015 International conference on
electrical engineering and informatics (ICEEI) (pp. 656–660).
18. Jiao, Y., Ma, L., & Xu, Y. (2014). Research on vertical handover in LTE two-tier macrocell/femtocell
systems based on fuzzy neural network. In 2014 International conference on computing, communication
and networking technologies (ICCCNT) (pp. 1–5).
19. Saaty, T. L. (1990). How to make a decision: the analytic hierarchy process. European Journal of
Operational Research, 48, 9–26.
20. Kalbkhani, H., Solouk, V., & Shayesteh, M. G. (2015). Resource allocation in integrated femto–
macrocell networks based on location awareness. IET Communications, 9, 917–932.
21. Elayoubi, S.-E., Haddada, O. B., & Fourestie, B. (2008). Performance evaluation of frequency planning
schemes in OFDMA-based networks. IEEE Transactions on Wireless Communications, 7, 1623–1633.
22. Itoh, K. I., Watanabe, S., Shih, J. S., & Sato, T. (2002). Performance of handoff algorithm based on
distance and RSSI measurements. IEEE Transactions on Vehicular Technology, 51, 1460–1468.
23. Lee, J. Y., Bae, S. J., Kwon, Y. M., & Chung, M. Y. (2011). Interference analysis for femtocell
deployment in OFDMA systems based on fractional frequency reuse. IEEE Communications Letters,
15, 425–427.
24. Claussen, H. (2007). Performance of macro-and co-channel femtocells in a hierarchical cell structure. In
IEEE 18th international symposium on personal, indoor and mobile radio communications, 2007.
PIMRC 2007 (pp. 1–5).
25. Mehta, N. B., Wu, J., Molisch, A. F., & Zhang, J. (2007). Approximating a sum of random variables
with a lognormal. IEEE Transactions on Wireless Communications, 6, 2690–2699.
26. Alouini, M.-S., & Goldsmith, A. J. (1999). Area spectral efficiency of cellular mobile radio systems.
IEEE Transactions on Vehicular Technology, 48, 1047–1066.
27. Oborkhale, L., Salatian, A., & Onoh, G. (2012). Ranking the quality of service of a wireless cellular
network for different periods of a day. International Journal of Energy, Information and Communi-
cations, 3, 55–66.
28. Leu, A. E., & Mark, B. L. (2004). A discrete-time approach to analyze hard handoff performance in
cellular networks. IEEE Transactions on Wireless Communications, 3, 1721–1733.
29. Tung, Y. A. (1998). Time complexity and consistency issues in using the AHP for making group
decisions. Journal of Multi-Criteria Decision Analysis, 7, 144–154.
30. Moon, J. M., & Cho, D. H. (2010). Novel handoff decision algorithm in hierarchical macro/femto-cell
networks. In Wireless communications and networking conference (WCNC) (pp. 1–6).
31. Camp, T., Boleng, J., & Davies, V. (2002). A survey of mobility models for ad hoc network research.
Wireless Communications and Mobile Computing, 2, 483–502.
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

Hashem Kalbkhani received the B.Sc. and M.Sc. degrees from Urmia
University, Iran, both in Electrical Engineering. He is now working
toward the Ph.D. degree with the Department of Electrical Engineer-
ing, Urmia University, Urmia, Iran. His research interests are cellular
networks and signal processing.
Sahar Jafarpour-Alamdari received the B.S. degree from University
of Tabriz, Iran, and M.S. degree from Urmia University, Iran, both in
Electrical Engineering. Her research interests include wireless com-
munication, handoff and resource management in cellular networks.
Mahrokh G. Shayesteh received the B.Sc. degree from the University
of Tehran, Tehran, Iran; the M.Sc. degree from Khajeh Nassir
University of Technology, Tehran, Iran; and the Ph.D. degree from
Amir Kabir University of Technology, Tehran, Iran, all in Electrical
Engineering. She is currently a Professor with the Department of
Electrical Engineering, Urmia University, Urmia, Iran. She is also
working with the Wireless Research Laboratory, Advanced Commu-
nication Research Institute (ACRI), Department of Electrical Engi-
neering, Sharif University of Technology, Tehran, Iran. Her research
interests include wireless communications, signal and image
processing.
QoS-Based Multi-criteria Handoff Algorithm for Femto-Macro…
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order

Vahid Solouk is an assistant professor at the department of IT and
Computer Engineering, Urmia University of Technology and currently
serves as Director of Graduate Campus. He received his Ph.D. degree
in Communication and Network Engineering from the department of
Communication Engineering, Universiti Putra Malaysia. He is a
member of IEEE. His research interests snap specific areas of wireless
and mobile communications including mobility management, resource
allocation, and channel coding.
H. Kalbkhani et al.
123
Downloaded from https://iranpaper.ir https://www.tarjomano.com/order