Content uploaded by Meenakshi Munjal
Author content
All content in this area was uploaded by Meenakshi Munjal on Jul 01, 2018
Content may be subject to copyright.
Vol.:(0123456789)
Journal of Network and Systems Management
https://doi.org/10.1007/s10922-018-9467-y
1 3
QoS andCost‑Aware Protocol Selection forNext Generation
Wireless Network
MeenakshiMunjal1 · NirajPratapSingh1
Received: 10 October 2017 / Revised: 15 June 2018 / Accepted: 24 June 2018
© Springer Science+Business Media, LLC, part of Springer Nature 2018
Abstract
In the growing area of Internet of Things (IoT), mobility management protocols
become necessary because in today’s environment everything is movable so there
is no dominance of static nodes. Mobility is a key aspect for the next-generation
wireless network. So, mobility management mechanism is necessary for providing
the seamless handoff to end user regardless of their location. Thus, Internet Engi-
neering Task Force (IETF) has developed the Mobile IPv6 (MIPv6) for providing
the mobility facility. But MIPv6 is not able to provide the mobility for next-genera-
tion wireless network due to large packet loss and high latency. Thus, Proxy MIPv6
(PMIPv6) is developed for providing the mobility for the next-generation network,
but it has some limitation as well. In this paper, firstly a survey of various types of
protocols based on PMIPv6 protocol is given. Then protocol is selected based on the
Quality of Service (QoS) and signaling cost parameters. Two applications are con-
sidered i.e. streaming and background traffic class. For streaming traffic class, QoS
(handoff latency) parameter is preferred and for background traffic class, signaling
cost is preferred. Analytical Hierarchical Process (AHP) method is used for assign-
ing the weight of different traffic class because user`s preferences play an important
role in the decision-making process as it enhances the quality of experience of the
user. Then multi-attribute decision making (MADM) and prospect theory are used
for selection of protocols. Results show that different protocols are selected for dif-
ferent applications. The performance of MADM and prospect theory is shown in
terms of accuracy.
Keywords MADM algorithm· Mobility Management protocols· Prospect theory·
Proxy Mobile Internet Protocol version 6 (PMIPv6)· Seamless handoff
* Meenakshi Munjal
meenakshi6150008@gmail.com
Niraj Pratap Singh
nirajatnitkkr@gmail.com
1 Department ofElectronics andCommunication Engineering, National Institute ofTechnology,
Kurukshetra, Haryana136119, India
Journal of Network and Systems Management
1 3
1 Introduction
For 5G wireless networks, there is a need of larger capacity and high mobility for
fast-moving vehicles. The cell densification in the 5G network is required for high
capacity support for low mobility users. For next-generation networks, mobility
for vehicles communication becomes necessary [1]. The mobility management is
required for seamless access to mobile services. The basic need of mobility man-
agement is for supporting mobility for all types of real and non-real applications
across homogeneous and heterogeneous wireless network without interruption of
the ongoing session. The session should be continued when the user moves from
one access point to another access point. Mainly, mobility management consists
of two steps: location management and handoff management [2]. The location
management takes care of Mobile Node (MN) reachability in every location and
it should be always connected to the network. The handoff management takes
care of active session which should be maintained at the time of MN roaming [3].
For providing the full access to information and to move from one network
to another, mobility management protocols are required [4]. For achieving con-
sistent and efficient handoff process of different mobile nodes, initially, Mobile
Internet Protocol version 4 (MIPv4) was proposed by Internet Engineering Task
Force (IETF). Then MIPv6 is designed to overcome the problems of MIPv4 such
as weak security mechanism and small size of IP address. MIPv6 support 128
bit long integers and it support mobility for next-generation networks [5]. MIPv6
protocol which is a host-based mobility management protocol, provide continu-
ous service to Mobile Nodes (MNs). For high mobility application, MIPv6 does
not provide best services because, at the time of handoff, it creates disruption.
For removing the drawback of MIPv6, the IETF Network-based Localized Mobil-
ity Management (NETLMM) developed another type of protocol which is called
Proxy Mobile IPv6 (PMIPv6). It is network-based mobility management proto-
col which has various advantages as compared to MIPv6 such as lower hand-
off latency and does not perform Duplicate Address Detection (DAD). PMIPv6
protocol suffers from packet ordering and packet loss problem. So to overcome
the problem of PMIPv6, several improved approaches are developed for seamless
handoff [6].
In literature [7–9], different protocols are studied but it is not defined any-
where that which protocol is suitable for the specific application. As there exists
a large number of protocols, so it becomes a difficult task to select the best pro-
tocol according to user’s requirement. In this paper, a brief description of sev-
eral mobility management protocols is given for handoff procedure improvement.
Every protocol has its own procedure for performing the handoff process [10] and
has different advantages and disadvantage in terms of different QoS and signal-
ing cost parameter [6]. The different protocols are analyzed based on the hand-
off latency and signaling cost. Then selection of protocol is done for different
application based on user’s requirement that enhances the QoE. As the 5G wire-
less network is based on increasing user`s QoE, so user preferences should be
taken into account for selection. The user can give preference by AHP method
1 3
Journal of Network and Systems Management
according to their requirements. Therefore, AHP based selection gives optimal
and efficient results on the basis of application requirements. Then MADM and
prospect theory are used for selection of best protocol according to the specific
applications. As there are different methods for selection, but here two methods
are used for validation of results. The selection results are equally valid for high
mobility and next-generation wireless network. Finally, accuracy has been used
as the performance metric.
The weight for different protocols parameters are computed by AHP [11]. The
weight can be calculated by subjective and objective methods. In the subjective
weight calculation method, user preference is considered while in the objective
method, user preference is not considered. AHP method is termed as subjective
method because the user can give preference according to their requirement.
The protocol selection problem in heterogeneous environment is simulated by
MADM algorithm and prospect theory. Different MADM algorithms include Sim-
ple Additive Weighting (SAW), Multiplicative Exponential Weighting (MEW), Grey
Relational Analysis (GRA), Technique for Order Preference by Similarity to Ideal
Solution (TOPSIS) and VIKOR (VlseKriterijumska Optimizacijia I Kompromisno
Resenje, in Serbian) are used for score calculation [12, 13]. In MADM algorithm,
the selection is done on the basis of user’s preferences and the user always tries to
maximize their utility so their decision may lead to inefficient performance [14].
However, MADM methods are known to suffer from ranking abnormalities, which
may result in the ranking variation of alternatives when an alternative/protocol is
added or removed [15].
The prospect theory was proposed to overcome the drawback of expected utility
theory [16]. It is a behavioral economic theory which is used for decision making
under risk. The prospect theory is well applied in the field of economics but in the
wireless communication, it is not well developed. MADM algorithm is also used for
decision making but in these methods, risk factor is not considered [17]. There may
be risk that a wrong protocol is selected by the user but this factor cannot be consid-
ered in the MADM algorithm. This risk factor is considered in the prospect theory.
The remainder of this paper is organized as follow: Sect.2 introduces the differ-
ent mobility management protocols. In the Sect.3, AHP method for weight calcula-
tion is discussed. Sections 4 and 5 describes the score calculation using MADM
algorithm and prospect theory respectively. Section6 describes the performance
metrics. In Sect.7, simulation results using weight calculation and score calculation
for different applications are discussed. Finally, in Sect.8, the conclusion is given.
2 Mobility Management Protocols
In early 2000, various mobility management protocols were proposed. Some of
the protocols are host-based, which require an active participation from the MN
while some protocols are network based in which another entity is the responsible
for mobility instead of MN [10]. PMIPv6 protocol is network-based mobility man-
agement protocol developed by IETF [18]. In any IP based mobility signaling, it
Journal of Network and Systems Management
1 3
provides mobility without the need of MN. Movement of MN is tracked by mobility
entities which are present in the network. The mobility signaling is initiated by it
and then it set the required routing state.
The main blocks involved in mobility management is Mobile Access Gateway
(MAG) and Local Mobility Anchor (LMA). MAG act as an access router which
keeps track the movement of MN and manages the signaling of MN mobility. The
MN routes are maintained by LMA. LMA is an agent gateway which is in charge
of mobile node’s prefix and manages the mobile node’s binding state. As shown
in Fig.1, if the mobile node moves from home network to another network, then
its movement is detected by MAG1. MAG1 is the previously attached MAG and
MAG2 is the new one when MN moves to a closer position. MAG1 send the Proxy
Binding Update (PBU) to LMA for deregistering the state of MN. After receiving
the PBU, LMA deletes the Binding Cache Entry (BCE) and send the Proxy Bind-
ing Acknowledgement (PBA) to MAG1. At the same time, the MN is detected by
MAG2 in its network range and it sends PBU to the LMA for new state registra-
tion of MN. A bidirectional tunnel is established between LMA and MAG2. Then,
the packets are transferred through the new PMIPv6 tunnel to destination MN. In
PMIPv6, there is no involvement of MN and if there is any tunneling overhead, then
it is removed from the air. During the handoff process, if the MN is not able to trans-
mit and receive packets, then there exists some delay. So, PMIPv6 suffer from hand-
off latency and packet loss. To overcome the drawback of PMIPv6 protocol, differ-
ent protocols are developed. A brief introduction of these protocols is given here.
2.1 Fast Hando inPMIPv6 (FH‑PMIPv6)
To overcome the drawback of PMIPv6 protocol such as large handoff latency
and packet loss, Fast handoff in PMIPv6 is applied. The handoff performance is
improved in FH-PMIPv6, mainly in that highway scenario in which MN’s movement
MAG2
MN
MAG1
LMA
New PMIPv6
Tunnel
1 U
PBU
CN
PBA
PMIPv6 Domain
Fig. 1 PMIPv6 mobility management protocol
1 3
Journal of Network and Systems Management
is in quasi-one Dimensional (1D) and Base Station (BS) is in 1D space. In that case,
chances of anticipated handoff are increased [6]. However, FH-PMIPv6 also has
some drawback such as false handoff initiation which causes unnecessary handoff
and extra signaling between MAG1 and MAG2.
2.2 Hybrid Mode Fast Hando forPMIPv6 (HF‑PMIPv6)
To overcome the drawback of PMIPv6 and FH-PMIPv6 scheme, a hybrid mode fast
handoff for PMIPv6 (HF-PMIPv6) is introduced [6]. HF-PMIPv6 has the advantage
of smaller handoff latency and packet loss as compared to PMIPv6 and FH-PMIPv6
while there is no additional cost as compared to FH-PMIPv6. In HF-PMIPv6, a
decision table is used for detection of mobility mode. The mobility mode can be
reactive and predictive. In the reactive mobility mode, the handoff is directly per-
formed without a tunnel setup. So this mode has the advantage of reducing tunnel
transmission cost, signaling cost, and handoff latency. In the predictive mobility
mode, a tunnel is established such as in PMIPv6 between MAG1 and MAG2 for
reducing packet loss. In HF-PMIPv6, buffer mechanism is used for avoiding the
packet losses and it performs authentication and registration process simultaneously
through which handoff latency is reduced [19].
2.3 Fast Localized Proxy Mobile IPv6 (FL‑PMIPv6)
The FL-PMIPv6 protocol is the host-based mobility management protocol in which
handoff process is initiated by the MN. The basic idea of FL-PMIPv6 is taken from
Fast-Handovers for Mobile-IPv6 (FMIPv6) and it uses message format of 802.21.
By using the MIH message signal, MN gives information to the next MAG before
starting the handoff process. Handoff related signal is sent from MAG1 to MAG2
[8]. If the handoff process is timely and correctly predicted, then packet loss can be
reduced. The handoff prediction time is approximately few tens of millisecond. This
scheme has the advantage of less packet loss and reduced handoff latency.
2.4 Packet Lossless PMIPv6 (PL‑PMIPv6)
PL-PMIPv6 is an analytical model which is designed for lossless data traffic with
authentication consideration. In this model, MAG1 do registration of MAG2 with
LMA in advance which has the advantages of reduced handoff latency. After that
for reducing packet loss, buffering mechanism is used by MAG2. Sometimes vari-
ous data packets are lost because before receiving the PBA, no data is accepted by
MAG2 [20]. The complete handoff success is depending upon the correct judgment
of MAG1 in which mobile nodes are traveling. PL-PMIPv6 has the advantage of
less cost.
Journal of Network and Systems Management
1 3
2.5 Mobile Node’s Home Network Prex (MN‑HNP) Based PMIPv6
In PMIPv6, a prefix is allocated for link between MN and MAG which is known
as MN-HNP. There can be more than one prefix which can be allocated for link
between MN and MAG but in these cases, all prefixes are handled as a group related
with the mobility session. The prefix length can be 128-bit in some configuration. In
this scheme, MN related information MN-HNP (MN-identifier) is stored in Authen-
tication, authorization, and accounting (AAA) server by which MAG2 can access to
that information after the re-association message from the MN via the access point.
If MN connects to PMIPv6 using different interfaces at the same time, then every
interface is allocated a different set of HNP. The entire prefix is allocated to unique
interface of MN which is handled with one mobility session. The MN constructs its
interface with their address by using HNP [9].
2.6 Proxy Mobile IPv6‑Multicast Hando Agent (PMIP‑MHA)
For seamless handoff, fast attachment to new MAG becomes necessary. So for fast
neighbor MAG attachment and reducing the unnecessary data transmission dur-
ing the handoff procedure, multicast handoff agent is proposed which is known as
PMIP-MHA. In PMIP-MHA, list of an active mobile node in all groups is main-
tained by every MHA cache which helps in fast joining during the handoff proce-
dure. For handoff, two triggers are taken i.e. Link Detected (LD) and Link Up (LU).
When LD signal is detected then mobile nodes send proxy Multicast Link Detection
(MLD) signal to MAG1 for preregistration before the handoff. After receiving MLD
signal, MHA of MAG1 update its cache and helps in fast joining during the handoff
procedure [21].
2.7 Smart Buering Scheme
In PMIPv6, the smart buffering scheme is proposed for reducing data traffic loss
and providing the seamless handoff. In this scheme, MN participation is not consid-
ered while Received Signal Strength (RSS) is measured on network side for handoff
purpose and a discovery mechanism is used for finding the next MAG. Smart buff-
ering scheme eliminates the redundant packet transmission by using the link layer
retransmission indication and large buffering time [22]. Thus, the smart buffering
scheme has the advantage of small packet loss ratio, small latency and avoid redun-
dant packet transmission but transmission cost of buffered data is high. For identify-
ing the MAG2 and forward procedure between MAG1 and MAG2, an additional
signal is used, so this scheme also causes signal traffic overhead.
2.8 Head MAG
Head MAG provides more reliable handoff as compared to other schemes. MAG
maintains the MN’s mobility information in the head MAG. The advantage of head
MAG is fast handoff and fewer traffic losses. The head MAG is that MAG which
1 3
Journal of Network and Systems Management
is located in the same LMA domain at an optimal position. All the MAG’s address
information is maintained in the Head MAG which updates MN new location in
MAG2. This procedure has the advantage of fast handoff and fewer traffic losses but
has the drawback of additional overhead and infrastructure cost. It also caused extra
signaling cost and transmission cost due to additional message signal and increasing
buffering traffic from MAG1 to MAG2 respectively [23].
A summary of major aspects, advantages and disadvantages of each of the proto-
cols as identified in this section, is illustrated in Table1.
3 QoS andCost Weighted Parameters
The AHP method is used for assigning the weight of QoS and cost aware protocol.
This weight calculation method is generally used and it is developed by Saaty [11].
Therefore, for selection of a protocol, it is used to provide preference weight to each
parameter on the basis of user`s preference It consists of different steps for weight
calculation:
Step 1: Construct an ordered structure: First of all, the objective of the problem
i.e. QoS and cost parameters are defined in terms of different parameters. Then the
protocol selection problem is broken down into an ordered structure where at the
middle level, parameters are placed and at a lower level, protocols are placed.
Step 2: Construct comparison matrix: The pairwise comparison matrix
is given by
L
=
[
l
mn]i×i
where
i
is the total number of protocols and
lmm =1, lmn =1∕xnm,lmn ≠0
. The element
lmn
is defined by Saaty’s scale from 1 to
9 values as shown in Table2. The Saaty’s scale number is given corresponding to
their importance of one parameter to another.
Step 3: Construct normalized matrix: For scaling the attribute in the same scale,
there is a need for normalization [34]. The matrix is normalized by:
Step 4: Calculate of relative weights: After normalization, the weights of different
parameters are calculated by:
where
i
denotes the number of parameters.
(1)
k
mn =
l
mn
∑
i
m=1
l
mn
(2)
w
m=
∑i
m=1kmn
i
where
i
�
m=1
wm=
1
Journal of Network and Systems Management
1 3
Table 1 Summary of different protocols
Protocol Major Aspects Advantages Disadvantages
FH-PMIPv6 [24] Improve handoff performance in highway
scenario
Reduce the handover latency and the packet
loss ratio as compared to PMIPv6
False handoff initiation which causes unneces-
sary handoff
Extra signaling between MAG1 and MAG2
HF-PMIPv6 [25] A decision able is used for detection of mobil-
ity mode (reactive and predictive)
Buffer mechanism is used for avoiding the
packet losses
Reduce tunnel transmission cost and signaling
cost
Reduced spent time on DAD procedure as
compared to FH-PMIPv6
In predictive mode, a tunnel will be established
which increase handoff delay
FL-PMIPv6 [26] Host-based mobility and initiated by MN
The network access devices need to perform L2
intelligently.
Reduce the handoff latency and packet loss as
compared to FH-PMIPv6 and HF-PMIPv6
Increase burden on MN
PL-PMIPv6 [27] MAG1 do registration of MAG2 with LMA in
advance
Reduce handoff latency
Less cost
For information transfer, wait for PBA from
MAG2
Data sent before PBA is rejected
MN-HNP [28] A prefix is allocated for link between MN and
MAG
MN constructs its interface with their address
by using HNP
Large buffer capacity have made the proposed
scheme cost-effective
Reduce the packet loss
Increase overhead cost
PMIP-MHA [29] List of an active mobile node in all groups is
maintained by every MHA cache which helps
in fast joining during the handoff procedure
Two triggers are taken i.e. Link Detected (LD)
and Link Up (LU).
Reduce the unnecessary data transmission
Support the fast neighbor attachment
Minimize handover delay
Less security
Need an additional cache which increases the
cost
Smart buffering [30] MN participation is not considered
RSS is measured on network side for handoff
purpose and a discovery mechanism is used
for finding the next MAG
Reduce data traffic loss
Small packet loss ratio and small latency
Avoid redundant packet transmission
Transmission cost of buffered data is high
Increase signal traffic overhead
Head MAG [31, 32] All the MAG’s address information is main-
tained in the head MAG which updates MN
new location in new MAG
Fast handoff
Fewer traffic losses
Additional overhead and infrastructure cost
Extra signaling cost and transmission cost
1 3
Journal of Network and Systems Management
4 MADM Algorithm
MADM methods are the most commonly used method for solving the selection
problem. Different MADM algorithms include SAW, MEW, TOPSIS, VIKOR, and
GRA are used for ranking of protocols.
4.1 SAW (Simple Additive Weighting)
The SAW is the simplest method for selection of different alternatives. For selection
purpose, firstly arrange the parameters range in one scale by normalization. Then the
score of different alternatives is calculated by adding the weighted product of all the
parameter [35]. The score obtained by the SAW method is given by
Where
wn
are the weights of different parameters and
amn
is the normalized matrix
of different parameters. In this paper, the normalization is done by
Here
Umn
is the
nth
parameter matrix of
mth
protocol.
4.2 MEW (Multiplicative Exponential Weighting)
MEW is another simplest MADM method in which the normalized parameters matrix
is powered by
wn
. It is also known as the weighted product method. The alternative
which has the highest score is selected as the optimal solution [13]. The scores for
mth
alternative is obtained by
(3)
S
SAW =argmax
m∈I
j
∑
n=1
wna
mn
(4)
a
mn =
U
minimum
U
mn
m=1, 2, ……in =1, 2, ……
j
(5)
S
MEW =argmax
m∈I
j
∑
n=1
awn
mn
Table 2 Saaty’s scale of
pairwise comparison [33]Saaty’s scale Relative importance of two sub-elements
1 Equally important
3 Moderately important with one over another
5 Strongly important
7 Very strongly important
9 Extremely important
2, 4, 6, 8 Intermediate value
Journal of Network and Systems Management
1 3
where
amn
denotes the normalized matrix obtained by using Eq.(4) and
wn
denotes
the weight of
nth
parameters.
4.3 TOPSIS (Technique forOrder Preference bySimilarity toIdeal Solution)
TOPSIS is also an MADM method which is used for measuring relative efficiency of
different protocol. In this method, the preference order is determined on the base of best
similarity to positive choice and worst similarity to negative choice [15].
Step 1: Firstly, the decision matrix is defined in normalized form for one scale trans-
formation by using Eq.(6).
Step 2: Then, the weights for different parameters are multiplied to the normal-
ized matrix.
Step 3: After that the positive and negative ideal solution is determined. The posi-
tive ideal solution maximizes the benefit parameter and negative ideal solution mini-
mize the cost parameter.
where
N
is the set of benefit parameters and N′ is the set of cost parameters.
Step 4: In this step, determine the distance from different alternatives to positive
and negative ideal solution by Eq.(10) and (11).
Step 5: The relative closeness to the ideal solution is calculated by Eq.(12) and
finally arranges the alternatives in the decreasing order for calculation of scores of
different protocols.
(6)
r
mn =
U
mn
�∑
i
m=iU2
mn
(7)
vmn =wm∗rmn
(8)
v+
n= {(maxvmn
m∈i
n∈N,(min
m∈i
vmn
n∈N
�)}
(9)
v−
n= {(minvmn
m∈i
n∈N,(max
m∈i
vmn
n∈N
�)}
(10)
S
+
m=
√∑
n∈j
(vmn −v+
n)
2
(11)
S
−
m=
√∑
n∈j
(vmn −v−
n)
2
(12)
C
j=
S−
m
S−
m
+S
+
m
1 3
Journal of Network and Systems Management
4.4 VIKOR
VIKOR method was developed for a complex system for achieving the multi-attrib-
ute optimization. It is able to determine a compromise ranking list. Even in the pres-
ence of conflicting criteria, it just focuses on ranking and selecting the optimal alter-
native [14]. Firstly, the positive and negative ideal solution is determined by Eq.(8)
and (9). Then, calculate
Sm
and
Rm
for
m=1, 2, ……,i
by using the Eq.(13) and
(14).
After that calculate
Dm
for each
m=1, 2, ……,i
by Eq(15)
where
where the parameter
𝛾
represents the preference given to the strategy and usually it
is taken as 0.5. Finally, arrange the alternatives in increasing order. The highest rank
is selected as the optimal solution.
4.5 GRA (Grey Relational Analysis)
GRA is the part of the grey game theory, which is used for selection of alternatives
having partial information. It solves the selection problem by taking the complete
range of attributes value of every alternative into one, single value [36].
Step 1: For one scale transformation, normalization is done by using Eq.(4).
Step 2: After normalization, the parameters matrix is scaled into [0, 1]. If any of
the elements is equal to, or nearer to one, it means the performance of corresponding
(13)
S
m=
∑
n∈j
wn
(
v
+
n−amn
)
(v+
n
+v−
n
)
(14)
R
m=max
n∈jwn
(
v
+
n−amn
)
(v+
n
+v−
n
)
(15)
D
i=𝛾
(S
m
−S+
S−−S+
)
+(1−𝛾)
(R
m
−R+
R−−R+
)
(16)
S+
=
min
m∈i
S
m
(17)
S−=max
m∈iSm
(18)
R+
=
min
m∈i
R
m
(19)
R−=max
m∈iRm
Journal of Network and Systems Management
1 3
alternatives is the best one for those parameters. But, this kind of alternatives does not
always exist. Due to this, the reference sequence is defined by B0 as
The aim of the reference sequence is to find that alternatives whose compatibility is
closest to the maximum value or reference sequence.
Step 3: Then, calculate the grey relational coefficient to determine how close nor-
malized matrix
bmn
is to
b0n
. The larger the coefficient value, the closer
bmn
and
b0n
are. Grey relational coefficient is calculated by
where
𝜌
is the distinguishing coefficient,
𝜌𝜀[0, 1]
and usually it is taken as 0.5.
Step 4: The score is calculated by multiplying the relational coefficient by the
AHP weight.
5 Prospect Theory
Prospect theory is proposed by Kahneman and Tversky in 1979 [37]. It was used for
decision making mostly in the field of economics [38] but in this paper, it is used for
protocol selection. It is the part of grey game theory in which the risk attitude is con-
sidered. The selection is based on the gain and loss prospective and divided into two
phases: editing and evaluation phase. The reference point is defined in the editing phase
and value function and weight function is defined in the evaluation phase. The value
function is defined by
Where α and β are the parameters which define the concavity and convexity of the
value function and λ is the loss aversion coefficient.
Another name given to the prospect theory is the “Risk MADM” because it is used
for selection from multi attributes factors with a risk factor and there are some steps
that are taken from the MADM method [39].
Step 1: First of all, the normalization is done by using Eq.(4).
Step 2: The positive and negative ideal solution is defined by Eq.(8) and (9) as in
TOPSIS method.
(20)
(b01,b02 ,……b0n)=(1, 1, ……1)
(21)
𝜉
mn =
min
m
(min
m
|
|
b0n−b
∗
mn
|
|)
+𝜌
max
m
(max
n
|
|
b0n−b
∗
mn
|
|)
(|
|
b
0n
−b∗
mn|
|)
+𝜌max
m(
max
n|
|
b
0n
−b∗
mn|
|)
(22)
S
GRA =
j
∑
n=1
wn𝜉mn
.
(23)
e
ij =
{
x𝛼x≥
0
−𝜆
(
−x𝛽
)
x<
0
1 3
Journal of Network and Systems Management
Step 3: After that grey relational coefficient is calculated for both positive and nega-
tive ideal solution.
The grey relational coefficient is used to determine the closeness of compared
sequence to the ideal solution. If the coefficient value is larger, then the two sequence
is closer and vice versa.
Step 4: Then positive and negative prospect value function is defined by
Step 5: After defining the prospect value function, prospect weight is calculated
by
where
𝛾
is the weighting function parameter lies in the range of 0.60 to 0.71 for gain
and 0.51 to .76 for loss. The author in the paper [40–42] tested the different values
of weight function and finally, it is concluded that value of
𝛾+
n
= 0.61 and
𝛾
−
n
= 0.69 is
fitted properly in the selection process.
Step 6: Finally, by using the SAW method the prospect score is calculated by
(24)
𝜉
+
mn =
min
m
(min
n
|
|
amn −v
+
n
|
|)
+𝜌
max
m
(max
n
|
|
amn −v
+
n
|
|)
(|
|
a
mn
−v+
n|
|)
+𝜌max
m(
max
n|
|
a
mn
−v+
n|
|)
(25)
𝜉
−
mn =
min
m
(min
n
|
|
amn −v
−
n
|
|)
+𝜌
max
m
(max
n
|
|
amn −v
−
n
|
|)
(|
|
a
mn
−v−
n|
|)
+𝜌max
m(
max
n|
|
a
mn
−v−
n|
|)
(26)
v
+
mn
=
(
1−𝜉−
mn)0.88
(27)
v+
mn
=−2.25[−
(
𝜉
+
mn
−1
)
]
0.88
(28)
𝜋
+
(
wn
)
=w
𝛾
+
n
n
[
w
𝛾+
n
n+
(
1−wn
)
𝛾+
n
]
1∕𝛾+
n
(29)
𝜋
−
(
wn
)
=w
𝛾−
n
n
[
w
𝛾−
n
n+
(
1−wn
)
𝛾−
n
]
1∕𝛾−
n
(30)
V
m=
j
∑
n=1
v+
mn𝜋+
(
wn
)
+v−
mn𝜋−
(
wn
)
Journal of Network and Systems Management
1 3
Then arrange the
Vm
in the descending order. The first alternative is selected as the
optimal solution.
6 Performance Metric
The performance of ranking scores obtained by different methods is shown in
terms of accuracy. The difference between the maximum and minimum rank-
ing value of two protocols is corresponding to the accuracy of obtained rank-
ing scores. As one protocol has maximum score while another protocol has the
minimum score in the whole scale of ranking, then accuracy is calculated by the
difference between maximum and minimum score value. When the difference
between the two-ranking value is very small, then it is very difficult to identify
which protocol is best. This situation may lead to confusion in the decision-mak-
ing process. So, the difference between two ranking value should be larger so that
it is easy to identify the best protocol. The difference between two ranking values
of protocol allows determining the accuracy of the algorithm.
7 Result andDiscussion
Several PMIPv6 schemes are proposed for reducing the handoff latency, packet
loss, DAD operation. Handoff latency is actually the time gap between IP packets
which is received from MAG1 and MAG2 i.e. at the time of handoff. Packet loss
is the number of lost packets in MN at the time of handoff. The packet loss can
be due to any reasons such as lost connection, butter overflow etc. At the time of
handoff, MN configures its address and DAD once. But in some cases, it config-
ures its DAD operation several times, which should be avoided. If DAD operation
is performed several times, then time spent on DAD operation is increased which
is not desirable. So, it should be configured only once. For reducing the handoff
Table 3 Different protocols ranking [24–32]
Protocol Handoff latency Packet loss rate Spent time on
DAD
Signaling cost
FH-PMIPv6 3 3 4 4
HF-PMIPv6 3 3–4 3 3–4
FL-PMIPv6 2 2–3 2–3 3–4
PL-PMIPv6 2 4 3 3–4
MN-HNP 3 3 3 4
PMIP-MHA 3 2–3 3 5
Smart buffering 2 2 3 4
Head MAG 2 2 3 5
1 3
Journal of Network and Systems Management
latency, packet loss etc. additional overhead is provided during handoff but due to
this, the signaling cost of handoff is increased.
For the evaluation, eight different mobility management protocols are consid-
ered. Every protocol has its own specification in terms of QoS parameters (hand-
off latency, packet loss rate, spent time on DAD procedure) and signaling cost.
The ranking of different protocols in terms of QoS parameters and signaling cost
is given in Table3 where ranking 5 indicate the maximum value and ranking 1
indicates the minimum value. These ranking scores are obtained by reviewing the
different papers. The simulation is performed on MATLAB software.
7.1 Weight Calculation results
AHP method is used for assigning the weights in which user preference is consid-
ered. For streaming traffic class, QoS parameters are taken as an important factor.
The pairwise comparison matrix for streaming traffic class by applying the AHP
method is represented in Table 4. The number 1,3,5,7 is allocated according to
Saaty`s scale using Table2. The parameters are placed according to the descending
order of weights. Streaming traffic class is used for video application which requires
a continuous flow of information so first preference is given to handoff latency and
second preference is given to packet loss rate. Similarly, third preference is given to
spent time on DAD and least preference is given to signaling cost. In the first row of
the matrix, the handoff latency is compared to the all parameters, Firstly, the handoff
latency is compared with itself i.e. with handoff latency, which is equally important
so number 1 assigned from the Table2. Next, the handoff latency is compared to the
packet loss rate and handoff latency is moderately important than packet loss rate so
number 3 is assigned from Table2. Now, the comparison of handoff latency is done
with DAD spent time and signaling cost. The handoff latency is strongly and very
strongly important than DAD spent time and signaling cost respectively, so number
5 and 7 is assigned to the third and fourth column of the first row. In the second row,
the packet loss rate is compared with the all parameters, firstly it is compared to
the handoff latency. In this case, the packet loss rate is less important than latency,
so it is assigned 1/3 as defined in step 2 of AHP. Next, it is compared with itself,
and a number 1 is assigned to the second column in the second row. In the third
and fourth column of the second row, the packet loss rate is compared to the DAD
time and signaling cost. The packet loss rate is moderately important than DAD
time and strongly important from the signaling cost and number 3 and 5 is assigned
Table 4 The pairwise comparison matrix for streaming traffic class
Parameter Handoff latency Packet loss rate Spent time on
DAD
Signaling cost
Handoff latency 1 3 5 7
Packet loss rate 1/3 1 3 5
Spent time on DAD 1/5 1/3 1 3
Signaling cost 1/7 1/5 1/3 1
Journal of Network and Systems Management
1 3
respectively from the Saaty’s scale. Similarly, the third and fourth row is generated
by comparing the DAD time and signaling cost with all parameters. The pairwise
comparison matrix for background traffic class is shown in Table5 in which the first
preference is given to the signaling cost, second preference is given to packet loss
rate and third and fourth preference is given to the handoff latency and DAD time
respectively. The matrix is constructed same as the Table4 but the preferences are
different. After constructing the pairwise matrix, the weight is calculated by apply-
ing the Eqs.(1) and (2) of AHP method. The calculated weight for streaming and
background traffic class is shown in Figs.2 and 3. In Fig.2, the maximum weight is
assigned to the handoff latency because we give most preference to that parameter.
For background traffic class, signaling cost is given most preference as shown in
Fig.3. Background traffic class is used for emails and telemetry application in which
QoS parameter doesn’t matter but the only cost should be minimum. The protocol is
selected based on the AHP weight calculation. The reason for using AHP method is
to increase the QoE of the user because in this case user can give preference accord-
ing to their application and requirement.
Table 5 The pairwise comparison matrix for background traffic class
Parameter Signaling cost Packet loss rate Handoff latency Spent time
on DAD
Signaling cost 1 3 5 7
Packet loss rate 1/3 1 3 5
Handoff latency 1/5 1/3 1 3
Spent time on DAD 1/7 1/5 1/3 1
Fig. 2 Weight of different protocols for streaming traffic class
1 3
Journal of Network and Systems Management
7.2 Score Calculation Results
In this paper, protocol selection problem is simulated by the MADM method and
prospect theory. The performance of these methods is shown in terms of different
application.
Handover LatencyPacket loss rate Spent time on DADSignaling Cost
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Protocol Parameters
Preference Weights
Background Traffic Class
Fig. 3 Weight of different protocols for background traffic class
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Scores (SAW)
Streaming Traffic Class Background Traffic Class
Fig. 4 Protocol scores for different application using SAW method
Journal of Network and Systems Management
1 3
The protocol selected by SAW and MEW method is shown in Figs. 4 and 5
respectively. As shown in weight calculation results, for streaming application,
the maximum weight is given to handoff latency, packet loss rate and spent time
on DAD procedure respectively, so smart buffering PMIPv6 protocol is selected.
As shown in Table3, smart buffering PMIPv6 protocol and head MAG PMIPv6
protocol has minimum value of handoff latency, packet loss rate and spent time
on handoff procedure but the cost of head MAG protocol has the highest signal-
ing cost. As these two protocol has the same range of first three preferences, so in
this case, fourth preference is considered and smart buffering protocol is selected
for streaming application. As shown in Table3, the smart buffering protocol has
least value of handoff latency, packet loss rate, spent time on DAD procedure, so
it is selected every time and it is suitable for the next-generation wireless network.
But it has larger signaling cost, so it is not selected for background traffic class. As
background application gives most preference to cost parameters, so FL-PMIPv6
protocol is selected. Although, various protocols such as HF-PMIPv6, FL-PMIPv6,
PL-PMIPv6 provide minimum cost then in this case after first preference, second
and third preferences are considered. As shown in Fig.3, the second preference is
given to packet loss rate and third preference is given to handoff latency so the pro-
tocol which has a minimum value of those parameters is selected. HF-PMIPv6 pro-
tocol has larger packet loss rate and larger handoff latency, PL-PMIPv6 has larger
packet loss rate and smaller handoff latency and FL-PMIPv6 has smaller packet loss
rate and handoff latency, so FL-PMIPv6 protocol is selected for background appli-
cation. SAW method is easy to implement and it can accommodate multiple crite-
ria for selection with less complexity. MEW method is the least sensitive method
and its complexity is somewhat greater than SAW method but fewer complexes than
another method.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Scores (MEW)
Streaming Traffic Class Background Traffic Class
Fig. 5 Protocol scores for different application using MEW method
1 3
Journal of Network and Systems Management
The scores of different protocols calculated by TOPSIS, VIKOR and GRA
method is shown in Figs.6, 7 and 8 respectively. For streaming application, high-
est weight is given to handoff latency, so the smart buffering protocol is selected for
the same reason. As smart buffering protocol provides best QoS, so it is suitable for
high mobility applications. For background application, due to maximum weight of
cost parameter, the FL-PMIPv6 protocol is selected because it provides minimum
value of signaling cost. The concept of TOPSIS method is simple and compre-
hensive. It gives an accurate result with high efficiency and flexibility. Among all
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Scores (TOPSIS)
Streaming Traffic Class Background Traffic Class
Fig. 6 Protocol scores for different application using TOPSIS method
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Scores (VIKOR)
Streaming Traffic Class Background Traffic Class
Fig. 7 Protocol scores for different application using VIKOR method
Journal of Network and Systems Management
1 3
MADM method, the accuracy of VIKOR method is largest. GRA method can han-
dle many parameters and give a precise solution. But it is very complicated because
as the number of level increases, length of the process also increases.
In MADM algorithm, selection is done on the basis of user’s preferences and
user always tries to maximize their utility so their decision may lead to inefficient
performance. However, MADM methods e.g. TOPSIS and GRA are known to
suffer from ranking abnormalities, which may result in the ranking variation of
alternatives/protocols when an alternative is added or removed. The problem of
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Scores (GRA)
Streaming Traffic Class Background Traffic Class
Fig. 8 Protocol scores for different application using GRA method
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
Scores(Prospect Theory)
Streaming Traffic Class Background Traffic Class
Fig. 9 Protocol scores for different application using prospect theory
1 3
Journal of Network and Systems Management
ranking abnormalities occurs due to removing of low ranking alternative from the
candidate list and then ranking order is changed. An efficient MADM algorithm
requires that the best alternative does not change due to removing or replacing a
low rank alternative. Therefore, if any algorithm is suffering from this problem,
then the ranking order is not stable.
For removing the drawback of MADM application, prospect theory is used. The
protocol scores obtained by prospect theory are shown in Fig.9. In the prospect
theory, the maximum score is given to smart buffering for streaming application
because it has best QoS parameter same as in the MADM method. Smart buffering
PMIPv6 protocol and head MAG PMIPv6 protocol both has best QoS parameter but
signaling cost is larger for head PMIPv6 protocol. So, by considering of all pref-
erence, the smart buffering protocol is selected. For background traffic class, FL-
PMIPv6 is selected due to least signaling cost parameter although HF-PMIPv6, FL-
PMIPv6, PL-PMIPv6 protocols provide minimum cost but HF-PMIPv6 protocol has
larger packet loss rate and larger handoff latency, PL-PMIPv6 has larger packet loss
rate and smaller handoff latency and FL-PMIPv6 has smaller packet loss rate and
handoff latency, so FL-PMIPv6 protocol is selected for background application. The
all user preference (not only highest preference/weight) is considered in the prospect
theory. The advantage of prospect theory is that it is simple. The prospect theory is
also used for risk calculation in case of unknown parameter weight [43]. In that case
objective method is used for determining the weight according to the deviation of
parameters range and then prospect theory is used for score calculation and risk of
wrong protocol selection.
7.3 Accuracy ofDierent Score Calculation Results
The accuracy is calculated for different ranking scores obtained by different meth-
ods in the last section. The accuracy of different ranking scores obtained by differ-
ent methods is shown in Table6. The difference between maximum ranking and
minimum ranking values of alternatives determine the accuracy of the algorithm.
As shown in Table6, prospect theory has maximum accuracy. In the MADM algo-
rithm, VIKOR gives the more accurate result. In comparison of MADM algorithm
and prospect theory, prospect theory gives most accurate result for all applications.
The prospect theory is 66% more accurate as compared to the MADM algorithm for
streaming application and 70% is more accurate for background application. So in
terms of accuracy, prospect theory gives the best selection results.
Table 6 Accuracy of ranking scores
Protocol SAW MEW TOPSIS VIKOR GRA Prospect theory
Streaming 0.2393 0.2490 0.743 1.000 0.4289 1.5695
Background 0.2176 0.2151 0.549 1.000 0.4764 1.6643
Journal of Network and Systems Management
1 3
8 Conclusion
Mobility management protocols are necessary for providing the seamless handoff
to end user regardless of their location. IETF has developed the different protocols
for the next-generation wireless network. In this paper, a brief introduction of differ-
ent protocols based on PMIPv6 protocol is given. Then the best protocol is selected
for specific applications based on QoS and signaling cost parameters. Two applica-
tions have been considered i.e. streaming and background application. For stream-
ing application; QoS parameters are more important so highest weight is given to
QoS parameters (handoff latency) and for background traffic class; the cost is more
important, so, the highest weight is given to signaling cost. AHP method is used for
assigning the weight for different applications that enhance the QoE of the user in
the selection process. Then MADM and prospect theory have been used for selec-
tion of protocols. The simulation results show that smart buffering PMIPv6 proto-
col is selected for streaming application due to finest QoS parameters i.e. minimum
value of handoff latency, packet loss rate, and spent time on DAD. Although smart
buffering PMIPv6 protocol has larger signaling cost, but, cost doesn’t matter here.
So, the smart buffering PMIPv6 protocol is suitable for those applications where
QoS is more important. For background application, the cost is more important
so more weight is given to cost parameter and in this application also MADM and
prospect theory both select FL-PMIPv6 protocol as the best protocol. The results are
verified in terms of accuracy. Results show that prospect theory provides 66–70%
more accurate results as compared to MADM algorithm. Here two selection meth-
ods have been used for validation of result. So, for next-generation mobile commu-
nication, smart buffering PMIPv6, and FL-PMIPv6 protocols are recommended for
applications where QoS and cost respectively are more important respectively.
References
1. Duan, X., Wang, X.: Authentication handover and privacy protection in 5G hetnets using software-
defined networking. IEEE Commun. Mag. 53(4), 28–35 (2015)
2. Astorga, J., Jacob, E., Toledo, N., Aguado, M.: Analytical evaluation of a time- and energy-efficient
security protocol for IP-enabled sensors. Comput. Electr. Eng. 40(2), 539–550 (2014)
3. Ko, H., Lee, G., Pack, S., Member, S., Kweon, K.: Timer-Based Bloom Filter Aggregation for
Reducing Signaling Overhead in Distributed Mobility Management. IEEE Trans. Mob. Comput.
15(2), 516–529 (2016)
4. Fernandez, P., Santa, J., Bernal, F., Gomez-Skarmeta, A.F.: Securing Vehicular IPv6 Communica-
tions. IEEE Trans. Dependable Secur. Comput. 13(1), 46–58 (2016)
5. Makaya, C., Pierre, S.: An Analytical Framework for Performance Evaluation of IPv6-Based Mobil-
ity Management Protocols. IEEE Trans. Wirel. Commun. 7(3), 972–983 (2008)
6. Kim, M.-S., Lee, S., Cypher, D., Golmie, N.: Performance analysis of fast handover for proxy
Mobile IPv6. Inf. Sci. (Ny) 219, 208–224 (2013)
7. Lee, J.-H., Ernst, T., Chilamkurti, N.: Performance Analysis of PMIPv6-Based Network Mobility
for Intelligent Transportation Systems. IEEE Trans. Veh. Technol. 61(1), 74–85 (2012)
8. J.-E. Kang, D.-W. Kum, Y. Li, and Y.-Z. Cho, “Seamless Handover Scheme for Proxy Mobile IPv6,”
IEEE Int. Conf. Wirel. Mob. Comput. Netw. Commun., pp. 410–414, 2008
9. Makaya, C., Pierre, S.: Emerging Wireless Networks:Concepts, Techniques, and Applications. CRC
Press, New York (2012)
1 3
Journal of Network and Systems Management
10. Bouaziz, M., Rachedi, A.: A survey on mobility management protocols in Wireless Sensor Net-
works based on 6LoWPAN technology. Comput. Commun. 74, 3–15 (2016)
11. Saaty, T.L.: The Analytic Hierarchy and Analytic Network Measurement Processes: applications to
Decisions under Risk. Eur. J. Pure Appl. Math. 1(1), 122–196 (2008)
12. Wang, L., Kuo, G.G.S.: Mathematical Modeling for Network Selection in Heterogeneous Wireless
Networks – A Tutorial. IEEE Commun. Surv. Tutorials 15(1), 271–292 (2013)
13. J. D. Martínez-Morales, U. Pineda-Rico, and E. Stevens-Navarro, “Performance comparison
between MADM algorithms for vertical handoff in 4G networks,” in 7th International Conference
on Electrical Engineering, Computing Science and Automatic Control, CCE, 2010, pp. 309–314
14. Qin, J., Liu, X., Pedrycz, W.: An extended VIKOR method based on prospect theory for multi-
ple attribute decision making under interval type-2 fuzzy environment. Knowledge-Based Syst. 86,
116–130 (2015)
15. Sgora, A., Gizelis, C.A., Vergados, D.D.: Network selection in a WiMAX–WiFi environment. Per-
vasive Mob. Comput. 7(5), 584–594 (2011)
16. Wang, L., Zhang, Z., Wang, Y.: A prospect theory-based interval dynamic reference point method
for emergency decision making. Expert Syst. Appl. 42(23), 9379–9388 (2015)
17. Zhou, L., Zhong, S., Ma, S., Jia, N.: Prospect theory based estimation of drivers’ risk attitudes in
route choice behaviors. Accid. Anal. Prev. 73, 1–11 (2014)
18. Cho, C., Kang, J.-J., Jeong, J.: Performance analysis of DNS-assisted global mobility management
scheme in cost-optimized proxy mobile IPv6 Networks. Inf. Syst. 48, 226–235 (2015)
19. T. Y. Wu, W. K. Liu, and W. T. Lee, “Advanced handover enhancement for 6LoWPAN PMIPv6
by HF-PMIPv6,” Proc. - 2014 Int. Conf. Inf. Sci. Electron. Electr. Eng. ISEEE 2014, vol. 3, pp.
1770–1774, 2014
20. Ryu, S., Kim, G.Y., Kim, B., Mun, Y.: A scheme to reduce packet loss during PMIPv6 handover
considering authentication. Proceedings - The International Conference on Computational Sciences
and its Applications, ICCSA 6, 47–51 (2008)
21. M. Gohar, S. Il Choi, and S. J. Koh, “Fast handover using multicast handover agents in PMIPv6-
based wireless networks,” in International Conference on Information Networking, ICOIN, 2011,
pp. 367–372
22. W. Xu, Wu, Daneshmand, Liu, “Smart Buffering for seamless handover in Proxy Mobile IPv6,”
Wirel. Commun. Mob. Comput., vol. 11, no. 4, pp. 491–499, 2011
23. Kwon, N., Kim, H., Oh, S., Choo, H.: Fast handover scheme based on mobility management of head
MAG in PMIPv6. Springer ICCSA 6, 181–192 (2011)
24. M. C. Chuang and J. F. Lee, “FH-PMIPv6: A fast handoff scheme in Proxy Mobile IPv6 networks,”
in International Conference on Consumer Electronics, Communications and Networks, CECNet,
2011, pp. 1297–1300
25. Yan, Z., Zhou, H., Wang, H.C., Zhang, H., Zhang, S.: Design and implementation of a hybrid
MIPv6/PMIPv6-based mobility management architecture. Math. Comput. Model. 53(3–4), 421–442
(2011)
26. I. Al-Surmi, M. Othman, N. A. W. A. Hamid, and B. M. Ali, Latency low handover mechanism con-
sidering data traffic lost preventing for proxy mobile IPv6 over WLAN, vol. 70, no. 1. 2013
27. S. Ryu and Y. Mun, “Scheme to prevent packet loss during PMIPv6 handover,” in Transactions on
Computational Science VI, 2009, pp. 131–142
28. I. Al-Surmi, M. Othman, and B. M. Ali, “Review on mobility management for future-IP-based next
generation wireless networks,” in 12th International Conference on Advanced Communication Tech-
nology (ICACT), 2010, vol. 2, pp. 989–994
29. Koh, S.J., Gohar, M.: Multicast Handover Agents for Fast Handover in Wireless Multicast Net-
works. IEEE Commun. Lett. 14(7), 676–678 (2010)
30. H. Choi, K. Kim, H. Lee, and S. Min, “Reactive Smart Buffering Scheme for Seamless Handover in
PMIPv6,” in International Conference On Principles Of Distributed Systems, 2008, pp. 568–571
31. Modares, H., Moravejosharieh, A., Lloret, J., Bin Salleh, R.: A survey on proxy mobile IPv6 hando-
ver. IEEE Syst. J. 10(1), 208–217 (2016)
32. Jabir, A.J., Subramaniam, S.K., Ahmad, Z.Z., Asilah, N., Hamid, W.A.: A cluster-based proxy
mobile IPv6 for IP-WSNs. EURASIP J. Wirel. Commun. Netw. 173(1), 1–17 (2012)
33. Meenakshi and N. P. Singh, “A comparative study of cooperative and non-cooperative game theory
in network selection,” in IEEE International Conference on Computational Techniques in Informa-
tion and Communication Technologies (ICCTICT), 2016, pp. 612–617
34. M. Munjal and N. P. Singh, “Utility aware network selection in small cell,” Wirel. Networks, 2018
Journal of Network and Systems Management
1 3
35. F. Bari and V. Leung, “Multi-Attribute Network Selection by Iterative TOPSIS for Heterogeneous
Wireless Access,” in 4th IEEE Consumer Communications and Networking Conference, 2007, pp.
808–812
36. Verma, R., Singh, N.P.: GRA Based Network Selection in Heterogeneous Wireless Networks. Wirel.
Pers. Commun. 72(2), 1437–1452 (2013)
37. Kahneman, D., Tversky, A.: Prospect Theory: an Analysis of Decision under Risk. Econometrica
47(2), 263–292 (1979)
38. Rosenfeld, A., Kraus, S.: Predicting Human Decision-Making: from Prediction to Action. Synth.
Lect. Artif. Intell. Mach. Learn. 12(1), 1–150 (2018)
39. M. Munjal and N. P. Singh, “Improved network selection for multimedia applications,” Trans.
Emerg. Telecommun. Technol., vol. 28, no. 5, 2017
40. Tversky, A., Kahneman, D.: Advances in Prospect-Theory - Cumulative Representation of Uncer-
tainty. J. Risk Uncertain. 5(4), 297–323 (1992)
41. Nilsson, H., Rieskamp, J., Wagenmakers, E.J.: Hierarchical Bayesian parameter estimation for
cumulative prospect theory. J. Math. Psychol. 55(1), 84–93 (2011)
42. Rieger, M.O., Wang, M., Hens, T.: Estimating cumulative prospect theory parameters from an inter-
national survey. Theory Decis. 82(4), 567–596 (2017)
43. Zhang, X., Liu, P.: Method for Multiple Attribute Decision-making under Risk with Interval Num-
bers. Int. J. Fuzzy Syst. 12(3), 237–242 (2010)
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.
Meenakshi Munjal is a research scholar in the Department of Electronics and Communication Engineering
at National Institute of Technology, Kurukshetra, Haryana, India. She received her B.Tech. and M.Tech.
degrees in Electronics and Communication Engineering from Kurukshetra University, Kurukshetra,
Haryana, India. Her research interest is wireless communication. She is currently doing research in radio
resource management and mobility management of next generation wireless network.
Niraj Pratap Singh is an associate professor in the Department of Electronics and Communication Engi-
neering at National Institute of Technology, Kurukshetra. He received his B.E. and M.E. degrees in
Electronics and Communication Engineering from Birla Institute of Technology, Mesra, Ranchi, India
in 1991 and 1994, respectively. He received Ph.D. degree in Electronics and Communication Engineer-
ing from National Institute of Technology, Kurukshetra, India. His research interests are radio resource
management, interworking architectures design, D2D communication, wireless sensor network, cognitive
radio; and mobility management of next generation wireless networks.
A preview of this full-text is provided by Springer Nature.
Content available from Journal of Network and Systems Management
This content is subject to copyright. Terms and conditions apply.
