multi-attribute network selection and evaluation models ... · multi-attribute network selection...

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 12, Number 16 (2017) pp. 5495-5510

© Research India Publications. http://www.ripublication.com

5495

Multi-Attribute Network Selection and Evaluation Models for Vertical

Handoff in Heterogeneous Networks

K.Aruna Kumari1 and Mesala Sravani2

1Assistant Professor, Department of Computer Science Engineering, Sagi Rama Krishnam Raju Engineering College, Bhimavaram, India.

2Software Engineer, India.

1Orcid: 0000-0002-0267-2594 & Researcher ID: Q-2757-2016 2Orcid: 0000-0003-2634-3354 & Researcher ID: N-3977-2017

Abstract

The extensive and stupendous development in the field of

mobile and wireless communication focuses on the intension

and desire to provide un interrupted seamless connectivity in

accessing numerous technologies in wireless communication

and to get connected to the best network which aims in

providing the best quality of service (QOS).The movement of

the mobile subscriber along several radio networks with

distinct features demands for a proper handoff to be carried

out. For performing handoff efficiently, selection of

destination network needs accuracy. In this work the focus is

laid on three major mobile networks WLAN(Wireless Area

Networks), WiMAX (Worldwide Interoperability for

Microwave Access), UMTS(Universal Mobile

Telecommunications Systems).Various network criteria like

Bandwidth, Jitter, Delay, Packet loss, and Cost are considered.

We have proposed several handoff decision MADM modules

for improving the performance and to decide the best

destination network to be connected. The simulated results

based on performance evaluation validate the efficiency of the

proposed models in choosing the best network during the

process of handoff.

Keywords: Vertical Handoff (VHO), Multiple Attribute

Decision Making (MADM), WLAN, WiMAX, UMTS,

Quality of Service (QOS).

INTRODUCTION

Next Generation Network (NGN) will inexorably integrate

triple-play services, which means that all traffic classes of

voice, video and data will be managed to meet the particular

Quality of Service (QoS) requirements, such as packet delay,

jitter and loss. In spite of the radio access technologies, no

single wireless network technology is considered to be more

favorable than other technologies in terms of QoS. Moreover,

due to competition between the infrastructure based on WIFI,

UMTS and WiMAX, the operators of telecommunication are

not yet willing to change their infrastructure based on 2G and

3G[1]. For instance WIFI provides better bandwidth with

limited area of coverage, while UMTS covers large area with

limited bandwidth and WiMAX provides moderate coverage

with better bandwidth consumption. There are two criteria

involved. The first one deals with users being benefited by the

concept of “Always Best Connected“(ABC) which allows the

users to various services anywhere at any time in the best way

with devices multi-interfaces[2]. The second criteria are

ensuring interoperability convergence between various

technologies with heterogeneous specifications. To satisfy the

first criteria, the systems have been designed and developed

with new terminals which are equipped with multiple

interfaces. Also, the VHO process maintains the convergence

between heterogeneous networks. The vertical handover

means that the calls can be transferred by the mobile terminal

from one base station (BS) the other. It does not mean a

change in the channel assigned. Thus handoff is the process of

transferring the call of a mobile station from one BS to the

other or one cell boundary to the other. The classification of

different types of Handoff’s is shown in figure1below.

Types of Handoff

HORIZONTAL HANDOFF (Symmetric handoff)

The cellular network is further classified as intra-cell and

inter-cell handoffs as shown in figure 1. In intra-cell handoff

means when a user moving with mobile terminal within a

network or cell and the radio channels changes in order to

minimize inter channels interference under the same base

station of same network technology. Furthermore the inter cell

handoff will occur when Mobile terminal moves into the

adjacent cell of the any base station of any other network.

mailto:[email protected]

mailto:[email protected]

http://orcid.org/0000-0002-0267-2594

http://orcid.org/0000-0003-2634-3354



5496

Figure 1: Handoff Classification.

Figure 2: Horizontal and Vertical handoff

VERTICAL HANDOFF (Asymmetric handoff)

The vertical handoff process involves three main phases,

namely system discovery, vertical handoff decision, and

vertical handoff execution as shown in figure 2. During the

system discovery phase, the mobile terminal determines

which networks can be used. In the vertical handoff decision

phase, the mobile terminal determines whether the

connections should continue using the existing selected

network or be switched to another network as shown in figure

3. During the vertical handoff execution phase, the

connections in the mobile terminal are re-routed from the

existing network to the new network in a seamless manner.

This phase also includes the authentication, authorization, and

transfer of a user’s context information.

Figure 3: Phases of horizontal handoffs



5497

Decision Making Parameters of Vertical Handoff

Figure 4: Parameters of Handoff

Quality of Service: Quality of service (QoS) is the key

parameter that decides the efficiency of a mobile network.

Some of the key parameters of QoS are Band Width, Delay,

Jitter, latency, loss and cost. In this paper the parameters

stated are consider and analysed as shown in figure 4.

Bandwidth: Bandwidth is the difference of the upper and

lower frequencies in a given set of frequency band and is

measured in Hertz. It is also referred as pass band bandwidth,

sometimes to baseband bandwidth, basing on practical

considerations.

Delay: Network delay is a major parameter that will decide

the performance characteristic of a cellular network. It is the

time taken for the packet to travel from the CBR (Constant Bit

Rate) source to the destination which is measured in seconds.

Jitter: It is change in packet delay during transit that is caused

by queuing, contention and serialization effects that are

caused due to the nature of the path of propagation through

the network. Generally jitter is high in the links that are slow

and have highly congested.

Signal Quality: It is the measure of Bit Error Rate (BER)

after the process of decoding, and is proportional to the S/N

ratio. The effect is severe under weak signal quality

conditions as the skew angle adjustment in the LNB (low

noise block) is not optimum.

Reliability: Reliability is a factor which is related to effective

delivery of data to the intended recipient(s), as opposed to an

unreliable protocol, which does not provide notifications to

the sender as to the delivery of transmitted data.

MULTIPLE ATTRIBUTE DECISION

MAKING(MADM) METHODS

Multiple attribute decision making (MADM) methods

consider problems where making preference decisions over

available alternatives that are characterized by multiple and

usually conflicting attributes are required. It is a part of a

general class of operation research models that deal with

decision problems under the presence of a number of decision

criteria.MADM is a branch of the field multiple criteria

decision making (MCDM). MADM problems are different

from one another in disciplines, but all of them have the

common characteristics like: various selection alternatives,

multiple attributes which describe various alternatives in a

variety of measurement, and a set of weights among

attributes[3]. For notation, let M be the set of alternatives and

N be the set of parameters or attributes. A MADM problem is

expressed in expressed in a matrix format, where columns

represent attributes and rows represent several alternatives.

The matrix element xij indicates the rating of the performance

of the ith alternative with respect to the jth attribute. Thus, a

MADM problem with |M| alternatives and each with |N| parameters is given by

[

𝑥11 𝑥12𝑥1𝑗 ⋯ 𝑥1𝑁

⋮ ⋱ 𝑥𝑖𝑗 ⋮𝑥𝑀1 𝑥𝑀2 𝑥𝑀𝑗 ⋯ 𝑥𝑀𝑁

]

A set weight is defined for the calculation of the ranking. The

value of these weights should represent various levels of

importance of the parameter for the decision making. The set

of weights must satisfy the constraint:

∑ 𝑊𝑗

𝑗∈𝑁

= 1

In the case of VHO decision, the weights should represent the

QoS requirements of the connection as well as the user´s

preferences. In general, all the following MADM decision

methods first calculate an index or score based on their

specific procedures, then, the network selected for vertical

handoff is the one with the best score value or the one in the

first place in the ranking of candidate networks.

The list of MADM methods that are proposed for evaluation

of the best network selection are.



5498

Euclidean Distance Based Network Selection

Algorithm(EDBNS)

Rank Reversal Technique based Algorithm (RRTA)

Parameter based Network Selection

Algorithm(PBNSA)

Oliver Blume Algorithm Method(OBAM)

Similarity Based Network Selection

Algorithm(SBNSA)

Models Formulation

a) Euclidean Distance Based Network Selection

Algorithm(EDBNS)

This method is based on calculation of Euclidean distance.

The distance of the decision matrix from ideal matrix and

non-ideal matrix is formulated. This is performed by

constructing ideal matrix that consists of maximum value of

positive attribute and lowest value for the negative attribute.

Similarly the non-ideal matrix consisting of lowest values is

calculated and the ideal matrix is constructed. The distance

between the decision matrix and ideal matrix is found.

Similarly the distance between decision matrix and non-ideal

matrix is also calculated. Finally the positive solution and

negative solution are found out. The network with highest cost

function ranked as the best network. The efficiency of the

network is also calculated.

1. Construct decision matrix.

D( i , j) = [

𝑥11 𝑥12𝑥1𝑗 ⋯ 𝑥1𝑁


]

2. Construct the positive ideal matrix with maximum and

minimum values from attributes. Similarly negative ideal

matrix.

𝐼+ = [𝑖11 𝑖12 … … … 𝑖1𝑛]𝐼− = [𝑖11 𝑖12 … … … 𝑖1𝑛]

3. Calculate the distance between decision matrix and positive

ideal matrix

𝑐𝑖+ = √∑(

𝑛

𝑗=1

𝐷(𝑖, 𝑗) − 𝐼+)2

4. Calculate the distance between decision matrix and

negative ideal matrix

𝑐𝑖− = √∑(𝐷(𝑖, 𝑗) − 𝐼−)2

𝑛

𝑗=1

5. Normalise the values of 𝐶𝑖 to obtain the value of C as

𝑐 =𝑐𝑖

+ −⁄

𝑚𝑒𝑎𝑛

6. Calculate the positive solution 𝑠𝑖+ = [

𝑐11

⋮𝑐𝑀1

]

7. Calculate negative solution 𝑠𝑖− = [

𝑐11

⋮𝑐𝑀1

]

8. Calculate cost function 𝑐𝑘=𝑆−

𝑆−+𝑆+

9. Rank the value with maximum 𝑐𝑘value.

b) Rank Reversal Technique Based Algorithm (RRTA):

RRTA is an acronym for Rank Reversal Technique based

algorithm. It is based on the concept of preference order by

similarity to Ideal solution[4][5]. This algorithm is used to

give ranking to the network by calculating the cost

function[6].

1. Construct the decision matrix, D

D = [

𝑥11 𝑥12𝑥1𝑗 ⋯ 𝑥1𝑁


]

Where 𝑥𝑖𝑗=elements of decision matrix D.

2. Obtain Cost Function, minimum the better

3. Construct Normalized Decision Matrix, R

𝑅𝑠𝑖 = 𝑠𝑖 √∑ 𝑠𝑖2𝑚

𝑖⁄

4. Calculate Weight Normalized Decision Matrix,

V= [W] *[R]

5. Determine +ve and –ve Ideal Solution

𝐴+= MAX (V) ,𝐴−= MIN (V) .

6. Calculate Separation Measure

𝑆𝑖+ = √∑

(𝐴𝑗+ − 𝑣𝑖𝑗)

𝑤𝑖𝑗

2𝑛

𝑗=1

𝑆𝑖− = √∑

(𝐴𝑗− − 𝑣𝑖𝑗)

𝑤𝑖𝑗

2𝑛

𝑗=1

7. Calculate Cost C=𝑆−

(𝑆−+𝑆+)

8. Rank the network on the basis of cost function.

c) PBNSA (Parameter Based Network Selection

Algorithm):

The proposed PBNSA algorithm is also a MADM method

which is based on the preference structure of PROMTHEE

and the criterion of Euclidian distance. Initially it compares

each pair of solutions (ai,ar) using a preference function(d),

where d= fj( ai)- fj( ar) is the relative difference between the

evaluation of two alternatives[7]. This algorithm involves in

dealing with both qualitative and quantitative analysis. The

pair wise comparison refers to qualitative analysis and the



5499

Euclidian distance calculation which specifies the degree of

preference through the quantity [8].

1. Construct the decision matrix 𝐷 and the weight

vector W.

D = [

𝑥11 𝑥12𝑥1𝑗 ⋯ 𝑥1𝑁


]

Wj= [0.0860 0.1672 0.4319 0.0531 0.2593 0.0021]

2. Define the preference function for each attribute.

3. Define the preference index for each couple of

alternatives:

n = V (𝑎𝑖 , 𝑎𝑟) = Wj (𝑝𝑗(𝑓𝑗(𝑎𝑖) − 𝑓(𝑎𝑟))).

4. Calculate the distance 𝑆𝑖+between each scheme and

positive ideal point and also calculate the distance

𝑆𝑖−between each scheme and negative ideal point.

𝑆𝑖+ = √∑ (𝑣𝑖𝑗 − 𝑣𝑗

+)2𝑛𝑗=1 , i∈ 𝑚,

𝑆𝑖− = √∑ (𝑣𝑖𝑗 − 𝑣𝑗

−)2𝑛𝑗=1 , i ∈ 𝑚,

5. Calculate the relative approach degree 𝐶𝑖+ of each

scheme to the ideal points.

𝐶𝑖+ =

𝑆𝑖−

(𝑆𝑖++𝑆𝑖

−), 0 <𝐶𝑖

+ < 1, i∈ 𝑚,

6. Rank the schemes based on 𝐶𝑖+.The larger is the

𝐶𝑖+the better is the scheme.

d) Oliver Blume Algorithm method(OBA)

The OBA method is an acronym for Oliver Bloom Algorithm

which suggests the network selection mechanism which is

based on the QOS parameters. The network performance for

various QoS criteria like bandwidth, delay, loss, jitter and cost

are observed[9]. The OBA is summarised as follows.

1. Construct the decision matrix.

D = [

𝑥11 𝑥12𝑥1𝑗 ⋯ 𝑥1𝑁


]

2. Construct weight matrix.

Wj= [0.0860 0.1672 0.4319 0.0531 0.2593 0.0021]

3. Construct the Ideal matrix I consisting of min/max value

for each attribute.

4. Calculate the cost function 𝐶𝑖 = ∏ (𝐷𝑖𝑗

𝐼𝑖𝑗)𝑤𝑖𝑗

𝑖

5. Rank the network with minimum cost.

e) Similarity Based Network Selection Algorithm

(SBNSA)

The similarity based algorithm is based on the determination

of ideal solution and non-ideal solution. The most important

attribute is to have similar to positive ideal solution and the

worst attribute is having the similar to non-ideal solution. This

is one of the multi criteria analysis method in which based on

different parameters to choose the best network [10] .This

method gives disagreement index is found between the ideal

solution and non-ideal solution. Fromfigure5 𝐵+and𝐵− are the

ideal and no ideal solutions. These two alternatives

𝐵+𝑎𝑛𝑑𝐵−are found to be two vectors in x-dimensional space.

This space between 𝐵+𝑎𝑛𝑑𝐵− gives the separation between

two alternatives. The angle between the 𝐵+𝑎𝑛𝑑𝐵− given

by 𝜃 .If 𝜃=0 which means 𝐵+𝑎𝑛𝑑𝐵− are increasing in same

direction with coincident. There is a problem when 𝜃 ≠ 0

then both the alternatives does lie in same line. The two

alternatives angle 𝜃 gives gradients of vectors.When the

angle 𝜃=0 both vectors lies in same line when 𝜃=90° then the

vectors will be 90°apart from each other[11]. The ideal

solution is which gives best alternative which is having degree

of resemblance between positive ideal solution and lowest

degree of resemblance with non-ideal solution.

Figure 5: Ideal and Non Ideal solutions

The positive and negative solution mainly has best

alternatives and worst alternatives it mainly has highest and

lowest values. This is mainly based on multiple attribute

decision making which helps in finding the performance

matrix. Similarly we can find resemblance degree 𝜃 can be

calculated from ideal and on ideal solution and finally

performance index can be calculated in which highest value is



5500

give highest priority similarly ranking is given.

1. Determine the Decision matrix

A=

11 1

1

n

m mn

a a

a a

2. Find the normalized decision matrix using the formula

𝑥𝑖𝑗 =𝑎𝑖𝑗

√∑ 𝑎𝑖𝑗2𝑛

𝑗=1

3. The weight matrix is given by.𝑃𝑗= [𝑃11 𝑃12⋯ 𝑃1𝑁]

4. Required to calculate the normalized weight matrix.

𝑉𝑖𝑗= [𝑃11 𝑃12𝑃1𝑗 ⋯ 𝑃1𝑁] [

𝑥11 𝑥12 ⋯ 𝑥1𝑁


]

5. Find the ideal solution and negative ideal solutions

𝐵𝑗+= MAX (𝑉𝑖𝑗) 𝐵𝑗

−= MIN (𝑉𝑖𝑗)

𝐵𝐽+= [𝑚11

+ 𝑚12+ ⋯ 𝑚16

+ ] 𝐵𝐽−= [𝑚11

− 𝑚12− ⋯ 𝑚16

− ]

6. Find the degree 𝜃+or𝜃− between each alternative for

positive and negative ideal solution

𝜃𝑗+ =

∑ 𝑉𝑖𝑗𝐵𝐽+𝑚

𝑗=1

(∑ 𝑉𝑖𝑗2𝑚

𝑗=1 )0.5

(∑ 𝐵𝐽+2𝑚

𝑗=1 )0.5 𝜃𝑗

− =∑ 𝑉𝑖𝑗𝐵𝐽

−𝑚𝑗=1

(∑ 𝑉𝑖𝑗2𝑚

𝑗=1 )0.5

(∑ 𝐵𝐽−2𝑚

𝑗=1 )0.5

7. Find positive ideal solution and negative ideal solution

degree of resemblance

𝑘𝑖 = 𝜃𝑖+ ∗ 𝑣𝑖𝑗 , 𝑙𝑖 = 𝜃𝑖

− ∗ 𝑣𝑖𝑗

8. Calculate the overall performance index for each

alternative.

𝑜𝑖+ =

𝑘𝑖

𝐵𝐽+ 𝑜𝑖

− =𝑙𝑖

𝐵𝐽−

9. Rank the networks which is having highest value of

𝑄𝑖as 𝑄𝑖 =𝑜𝑖

+

𝑜𝑖++𝑜𝑖

−

CASE STUDY

The above section a detailed description of various vertical

hand off decision schemes and MADM methods like

Euclidean Distance Based Network Selection Algorithm

(EDBNS), Rank Reversal Technique based Algorithm

(RRTA), Parameter based Network Selection Algorithm

(PBNSA), Oliver Blume Algorithm Method (OBAM), and

Similarity Based Network selection Algorithm (SBNSA) are

proposed .As an example the case of a mobile terminal

currently connected to a Wi-Fi cell is considered. It has to

make decision among six candidate networks Wi-Fi 1,Wi-Fi

2, UMTS-1, UMTS-2, W-LAN 1, W-LAN 2 . Vertical

handover QOS criteria that are considered here are Packet

delay, bandwidth, cost, Packet jitter , Packet loss. The above

said algorithms are applied to the data below for six different

networks for different QOS parameters in Table 1

The weights for the above data are calculated and taken

according to IEEE standard calculations [12] as

Wj= [0.0860 0.1672 0.4319 0.0531 0.25930.0021]

Calculation of Entropy weights are as follows:

a. Let 𝑞𝑖𝑗 denote the contributing degree of scheme I to

target attribute j

𝑞𝑖𝑗 =𝑢𝑖𝑗

∑ 𝑢𝑖𝑗𝑚𝑖=1

⁄ .

b. Entropy 𝑋𝑗 can be used to denote the total of

contribution of all schemes to target attribute

𝑋𝑗 = −C ∑ qij ln qij,m

i=1 j = 1,2, … . . , n

Where C=1/ln m to guarantee 0≤ 𝑋𝑗 ≤ 1

c. Define ℎ𝑗as the inconsistent degree of each scheme’s

contributing degree corresponding to target attribute

jas

ℎ𝑗 = 1 − 𝑋𝑗

d. The entropy weight of every target attribute can be

obtained by:

𝑔𝑗 =ℎ𝑗

∑ ℎ𝑗𝑛𝑗=1

⁄ .

e. According to AHP weight 𝑤𝑗 and the entropy weight

𝑡𝑗, the modified comprehensive weight 𝑊𝑖is as

follows:

𝑊𝑖 = 𝑔𝑗𝑤𝑗

∑ 𝑔𝑗𝑤𝑗𝑛𝑗=1

The Entropy weights for the data shown in Table.1 are

calculated as

We= [0.0645 0.0103 0.83000.0370 0.0589- 0.0015]

Table 1: QoS parameters

Parameter

/Network

Available

Band

Width

(Mbps)

Total

Band

Width

(Mbps)

Packet

Delay

(ms)

Packet

Jitter

(ms)

Path

Loss

(per 106)

Cost per

Byte

(price)

UMTS 1 1 2 37 7 50 0.6

UMTS 2 1.2 2 38 8 51 0.8

WiLan 1 6 11 125 15 50 0.1

WiLan 2 27 54 126 16 49 0.05

WiMax 1 32 60 80 6 48 0.5

WiMax 2 30 60 82 8 45 0.4

A comparative analysis of the networks Wi-LAN, Wi MAX,



5501

UMTS which are considered in this paper is given in table.2.

Various parameters that are compared are based on related

technologies, penetrations in the market, difficulties of the

vendor, buyers capacity, threat from new substitutes in the

market

Table 2: Comparison analysis of various networks

WLAN W iMAX UMTS

Peak Data Rate 802.11a,g=54Mbps

802.11b=11Mbps

UL:70Mbps

DL:70Mbps

DL:2Mbps

UL:2Mbps

Band Width 20Mhz 5-6Ghz 5Mhz

Multiple Access CSMA/CA OFDM/OFDMA CDMA

Duplex TDD TDD FDD

Mobility Low Low High

Coverage Small Mid Large

Standardization IEEE802.11x 802.16 3GPP

Target Market Home/Enterprises Home/Enterprises Public

RESULTS AND ANALYSIS

The performance of the algorithms described in the literature,

the selection of the best network by those algorithms for

various parameters like bandwidth, delay, jitter, loss, cost

factors specified in table.1 are calculated using standard

weights(SW) and Entropy Weights(EW). The efficiencies of

these algorithms are also calculated considering both the

weight matrix i.e. standard weights and entropy weights (SW,

EW).

a) Euclidean Distance Based Network Selection

Algorithm(EDBNS)

Figure 6: Network selection comparison for SW and EW for

Original base data


ABW


TBW

UMTS1 UMTS2 WLAN1 WLAN2 WIMAX1 WIMAX20

0.1

0.2

0.3

0.4

0.5

0.6

0.7

NETWORKS

VALU

ES O

F CO

ST F

UNCT

ION

EUCLIDEAN DISTANCE BASED NETWORK SELECTION

STANDARD WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

COST

FUN

CTIO

N

NETWORKS

PLOT FOR ORIGINAL DATA

STRANDARD WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

CHANGE OF AVAILABLE BAND WIDTH

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

COST

FUN

CTIO

N

NETWORKS

PLOT FOR AVAILABLE BANDWIDTH

STRANDARD WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

CHANGE OF TOTAL BAND WIDTH

VA

LUE

S O

F C

OS

T FU

NC

TIO

N


STANDARD WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

CO

ST

FUN

CTI

ON

NETWORKS

PLOT FOR TOTAL BANDWIDTH

STRANDARD WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

CHANGE OF PACKET DELAY

VA

LUE

S O

F C

OS

T FU

NC

TIO

N


STANDARD WEIGHT



5502


Packet Delay


Packet Jitter


Packet Loss


Cost

b) Rank Reversal Technique based Algorithm (RRTA)

Figure 13: Network selection comparison for SW and EW

for Original base data

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

0.4

0.5

0.6

0.7

0.8

0.9

1

CO

ST

FUN

CTI

ON

NETWORKS

PLOT FOR PACKET DELAY

STRANDARD WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

CHANGE OF PACKET JITTER

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

COST

FUN

CTIO

N

NETWORKS

PLOT FOR PACKET JITTER

STRANDARD WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

CHANGE OF PACKET LOSS

VALU

ES O

F C

OST

FU

NC

TIO

N


STANDARD WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

COST

FUN

CTIO

N

NETWORKS

PLOT FOR PACKET LOSS

STRANDARD WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

CHANGE OF COST

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

COST

FUN

CTIO

N

NETWORKS

PLOT FOR COST

STRANDARD WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

NETWORKS

VALU

ES O

F CO

ST F

UNCT

ION

RANK REVERSAL TECHNIQUE ALGORITHAM

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

CHANGE OF AVAILABLE BANDWIDTH

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT



5503

Figure 14: Network selection comparison considering SW

and EW for ABW


and EW for TBW


and EW for Packet Delay


and EW for Packet Jitter


and EW for Packet loss

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

CHANGE OF TOTAL BANDWIDTH

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FU

NCTI

ON

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


VALU

ES O

F CO

ST F

UNCT

ION

RANK REVERSAL TECHNIQUE ALGORITHM

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

CHANGE OF COST

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT



5504


and EW for Cost

c) Parameter Based Network Selection Algorithm

(PBNSA):


and EW for Original base data

Figure 21: Network selection comparisons for SW and EW

for change of ABW


for change of TBW


for change of Packet Delay


for change of Packet jitter

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS

PLOT FOR COST

STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

NETWORKS

VALU

ES O

F CO

ST F

UNCT

ION

PARAMETER BASED NETWORK SELECTION

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT



5505


for change of Packet Loss


change of Cost

d) Oliver Blume Algorithm Method(OBA)


Original base data


and EW for change of ABW


and EW for change of TBW


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

CHANGE OF COST

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

COST

FUN

CTIO

N

NETWORKS

PLOT FOR COST

STRANDARD WEIGHT

ENTROPY WEIGHT


0.5

1

1.5

2

2.5

3

NETWORKS

VALU

ES O

F CO

ST F

UNCT

ION

OLIVER BLUM METHOD

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.5

1

1.5

2

2.5

3

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.5

1

1.5

2

2.5

3


VALU

ES O

F CO

ST F

UNCT

ION

OLIVER BLUM METHOD

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60.5

1

1.5

2

2.5

3

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF AVAILABLE BANDWIDTH

STRANDARD WEIGHT

ENTROPY WEIGHT


0.5

1

1.5

2

2.5

3


VALU

ES O

F CO

ST F

UNCT

ION

OLIVER BLUM METHOD

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60.5

1

1.5

2

2.5

3

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF TOTAL BANDWIDTH

STRANDARD WEIGHT

ENTROPY WEIGHT



5506


and EW for change of Packet Delay


and EW for change of Packet Jitter


and EW for change Packet Loss


and EW for change of Cost

e) Similarity Based Network Selection Algorithm

(SBNSA)


and EW for Original base data


0.2

0.4

0.6

0.8

1

1.2

1.4


VALU

ES O

F CO

ST F

UNCT

ION

OLIVER BLUM METHOD

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60.4

0.5

0.6

0.7

0.8

0.9

1

1.1

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF PACKET DELAY

STRANDARD WEIGHT

ENTROPY WEIGHT


0.5

1

1.5

2

2.5

3


VALU

ES O

F CO

ST F

UNCT

ION

OLIVER BLUM METHOD

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.5

1

1.5

2

2.5

3

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF PACKET JITTER

STRANDARD WEIGHT

ENTROPY WEIGHT


0.5

1

1.5

2

2.5

3


VALU

ES O

F CO

ST F

UNCT

ION

OLIVER BLUM METHOD

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.5

1

1.5

2

2.5

3

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF PACKET LOSS

STRANDARD WEIGHT

ENTROPY WEIGHT


0.5

1

1.5

2

2.5

3

CHANGE OF COST

VALU

ES O

F CO

ST F

UNCT

ION

OLIVER BLUM METHOD

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.5

1

1.5

2

2.5

3

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF COST

STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

NETWORKS

VALU

ES O

F CO

ST F

UNCT

ION

MODIFIED SIMILARITY BASED NETWORK SELECTION

STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT



5507


change of ABW


and EW for change of T BW


change of packet delay


change of packet jitter


change of packet loss


and EW for change of cost


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

CHANGE OF AVALIABLE BAND WIDTH

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

COST

FUN

CTIO

N

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

CHANGE OF TOTAL BAND WIDTH

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF AVAILABLE BANDWIDTH

STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF TOTAL BANDWIDTH

STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF PACKET DELAY

STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8


VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF PACKET JITTER

STRANDARD WEIGHT

ENTROPY WEIGHT


0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

CHANGE OF COST

VALU

ES O

F CO

ST F

UNCT

ION


STANDARD WEIGHT

ENTROPY WEIGHT

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

COST

FUN

CTIO

N

NETWORKS

PLOT FOR CHANGE OF COST

STRANDARD WEIGHT

ENTROPY WEIGHT



5508

NETWORK SELECTION BY VARIOUS

ALGORITHMS

From the simulated results, the selection criterion of the best

network is given in table.3.According to the IEEE standards

specified in table.2 UMTS and Wi-Max are the best networks

in terms of various parameters that are considered like

available bandwidth, total bandwidth, delay, jitter, packet loss

and cost.

N1: UMTS-1, N2: UMTS-2, N3:WiLAN-1, N4:WiLAN-2,

N5:WiMax-1, N6:WiMax-2,

SW: Standard Weights, EW: Entropy weights

Table 3: Selection Of best Network by various algorithms

Parameters/

Networks Selection

Original Data Available

Band-width

Total

Band-width Packet delay Packet jitter Packet loss Cost

SW EW SW EW SW EW SW EW SW EW SW EW SW EW

OBAM N4 N4 N4 N4 N4 N4 N4 N4 N4 N4 N4 N4 N4 N4

RRTA N1 N1 N1 N1 N1 N1 N1 N1 N2 N1 N1 N1 N1 N1

EDBNS N5 N5 N1 N6 N5 N5 N5

PBNSA N1 N1 N1 N1 N1 N1 N6 N6 N1 N1 N1 N1 N1 N1

SBNSA N5 N5 N5 N1 N1 N5 N5 N5 N5 N5 N5 N5 N5 N5

EFFICIENCIES OF VARIOUS PROPOSED

ALGORITHMS

Figure 41: Efficiencies of EDBNS for various parameters

considering SW and EW.

Figure 42: Efficiencies of RRTA for various parameters


Figure 43: Efficiencies of PBNSA Algorithm for various

parameters for SW and EW.

ORG CHAB CHTB CHPD CHPJ CHPL CHCT0

5

10

15

20

25

30

35

CHANGE OF PARAMETRS

VALU

ES O

F EF

FICE

NCY


STANDARD WEIGHT

1 2 3 4 5 6 716

18

20

22

24

26

28

30

32

EFFI

CIEN

CES

NETWORKS

PLOT FOR CHANGE OF PARAMETRS

STRANDARD WEIGHT


5

10

15

20

25

30

35

40

45

CHANGE OF PARAMETRS

VALU

ES O

F EF

FICE

NCY

RANK REVERSAL TCHNIQUE ALGORITHAM

STANDARD WEIGHT

ENTROPY WEIGHT

1 2 3 4 5 6 734

35

36

37

38

39

40

41

42

43

EFFI

CIEN

CES

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


5

10

15

20

25

30

35

40

45

50

CHANGE OF PARAMETRS

VALU

ES O

F EF

FICE

NCY


STANDARD WEIGHT

ENTROPY WEIGHT

1 2 3 4 5 6 730

32

34

36

38

40

42

44

46

48

EFFI

CIEN

CES

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT



5509

Figure 44: Efficiencies of OBAM for various parameters


Figure 45: Efficiencies of SBNSA for various parameters


Table 4: Efficiencies of the networks based on parameter criterion

Parameters/

Network

Efficiency

Original Data Available

Band-width

Total

Band-width Packet delay Packet jitter Packet loss Cost

SW EW SW EW SW EW SW EW SW EW SW EW SW EW

Efficiency of

RRTA 35.9 39.4 35.8 42.5 34.9 39.7 38.7 39.5 37.6 39.5 36.2 39.7 36.2 37.2

Efficiency of

PBNSA 30.9 42.6 31.2 43.8 39.8 42.7 46.2 40.6 30.9 42.6 30.9 42.6 30.8 42.7

Efficiency of

OBAM 55.8 85.6 49.4 83.1 42.2 85.2 28.8 11.8 53.1 82.1 54.6 85.2 55.2 85.6

Efficiency of

MSBA

24.2 26.9 24.6 26.9 25.4 26.9 19.3 19.4 24.2 26.8 23.1 26.9 24.2 26.8

Efficiency of

EDBNS

21.6 16.6 17.3 30.1 20.3 21.1 21.5

ELUCIDATION OF RESULTS

The vertical handoff mechanisms based on the five algorithms

that were proposed in the literature are implemented. Two

different weight vectors are considered, one Wj taken from

IEEE standard calculations and the other We, which is

calculated using entropy method as discussed in section 3.

Using these weight matrices the performance of the five

algorithms which were are tested. The evaluation is done for

various QoS triggers like Bandwidth, delay, jitter, loss and

cost. The algorithms and their performance based on various

QoS parameters are simulated using MATLAB. Using the

evaluated results the best network among UMTS, WiFi, Wi-

MAX is selected. The efficiency of each algorithm is also

calculated.

For simulating the VHO models, the five QoS

parameters i.e. bandwidth, delay, jitter, loss and cost are

used. In this paper the investigation is to select the best

network among the networks that are taken (UMTS,

WiFi, Wi-MAX).

From figure 6, the performance of Euclidean Distance

Based Network Selection Algorithm (EDBNS) is

analysed considering the original data in table1.Figure

(7-12) show the performance of the algorithm in

selecting the best network considering the QoS

parameters like bandwidth, delay, jitter, loss and cost.

The algorithm performance is evaluated for both weights

(IEEE standard weights and Entropy weights).This

algorithm performance shows that N1 (Wi-MAX) is the

best network.

From figure 13 and figure 20 the performance of Rank


10

20

30

40

50

60

70

80

90

CHANGE OF PARAMETRS

VALU

ES O

F EF

FICE

NCY

OLIVER BLUM METHOD

STANDARD WEIGHT

ENTROPY WEIGHT

1 2 3 4 5 6 710

20

30

40

50

60

70

80

90

EFFI

CIEN

CES

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT


5

10

15

20

25

30

CHANGE OF PARAMETRS

VALU

ES O

F EF

FICE

NCY


STANDARD WEIGHT

ENTROPY WEIGHT

1 2 3 4 5 6 719

20

21

22

23

24

25

26

27

EFFI

CIEN

CES

NETWORKS


STRANDARD WEIGHT

ENTROPY WEIGHT



5510

Reversal Technique Algorithm (RRTA) and Parameter

Based Network Selection Algorithm (PBNSA) is

analyzed considering the original data in table1.Figure

(14-19) and Figure (21-26) show the performance of

these algorithms in selecting the best network

considering the QoS parameters. The performance of

these algorithms is evaluated for both weights(IEEE

standard weights and Entropy weights).This algorithm

performance shows that N1(UMTS) is the best network

From figure27, the performance of Oliver Bloom

Algorithm Method (OBAM) is analyzed considering the

original data in table1.Figure (28-33) show the

performance of the algorithm in selecting the best

network considering the QoS parameters. The algorithm

performance is evaluated for both weights(IEEE

standard weights and Entropy weights).This algorithm

performance shows that N4 (Wi-FI) is the best network

From figure34, the performance of Similarity Based

Network Selection Algorithm (SBNSA) is analyzed

considering the original data in table1.Figure (35-40)

show the performance of the algorithm in selecting the

best network considering the QoS parameters. The

algorithm performance is evaluated for both weights.

This algorithm performance shows that N5(Wi-MAX)

is the best network

Similarly in figure (41-45) shows the efficiencies of all

the five proposed algorithms are evaluated.

In table 3 and table 4 in section 6 summarizes the

performance of all the five algorithms. Results show the

efficiencies of all the five algorithms based on various

parameter criteria are calculated for both Standard

Weights and Entropy weights.

CONCLUSION

In this paper, five novel vertical handoff algorithms EDBNS,

RRTA, PBNSA, OBAM, SBNSA are proposed and compared

with each other. The simulation results display that the

performance of these algorithms are affected by the allocated

weight vector. According to the analysis and simulation

results, the five algorithms can achieve the satisfactory

performance in selecting the best network to which handoff is

to be performed, considering all the QoS criteria i.e.

bandwidth, delay, jitter, loss and cost. The work done in this

paper explores the multi-criteria approach based on QoS

which is used for initiating vertical handoff. The analysis

suggests that OBAM algorithm is the most efficient algorithm

with 85% efficiency (42%more than other algorithms) when

both standard weights and entropy weights are considered.

The work results also show that the PBNSA algorithm is

better than the other three proposed algorithms with an

efficiency of (43%).For future work, more comparisons with

other vertical handoff methods can be further discussed and

other techniques to solve the decision problem can also be

taken into account.

REFERENCES

[1] Aktu l Kavas “Comparative Analysis of WLAN,

WiMAX and UMTS Technologies” PIERS

Proceedings, August 27-30, Prague, Czech Republic,

2007

[2] E. Gustafsson and A. Jonsson, “Always best

connected”, IEEE Wireless Communications

Magazine, vol.10, no.1,pp.49-55, Feb. 2003.

[3] E.S. Navarro, V.W.S. Wong, Comparison between

vertical handoff decision algorithms for

heterogeneous wirelss networks, IEEE 63rd

Vehicular Technology Conference, Melbourne,

2006.

[4] F. Bari, and V. C. M. Leung, ―Multi-Attribute

Network Selection by Iterative TOPSIS for

Heterogeneous Wireless Access, in Proc. IEEE

CCNC’07 Las Vegas, NV, USA, January 2007.

[5] J. S. Dyer, ―Remarks on the Analytic Hierarchy

Process, Management Science, Volume 36 Number

3, Page 249-258, 1990

[6] Z. Ozturk, “A Review of Multi Criteria Decision

Making with Dependency between Criteria”, in Proc.

of MCDM 2006, Chania, Greece, June 2006.

[7] J.P.Brans ,P.Vincke, and B. Mareschal,“ How to

select and how to rank projects: the Promethee

method,” European Journal of

OperationalResearch,vol.24,no.2,pp.228–238,1986.

[8] E. Stevens-Navarro and V. W. S. Wong,

“Comparison between vertical handoff decision

algorithms for heterogeneous wireless networks,” in

Proceedings of the IEEE 63rd Vehicular Technology

Conference (VTC- Sping’06), pp.947–951,

Melbourne, Australia, May2 006.

[9] Oliver Blume, Method of Identifying a radio link,

Inventor: Oliver, Stuttgart (DE), patent US 7,616,

928 B2, date of patent Nov. 10, 2009.

[10] Deng,c.h.yeh and R.J willis,2000.Intercompany

comparison using modified Topsis with objective

weights. Computers and operations research, 27:963-

973

[11] Yeh, C.H., H. Deng and H. Pan, 1999. Multi-criteria

Analysis for Dredger Dispatching under Uncertainty.

Journal of the Operational Research Society, 50: 35-

43

[12] www.ieee.org

http://www.ieee.org/

multi-attribute network selection and evaluation models ... · multi-attribute network selection...

Documents