NETWORK PERFORMANCE ENHANCEMENT WITH OPTIMIZATION SENSOR PLACEMENT IN WIRELESS SENSOR NETWORK

International Journal of Wireless & Mobile Networks (IJWMN) Vol. 9, No. 1, February 2017
DOI:10.5121/ijwmn.2017.9702 9
NETWORK PERFORMANCE ENHANCEMENT WITH
OPTIMIZATION SENSOR PLACEMENT IN WIRELESS
SENSOR NETWORK
Gholamreza Farahani
Department of Electrical Engineering and Information Technology, Iranian Research
Organization for Science and Technology (IROST), Tehran, Iran
ABSTRACT
From one side, sensor manufacturing technology and from other side wireless communication technology
improvement has an effect on the growth and deployment of Wireless Network Sensor (WSN). The
appropriate performance of WSN has abundant necessity which has dependent on the different parameters
such as optimize sensor placement and structure of network sensor. The optimized placement in WSN not
only would optimize number of sensors, but also help to reach to the more precise information. Therefore
different solutions are proposed to reduce cost and increase life time of sensor networks that most of them
are concentrated in the field of routing and information transmission. In this paper, places which they need
new sensors placement or sensor movements are determined and then with applying these changes,
performance of WSN will calculate. To achieve the optimum placement, the network should evaluate
precisely and effective criteria on the performance should extract. Therefore the criteria should be ranked
and after weighting with using AHP algorithms, with use of Geographical Information System (GIS), these
weighted criteria will combined and in the locations which WSN doesn’t have enough performance, new
sensor placement will create. New proposed method, improve 21.11% performance of WSN with sensor
placement in the low performance locations. Also the number of added sensor is 26.09% which is lowest
number of added sensors in comparison with other methods.
KEYWORDS
Sensor, Wireless Sensor Network, Geographic Information System, Network performance, Analytical
Hierarchy process, Overlap Index
1. INTRODUCTION
Growth and progress of sensor technology and wireless communication causes movement to the
new generation of networks named WSN. Sensor placement in WSN is one of the important
issues in network design and implementation. Therefore determination of sensor placement
strategy is an essential step in design and optimization of network. As it is clear, optimum
placement of sensors lead to increment and enhancement of WSN performance. There are many
map points which it could be used to install sensors there that it will made a complicated problem,
also all of these map points will not improve the network performance. To increase network
performance, it requires considering different parameters to find the optimum place. For sensor
placement, the network should study carefully and extract effective parameters which with help
of these parameters, some sensors will install in locations that require new sensors. Geographical
Information System (GIS) is a set of combination of computer hardware and software, spatial

10
information and experts which designed and created for storage, update, processing, analyze and
presentation of spatial information.
Reference [1] achieves a sensor placement in water distribution network with use of ant colony
algorithm. In this paper with consideration limited number of sensors and limitation such as cost,
physical condition of sensors and possibility or impossibility of sensor installation between nodes,
proper location of sensors is chosen. Reference [2] proposed a placement of Tehran city pollution
measurement stations that with use of population, they determined stations and location of them,
also it proposed new stations with using stations measurement to obtain a better performance.
Reference [3] presents an algorithm for placement of disaster management sites. In this algorithm
with use of effective criteria and AHP method, locations for new disaster management sites are
proposed. Reference [4] proposed a program to solve a WSN problem to obtain a network
optimal configuration. The aim of this paper is finding a set of optimum placements to install
sensors and base stations to recover connection and covering requirements. The parameters which
are considered for this work were measurement range and field of view. Reference [5] proposed a
method for sensor placement and optimization of related variables in water distribution system
based on different scenarios and different optimum variables. They analyzed distribution of
network sensor location and with use of spatial information system; the best sensors placement is
calculated. Reference [6] presents a new strategy to determine optimal sensor placement in WSN.
This will carry out by minimization of spatial analysis of variance method (ANOVA) base on
selection of random points to determine sensor locations. Spatial analysis is carried out with use
of geostatistic analysis and optimization with Monte Carlo analysis. Reference [7] presents a
method for sensor implementation and programing to increase sensors coverage. They used a bee
algorithm to find a sensor placement and shown use of bee algorithm will provide worthy results.
Reference [8] states that different optimum algorithms are developed and used which most of
these algorithms are based on sensor models and simple network. They don't use environment
spatial information such as earth model. Therefor to implement an effective sensors network, it
proposed sensor network deployment with considering sensor spatial information and obtained
optimum sensor placement.
In optimization of sensors placement, important parameters which are effective on the sensor
placement such as sensors lifetime, sensor coverage area, slope, slope direction, distance from
adjacent node and distance from base station should be extracted. After parameter extraction,
these parameters should be delivered to experts on the sensor network to determine importance
degree of them. After it, with use of AHP algorithm, weight of each parameter related to the
WSN performance should be calculated. Then with GIS, the locations of network which have low
performance will determine and new sensors will place on those locations to obtain a better WSN
performance.
2. EFFECTIVE PARAMETERS IN SENSORS NETWORK
Several factors are involved in placement of sensors on the WSN that the most important which
has very effect in the sensors location are bring at continuation.
2.1. Network Structure
A sensor network can be built as hierarchical or flat. In hierarchical structure some sensors are as
a central sensor and collect environmental report of their neighboring sensors. The collected data
will send to base station via these central sensors. In flat structure, unlike the hierarchical

11
structure, role and operation of all sensors are same and information processing and sending
information to base station is sensors responsibility [9]. In this paper the network structure is
considered as flat.
2.2. Coverage
Coverage commonly defines as that how well the sensor network covers the intended region.
Appropriate coverage of sensors; will cause optimum energy consumption of sensors. Generally
targets in sensors network divide into three common categories, full supervisory on the region,
supervisory on set of targets and intrusion detection between natural barriers and environment.
Therefore coverage in sensor networks also divides to three categories regional coverage, target
coverage and barrier coverage. Regional coverage means each point inside intended region is at
least on the measuring range of one sensor and ideal is to cover whole region with minimum
sensor. Target coverage includes observation and supervisory on the defined number of targets.
Generally these type of coverage used in military applications [10-12]. Coverage evaluation
according to sensor coverage model which use for commutative coverage of placed sensors is
different. Therefore sensor measuring model has an important role in the evaluation and
calculation of coverage. Different models are proposed for coverage. In some studies coverage of
each sensor will decrease with increasing the distance from sensor. Equation (1) shows sensor
sensitivity values S of sensor S
i
at point P; ( , )d S P
i
is the Euclidean distance between sensor
and the point, λ is a positive parameter and γ is between 2 to 5 related to environment features
[13].
[ ]
( , )
( , )
S S Pi
d S Pi
λ
γ
= (1)
Other method is binary code which is used in this paper. In this model for each sensor a coverage
radius is considered and environment will convert into binary clusters. Points inside of the range
are covered points and points outside the range are outsight points. This model also named disc
model. Equation (2) shows a binary model which ( )C S
i
is S
i
sensor coverage with distance d
from point P and r
s
is coverage distance [14].
1 ( , )
( )
0
i s
i
d S P r
C S
otherwise
<
= 

(2)
In this paper with consideration measuring range of each sensor, some buffers (circular rings
around each sensor) will make. Regions which are not inside of buffers they will not be in the
sensor coverage and they are coverage gaps.
2.3. Type of Placement
Sensor placement has an important effect on the network efficiency and sensors life time. There is
different strategy for sensor placement which can be categorized into two general categories:
static placement and dynamic placement. In this paper the initial sensor placement is static,
although placement for new added sensors is dynamic. In static placement, first sensor
placements determine randomly or planned, and then with goal of network performance
improvement, new dynamic sensor placement should carry out.

12
2.4. Initial Sensor Placement
Sensors in WSN can be place in two types: planned placement and random placement. In
programmable implementation, sensor placement of flat architecture will choose selectively to
reach the goals of specific application. Generally the goals of this design is maximum coverage
range, minimum energy consumption and powerful network communication. Programmable
implementation is appropriate for a wide range of WSN applications. In some environments, it is
not possible to present there and install the sensor in specified location; therefore random
placement of sensors in environment is only possible solution [15, 16].
2.5. Energy Consumption
WSNs are constituted with hundred or thousand nodes which are spread out in the far off or
dangerous areas. Each sensor node includes sensor, an arithmetic logic unit, memory and wireless
communication unit with limited range which all of these units for continuation of their works
need an energy source. One of the important factors in sensor failure is finishing power supply.
Generally sensor energy is supplied with batteries which are not large and if they finish, should
be changed or charged. It is important to have sensor with low energy consumption; because the
sensor node should stay alive until end on its mission [17, 18]. In this paper, criterion will
consider for energy consumption and if the sensor has a more energy, measuring around it will be
more stable and probability of coverage gap around it will reduce.
2.6. Network Scalability
In sensor network, there are hundred or thousand nodes that each sensor in the necessary time
should connect to the destination. In such a network, may be some nodes shut down or new nodes
add to the network and network topology change. A network which tolerate these variations and
continue to work, named scalable network. Therefore node density should change from a few to a
few hundreds. Node density will calculate according to equation (3):
2
. .
( )
N R
n R
A
π
= (3)
where ( )n R is average number of nodes which are in the node range. In the equation (3), A is
working area, N is number of nodes and R is the radio transmission range [19].
2.7. Digital Elevation Model (DEM)
DEM is a digital file included elevations terrain without any extra definition about surface. DEM
files could be use to determine landform, slope and other aspects of terrain. Results show
elevation and slope have a destructive effect on the sensors network and whatever the slope is
more, the sensor network performance will decrease. There are many equations to review effect
of terrain on the sending signal over the sensor network which one the important of them is
Durkin algorithm. In this algorithm, deterioration of the path included two parts. Part one of
algorithm convert topographical DEM file to topographical database and reconstruct ground
profile information during the path between sender and receiver. Part two of algorithm calculates
expected failure during the path [20].

13
2.8. Communication range
Sensors have a range and radius measurement, moreover they have communication range which
is not constant in some sensors and can be changed based on conditions. Generally
communication range is more than sensor measurement range. As shown in figure 1, two Si and Sj
sensors have ability to communicate together when Euclidean distance of them be less than or
equal of minimum amount of communication rang [21].
Figure 1. Communication range between two sensors
3. MULTI-CRITERIA DECISION MAKING METHODS
Placement is a process which many factors and criteria have impact on it. It is clear that
evaluation of total factors which might help to choose best place is too hard. Therefore the model
should be used that consider most effective factors on the sensor placement. Placement is decision
making process with long term time, because many criteria should be tested and results of these
tests use for decision and the decision to be evaluated. This evaluation could be carried out in
different ways which intended criterion and aspects to be assessed simultaneously. These
evaluation methods named multi-criteria decision making methods. One of the common problems
in multi-criteria decision is different importance of criteria and sub-criteria for deciders.
Therefore relative importance for each of these criteria and sub-criteria to one another is needed.
Criteria extraction and its weighting are very important in multi-criteria decision. Each weight
will insert as a number in evaluation which determine the relative importance of that criterion
rather than other criteria.
There are different weighting methods to evaluate importance of criteria. Between these methods,
hierarchical analysis methods are on the effective decision technique which is stated with Thomas
L. Saaty in 1980 and until now it had variety applications in different sciences. This technique is
based on paired comparisons and gives an opportunity to managers to compare different
scenarios. Hierarchical analysis process begins with identification and prioritizing decision
elements. These elements includes: goals, criteria or features and probable options which is used
in priority. Elements identification process and relation between them lead to creation of
hierarchical structure which is called making hierarchy. Steps of this process are as follow.
3.1. Hierarchical Making
In the first step, hierarchical structure of subject will determine in the diagram format. In this
diagram, there is four level hierarchical includes: goals, criteria, sub-criteria and options. In
hierarchical structure, analysis process of complicated problems will carry out with synthesis of it
to simple details elements which are related together as hierarchical and main goal relation is
identified with the lowest level of hierarchical.

14
3.2. Determination of Importance Coefficient of Criteria and Sub-Criteria
To determine importance coefficient (weight) of criteria and sub-criteria, they compare pair wise
together. Base of judgment for comparison is according to Table 1, that dominant intensity of
criterion i to criterion j named aij. Pair wise comparison will record in n×n matrix. This matrix
will name binary comparison criteria matrix [ ]ij n nA a ×= . All elements of this matrix are positive
and with respect to inverse conditions principle in hierarchical analysis process, in each binary
matrix, there is two numerical values ija and 1 ija (if significance of i than j is equal k,
importance of j than i is 1 k ). Diameter elements of this matrix, with respect to equal importance
of each criterion to itself to reach the goal are one.
Table 1. Nine quantity criteria comparison
DescriptionDefinitionScore
Two activities contribute equally to the objectiveEqual importance1
Weak or slight2
Experience and judgement slightly favour one activity
over another
Moderate importance3
Moderate plus4
Experience and judgement strongly favour one activity
over another
Strong importance5
Strong plus6
An activity is favoured very strongly over another; its
dominance demonstrated in practice
Very strong or
demonstrated importance
7
Very, very strong8
The evidence favouring one activity over another is of
the highest possible order of affirmation
Extreme importance9
3.3. Determination of Options Importance Coefficient
In this step, priority of each option related to sub-criteria will judge. If there is no sub-criterion for
criterion, directly judge with criterion. The base of this judgment is nine quantity criteria of Saaty,
with this difference that in comparison with other options related to each sub-criteria or criteria,
priority intensity of option is matter not importance of it.
3.4. Determination of Final Score Options
In this step, by combination of importance coefficients, final score of each option will determine.
For this, with using Saaty hierarchical combination principle, a priority vector by considering all
judgments in all hierarchical levels will calculate as equation (4).
1 1
( ) ( )
n m
k i ij
k i
priority Final Score of j W W g
= =
= ∑ ∑ (4)
where kW is importance coefficient of criterion k, iW is importance coefficient of sub-criterion i
and ijg is score of option j related to sub-criterion i.

15
3.5. Reviewing Compatibility in Judgments
One of the advantages of hierarchical analysis process is capability of reviewing judgment
compatibility for calculation of criteria and sub-criteria importance coefficient. When criteria
significant rather others meet, there is probability of inconsistently in judgments, it means if iA is
important than jA and jA is important than kA , logically iA must be important than kA , but unlike
of all efforts, often priorities and emotions of people is inconsistent.
Mechanism which Saaty is considered for reviewing incompatibility in judgments is calculation
of compatibility indices named I.R., which will calculate with dividing Incompatibility Indices
(I.I.) by Random Indices (R.I.). If this indices is less or equal 0.1, compatibility in judgments are
acceptable, otherwise judgments should revise. In other words the binary comparison matrix
should calculate again. Incompatibility Indices will calculate as equation (5).
max
. .
1
n
I I
n
λ −
=
−
(5)
where maxλ is maximum of Eigen value and n is number of criteria. Random Incompatibility with
respect to number of criteria (n) is extractable from table 2.
Table 2. Random Incompatibility with respect to number of criteria
In geometric mean method which is approximate method, instead of calculation maximum Eigen
value (λmax), L will use as equation (6).






= ∑=
)(
1
1
n
i i
i
W
AW
n
L (6)
where iAW is a vector which obtain from multiply binary comparison criteria matrix (matrix A)
by iW (weight vector or criteria importance indices) vector. This technique has some limitations
which main of this is related to reversing scores in heirechical analysis process method . AHP
assumes goals, criteria and options are in the hierechical structure and relations between them is
linear and one way. This assumption always is not correct. In this condition, results of AHP
method may be causes ranks reverse. Saaty presens other mehtod named Analytic Network
Process (ANP) which could be used in this condition. AHP method is usefull which creates
condition for analysis and conversion of complicated problems to logical and simpler hierechy
that programmer could evaluate options with help of criteria and sub-criteria easily [22].
4. GEOGRAPHIC INFORMATION SYSTEM (GIS)
GIS is a computerized system to record, storage, edit, update, analysis and monitoring of spatial
data with coordinate. Spatial data with coordinate includes data which have position and
geographical features such as river, roads and land form. Since the spatial data in GIS inspire of

16
traditional methods store as digital, therefore updating, analysis, monitoring and editing of them
will feasible with more speed and precision.
4.1. GIS Components
GIS apply four fundamental components about terrain data which are data input, data
management, data analysis and output data.
4.1.1. Data Input
Data input component, convert data to usable form in GIS. Generally georeference data are in
paper maps, table of descriptive information, electronic files of maps and descriptive information
of them, aerial photos or satellite images. Data insertion may be easier than changing the file
format or may be more complicated. Large databases creation may be have cost five or ten times
of GIS hardware and software.
4.1.2. Data Management
Data management includes functions for storage, keeping and retrieval information inside of
databases. There are different methods for data organization as files which computer could read
them. Data structure is a method which data are organized based on it and files relation to each
other will determine limitation on the information retrieval and speed of retrieval operation.
4.1.3. Data Analysis
The main goal in GIS is preparing the situation for spatial analysis. Therefore, information
protection capability and map production is limited. In normal condition, spatial analysis could
make easier the decisions making. Analysis is powerful component that GIS give to programmer
and designer. Quantitative feature of GIS, give this capability to researches to use mathematics
and statistics as a complementary. Analysis functions and works with data in GIS, is determinant
of information that this system could creates.
4.1.4. Output Data
Data output in different GIS based on quality, precision and ease of use has a much more
diversity than these system capabilities. Output data might be in different forms such as map or
table of values or writing, and presents on the paper or digital. Required output functions will
determine based on user requirement, therefore user interference is very important in
determination of required outputs.
4.2. Data Types in GIS
One of the main difference of GIS with other informative systems is the nature of data used in
this system that are recognize as georeference spatial data or coordinate spatial data that are
belong to specified place and position on the ground. In GIS, information about situation or
spatial phenomena is called spatial data and qualitative and quantitative characteristics of
phenomena are called descriptive data.

17
4.3. Layers Combination
In informative systems, data store in different tables. Available spatial data in spatial information
systems for increasing their performance and operational capabilities will store in different layers.
Each of the informative layers, store one of the data and these layers in interact with one another,
provides capability of various analyses. Mostly for modeling and analysis, it is needed to combine
different information layers. For combination of layers, there is different methods that most
important of them is Overlap Index method.
4.3.1. Overlap Index Method
In overlap index model, in addition to applying weight of each single factor than other factors,
each classes and spatial units in factors will weight based on relative importance and expert
comments. After weighting operation, factors combine together according to equation (7).
∑
∑
i
iji
W
SW
=S (7)
where iW is i-th map weight, ijS is j-th class of i-th map and S is the value of each spatial unit in
output map. This model have a more priority and flexibility than Boolean logic model, but linear
nature and inability in determination of correct weighting variations relevant to border classes of
each factor is disadvantages of it.
4.4. Clustering
Clustering defines as the data set classification process based on similarity to subsets named
cluster. This classification will carry out such that member of each set had a more similarity
together and more difference to other members. There are different techniques for data clustering.
Among these methods, Agglomerative Hierarchical Clustering (AHC) is an appropriate method
which without any pervious information of data pattern, categorization is possible. Hierarchical
clustering algorithms display data as a tree, which is called Dendogram hierarchical tree.
Dendogram composed of nodes layers which each of them shows a cluster. To obtain similarities
between data in hierarchical clustering, there is different methods that are Ward's method,
Complete linkage, Simple linkage, Strong linkage, Flexible linkage, Unweighted pair-group
average and Weighted pair-group average.
In this paper, Ward's method is used. This method uses a criterion to decrease outlier data losses.
Ward’s method uses the sum of squared difference between each data of cluster and mean vector
of that cluster as a criterion for assessing a cluster. Ward's algorithm is as follows.
1. First each data will consider as a cluster.
2. For total possible clusters pair of clusters set, two clusters will select which sum of
squared difference between cluster data of community of them and the mean cluster is
minimum.
3. Two selected cluster will combine
4. Until number of clusters reach to intended number, steps 1, 2 and 3 will repeat.

18
Ward's method states that the distance between two clusters, A and B, how the distance of squares
will increase when they combine [23]. The equation (8) shows the relation.
2 2 2 2
( , ) A B
i A B i A i B A B
A Bi A B i A i B
n n
A B x m x m x m m m
n n∈ ∈ ∈
∆ = − − − − − = −
+
∑ ∑ ∑U
U
r r r r r r r r
(8)
where jm
r
is the centre of cluster j, jn is the number of points in cluster j and ∆ is merging cost
of combining the clusters A and B.
5. GEOSTATISTICS
With use of geostatistic analysis, easily could create a map or continuous area of sampled points.
Sampled values can be hauteur, temperature, air pollution and etc. Also geostatistic analysis could
be used for spatial modeling in different applications. One of the advantages of this analysis is
fitness of model area to sampled points. Geostatistic analysis is going to find the way for spatial
continuously description and also statistic and deterministic tools collections and spatial
variations modeling.
5.1. Geostatistic Definition
Geostatistic in European concept is a branch of statistics which data coordinate of under study
population and consequently spatial structure of data will study. In classical statistic, generally
samples which taken from population assume random. In other words, it assumes samples are
independent of each other, therefore existence of one sample, will not provide any information
about next sample. But in geostatistic, samples are not independent and according to this theory,
adjacent samples to the certain distance have a spatial dependency. Therefore spatial position of
samples with intended quantity value can be analyzed together. This spatial communication in
mathematical formats will name spatial structure. Thus in geostatistical, variables will be
discussed that they have spatial structure.
5.2. Interpolation
Sensors in WSNs have been placed in specific locations and they have ability to measure intended
parameters in that point up to limited radius. Therefore these points information should convert to
area information. One of the effective methods in this field is usage of interpolation techniques.
According to the researches, geostatistics methods are one of the best methods for spatial data
interpolation. There are different geostatistics methods which most important of them are: Inverse
Distance Weight (IDW), Spline and Kriging. Between these methods Kriging is one of the best
and most applicable for interpolation. Therefore in this paper for map creation, this interpolation
method is used.
5.2.1. Kriging Interpolation
Generally, interpolation which carry out based on spatial statistical estimation models is process
that it could be possible to obtain quantity in undefined coordinate with use of same quantity in
other points with define coordinate. Most important spatial statistical estimator is Kriging.
Kriging is an estimator method which is based on weighted moving average and it is the best
unbiased estimator. Equation (9) shows the kriging equation.

19
∑=
=
n
i
iiZZ
1
ˆ λ (9)
where ˆZ is an approximated parameter, iλ is a weight or importance quantity related to i-th
sample and iZ is a specified parameter. This type of kriging is linear, because it is linear
combination of n data. For using this estimator, Z variable should have a normal distribution
otherwise nonlinear kriging should be used or normalized the variable with usage of statistical
conversions. Kriging estimator has two brilliant features. Firstly it doesn't have systematic error.
Secondly it has minimum estimation variance.
Necessities of first condition establishment is average of estimation error be zero. Therefore it can
be written as equation (10).
1
( ) 0
n
i i i
i
E Z Zλ
=
− =∑ (10)
By simplification of equation (10), equation (11) is as below.
1
( ) 0
n
i i i
i
E Z E Zλ
=
 
− = 
 
∑ (11)
Also ( )iE Z m= , therefore
1
0
n
i
i
m mλ
=
− =∑
Because m≠0, thus equation (12) must establish.
1
1
n
i
i
λ
=
=∑ ( 12)
Therefore to minimize kriging estimate variance, it is necessary to minimize it in term of kriging
indices. There are different methods for kriging which are simple kriging, ordinary kriging, token
kriging, probabilistic kriging and clustered kriging [24, 25].
6. IMPLEMENTATION
The study area in this paper is Grand Saint Bernard Pass with 2400 meter length in border of
Switzerland and Italy. This passage is in the mountain ranges between two countries which from
north lead to Switzerland and from south lead to Italy. The studied WSN in this paper is the
established WSN around the Grand Saint Bernard Pass named SensorScope which is
implemented with EPFL University in 2007. Data of this sensor network download from web site
with address: http://lcav.epfl.ch/page-86035-en.html. This WSN has 23 sensors which measure
environmental data such as ambient temperature, land surface temperature, sun radiation, relative
humidity, soil humidity, speed and direction of wind.
6.1. Data Collection and Preparation
To evaluate and locating of sensors in WSN, effective criteria should identified and their impact
to be studied. Table 3 shows the most important effective parameters and their impact. According
to effective criteria, required data must be collected and prepare for usage in ArcGIS software.
For calculation and analysis, types of all information layers in ArcGIS software will raster. In
raster model, whole map area divide to network of tiny and regular cells that named pixel. In fact,

20
raster type is n×m arrays of pixels. Each of these pixels is indicator of one position in the earth
area and corresponds to spatial range in real world. In each pixel, a numerical value will save
which indicates nature of that place in real world that is called Digital Number (DN).
Table 3. Most important parameter with their impact on WSN performance
Impact of criteria on WSN performanceCriteria Name
Whatever hauteur is higher, the network performance is lowerDEM
Whatever coverage is less, measuring cavity is moreSensor coverage
range Whatever communication range is less, network performance decreasesSensor
communication Sensors energy has a more effect on the network performanceSensor energy
consumption More scalability increase the network performanceScalability
The reason for usage of raster format instead of vector format is that calculation and overlap
analysis in raster model is easier. Station coordinate and features of them is prepared in Excel
format. Then this file is inserted into GIS environment and layer is created. The contents of this
layer are shown in table 4. This table includes station coordinate in UTM and WGS-84, hauteur,
coverage, energy and communication range of stations
6.2. Classification
6.2.1. DEM
Required DEM for study area is obtained from Earth Data website which belongs to NASA and
included 30 meter DEM of whole earth [26]. This region is in Europe with N45E008. Figure 2
shows a categorized DEM of region which is used in this paper. As it shown in figure 2, different
hauteur has different colors which are classified in five classes. Low hauteur has a white color
and high hauteur is darker. These DEM data are prepared in a raster format which will use in GIS
and ArcGIS software for continuation of simulation in the paper.
Table 4. Data collection from 23 stations in studied area

21
Figure 2. DEM of studied area
6.2.2. Coverage range
The coverage range of sensor in simulation is 50 meter and sensor measurability is considered 50
meter. To model the coverage range, the buffering feature of GIS is used. Buffering in GIS is in
the spatial analysis. Figure 3 shows a buffering map of sensors (coverage of sensors). In figure 3,
white color means better coverage and darker color becomes fewer converges.
Figure 3. Coverage map of network sensors
6.2.3. Communication Range
Sensor communication range is considered 100 meters. For communication range, same as
coverage, buffering is used and then the map of it is shown in figure 4. Same as figure 3, in figure
4, white colour means better communication and darker color becomes less communication.
Figure 4. Sensor communication range map
6.2.4. Energy Consumption
Energy consumption has an importance in WSN. Sensors of SensorScope network have 3-state
for energy consumption, State 1 – using a constant battery, State 2 – rechargeable batteries and
State 3 – solar panel.

22
The idealist one is using these three states. In the studied network at this paper, some sensors have
batteries as state one, some have power provision as state two and some include all three states
for power generation. However the sensor uses a more sustainable energy, measurement around it
will carry out with more stability. Therefore to provide map of energy consumption of area,
sensors are classified to three categories and according to it, map of network energy consumption
will produce. Number one in this classification means lower energy stability and number three
means highest energy stability. To calculate network energy consumption map, interpolation
methods is used. One of the effective methods is Kriging interpolation that is used in this paper
for preparing energy consumption map. Interpolation methods in GIS are as statistical analysis of
geostatic. Energy consumption map of studied area with use of Kriging interpolation is shown in
figure 5. Darker color has a lower energy and white color has a higher energy.
Figure 5. Energy consumption map of studied area
6.2.5. Scalability
Network scalability is a withstanding of network against variations. To prepare scalability map in
different points of network, first scalability of each sensor should be calculated and then with use
of interpolation methods, it will extend to the others and obtain a scalability map. To count
number of sensors in each sensor range for scalability, spatial joint is used which will create a
field includes number of sensors as normalized in Table 5. Finally to obtain scalability for whole
points in the area like as energy consumption map, kriging interpolation is used. Scalability map
is calculated for studied area and shown in figure 6. Scalability in stations seven and seventeen is
zero which is a darkest region in the map.
Table 5. Normalized network scalability in each sensor

23
Figure 6. Scalability map of studied area
6.3. Reclassification
The maps that are prepared in GIS are called layer. The prepared layers are in the raster format. A
raster layer with respect to its resolution is made with some isometric pixels which each of them
has its own numerical value. The operation of homogenization and clustering of data in rasterized
layer will call reclassification. With use of reclassification, after separating pixels worth in
classes, a rasterized layer will produce. With combination of rasters in one class, homogeneous
classes will make. In reclassification, number of classes, distance between classes and numerical
worth of each class will determine. Each produced maps will classify in five class, and value of
each class with respect to effect of it on the WSN is valued one to five that number one has
lowest worth and number five has highest worth. At continuation, reclassification for each map is
shown.
6.3.1. DEM
However hauteur is more, the communication between sensors is harder. Therefore in
reclassification of DEM, classes with higher hauteur will have lower worth. Table 6 shows a
reclassification of DEM in GIS.
Table 6. Reclassification of DEM in GIS
Old values 2276-2330 2330-2375 2375-2418 2418-2466 2466-2534
New values 5 4 3 2 1
In figure 7 output of DEM reclassification is shown. Higher hauteur has a darker color, which
means is not worthy situation for WSN.
Figure 7. Reclassification output of DEM

24
6.3.2. Coverage Range
In coverage reclassification, the aim is whole region in the studied area lay inside the coverage
range. Hence the regions inside of coverage range have maximum worth five and regions outside
of coverage range have a minimum value zero. Therefore sensor coverage range map will classify
into two classes. Figure 8 shows a reclassification output of sensor coverage range map that white
colour is points within and black color is point out of coverage range.
Figure 8. Reclassification output of sensor coverage range map
6.3.3. Communication Range
Reclassification of communication range map is similar to coverage range map and it has to
classes. Class 0-100 meter has most worth five and outside of this range has a least worth zero.
Figure 9 shows a reclassification for communication range. As shown in figure 9, most study area
is inside of sensor communication range and only regions with dark colors are cavity
communication.
Figure 9. Output map of reclassification communication range
6.3.4. Power Consumption
As explained in section 6.2.4, however number assigned to the energy consumption in energy
map is higher; it means that region has more stability, therefore in reclassification energy
consumption map; higher classes have a more worth. In figure 10, output map of reclassification
for energy consumption is shown. Whatever the colours are brighter, energy consumption in the
region is better.

25
Figure 10. Output map of reclassification energy consumption
6.3.5. Scalability reclassification
However scalability in each region of network is higher, the network will be more stable. Hence
for scalability reclassification map, higher classes have more worth. Figure 11 shows output map
for scalability reclassification. As shown in figure 11, brighter color means more scalability.
Figure 11. Output map of scalability reclassification
6.4. Calculation of Criteria Weights
After selecting criteria and collecting data, implementation and sensor placement process will
carry out. To improve a performance of WSN, the criteria should be weighted. These weights will
acquire with experts in the WSN field. Then with use of AHP method, in the expert choice
software, the optimum placement for sensor will obtain. Finally with use of Ordered Weighted
Averaging (OWA) method, layers in ArcGIS software will combine and places that they require
to place new sensor on them will determine. After providing, preparing and reclassification of
criteria map, these maps should be combined together to determine regions which require to
install new sensors. For considering the weight for each criterion, multifactor decision methods
should be used. One of the best and effective methods is AHP. To implement AHP method, some
questionnaire prepared and experts in WSN field, with comparison of each two criteria together,
they determined importance value of each criterion. Then the answers to these questions were
inserted to the expert choice software to get a weight of each criterion. Table 7, shows pair wise
comparison and inconsistent coefficient which has obtained from questionnaire.
Table 7. Pair wise comparison and inconsistent coefficient
Criterion Energy DEM Scalability Coverage Range
Energy 3.0 3.0 5.0 5.0
DEM 5.0 7.0 3.0
Scalability 3.0 5.0
Coverage 5.0
Range Incon : 0.10

26
As it shown in Table 7, the inconsistent coefficient is 0.10 which is in the allowable limit for this
coefficient.
6.5. Combination of criteria map and preparing final map
Table 8 shows a final weight for each criterion. After determination of criteria importance, with
use of hierarchical analysis process, the criteria should combine together to determine the
locations that require new sensor establishment. Output map is obtained from criteria map
combination such that each pixel shows a value to determine the suitability of the location where
the new sensor is placed with respect to all effective factors. Considering that the criteria used in
different ranges have different worth; therefore in this paper with use of overlap index method for
criteria map combination, the output is calculated and shown in figure 12. The darker area in
figure 12 means installation of new sensor in that region is more necessary. Therefore five
sensors were installed in the regions which they sever need to the sensor installation. Figure 13
shows locations of new sensors (sensors 24 to 29).
Table 8. Final weight calculated for each criterion with AHP method
Criterion Energy DEM Scalabilit Coverage Range
Weight 0.139 0.068 0.258 0.490 0.045
Figure 12. Criteria combination output with overlap index method
Figure 13. New sensor installation map
After installation of these new sensors, all implementation steps have been carried out and results
are shown in figure 14. As it is clear in this figure, after adding some new sensors in the dark
region which had low performance, network performance is increased.

27
Figure 14. New sensor locations based on output of overlap index method
Table 9 shows a pixel value in each station before and after adding sensors and also the amount of
network average performance improvement. As it is clear, network performance is increased
averagely 21.11%. This performance improvement for the new sensors installation is
considerable.
Table 9. Performance improvement for sensor placements
The ant colony algorithm with restriction of 30 sensors and cost as proposed in reference [1]
implemented in this study area. Also the proposed method of reference [4] which considers only
coverage parameter (same as our propose method in section 6.2.2 (50 m)) is implemented. Also
ANOVA method is implemented as stated in [6] with DEM parameter as section 6.2.2. Bee
algorithm is also use for sensor placement to get the best coverage (section 6.2.2) as reference [7].
In reference [8] which spatial information is used is also implanted for the studied area section 6.
Table 10 shows the performance improvement for these methods. As it is clear the proposed
method has a better performance improvement in comparison to other proposed methods. The
main reason which proposed method in this paper has a better improvement is that it considers all
of the important parameters together with proper weight for sensor placement which other
methods only consider some of these parameters for network performance improvement. Also in
Table 10, the percentage of added number of sensors is written which proposed method has a
lowest percentage of added sensors with 26.09%.

28
Table 10. Performance Comparison of Different Methods
For better clearance in figures 15 and 16 the bar chart of the results are shown.
Figure 15. Performance improvement percentage of different methods
Figure 16. Percentage of added number of sensors

29
7. CONCLUSIONS
In this paper, first most important and effective parameters which have effect on the network are
studied that these parameters are: coverage, communication range, scalability and DEM. After
determination of effective criteria, required data are prepared to insert to GIS. In fact GIS creates
an environment to combine and carry out different analysis on data. After insertion data into GIS,
with use of AHP method, to compare different criteria with themselves, relative worth of each
criterion is determined. At the end, with use of analysis and toolboxes in GIS, these criteria are
combined together and places which require a new sensor installation are determined. Results
shown whenever hauteur is more, connections between sensors are harder and related region
requires more sensor installation. The performance of proposed method, will improve 21.11%
with sensor placement in the low performance locations. Also the percentage for number of added
sensor is 26.09% that is lowest number between others compared methods.
REFERENCES
[1] Miri S. M. & Afshar A., (2014) "Optimum Layout for Sensors in Water Distribution Networks
through Ant Colony Algorithm: A Dual use Vision", Water and Sewage Quarterly, Vol. 25, No. 3, pp
67-75.
[2] Ashrafee Kh., Ghader S., Esfehanian V. & Motasadi S., (2007) "Air Pollution Measurement Layout in
Large Tehran", Journal of Environmental Study, Vol. 33, No. 44, pp 1-10.
[3] Bahrampour M. & Bemaniam M. R., (2012) "Study on Optimum Location of Disaster Management
Sites with use of GIS", Journal of Emergency Management, Vol. 1, No. 1, pp 51-59.
[4] Osais, Y.E., M. St-Hilaire & R.Y. Fei, (2010), "Directional sensor placement with optimal sensing
range, field of view and orientation", Mobile Networks and Applications, Vol. 15, No. 2, pp 216-225.
[5] Isovitsch, S.L. & J.M. VanBriesen, (2008), "Sensor placement and optimization criteria dependencies
in a water distribution system", Journal of Water Resources Planning and Management, Vol. 134, No.
2, pp 186-196.
[6] Castello, C.C. & Fan J., (2010), "Optimal sensor placement strategy for environmental monitoring
using wireless sensor networks", 42nd Southeastern Symposium on System Theory (SSST), pp 275-
279.
[7] Mini, S., S.K. Udgata & S.L. Sabat, (2014), "Sensor deployment and scheduling for target coverage
problem in wireless sensor networks", IEEE sensors journal, Vol. 14, No. 3, pp 636-644.
[8] Argany, M., (2015), Development of a GIS-Based Method for Sensor Network Deployment and
Coverage Optimization, PhD Thesis, Université Laval.
[9] M. Marks, (2010), "A survey of multi-objective deployment in wireless sensor networks", Journal of
Telecommunications and Information Technology, pp 36-41.
[10] C.Zhu, C. Zheng, L. Shu & G. Han, (2012),"A survey on coverage and connectivity issues in wireless
sensor networks", Journal of Network and Computer Applications, Vol. 35, No. 2, pp 619-632.
[11] S.Megerian, F. Koushanfar, M. Potkonjak & M.B. Srivastava, (2005), "Worst and best-case coverage
in sensor networks", IEEE transactions on mobile computing, Vol. 4, No. 1, pp 84-92.
[12] S.R.Angajala, (2012), "Coverage Problems in Sensor Networks", IJECCE, Vol. 3, No. 1, pp 104-110.
[13] A. Ghosh & S.K. Das, (2006), "Coverage and connectivity issues in wireless sensor networks",
Mobile, wireless, and sensor networks: Technology, applications, and future directions, pp 221-256.
[14] K.Chakraborty, S.S. Iyengar, H. Qi & E. Cho, (2002), "Grid coverage for surveillance and target
location in distributed sensor networks", IEEE transactions on computers, Vol. 51, Vol. 12, pp 1448-
1453.
[15] D.S. Deif & Y. Gadallah, (2014), "Classification of wireless sensor networks deployment techniques",
IEEE Communications Surveys & Tutorials, Vol.16, No. 2, pp 834-855.
[16] M. Leoncini, G. Resta & P. Santi,(2009), "Partially controlled deployment strategies for wireless
sensors", Ad Hoc Networks, Vol. 7, No. 1, pp 1-23.
[17] F.Bouabdallah, N. Bouabdallah & R. Boutaba, (2009), "On balancing energy consumption in wireless
sensor networks", IEEE Transactions on Vehicular Technology, Vol. 58, No. 6, pp 2909-2924.

30
[18] F.Ishmanov , A.S. Malik & S.W. Kim, (2011), "Energy consumption balancing (ECB) issues and
mechanisms in wireless sensor networks (WSNs): a comprehensive overview", European
Transactions on Telecommunications, Vol. 22, No. 4, pp 151-167.
[19] L.K.Alazzawi, A. M. Elkateeb & A. Ramesh, (2008), "Scalability analysis for wireless sensor
networks routing protocols". International Conference on Advanced Information Networking and
Applications-Workshops (AINAW), pp 139-144.
[20] M.Vuckovik, D. Trajanov & S. Filiposka, (2010), "Durkin’s propagation model based on triangular
irregular network terrain", In International Conference on ICT Innovations, Springer.
[21] R.Shorey, A. Ananda, M.C. Chan & W.T. Ooi, (2006), Mobile, wireless, and sensor networks:
technology, applications, and future directions. John Wiley & Sons.
[22] E. Triantaphyllou & S. H. Mann, (1995), "Using the analytic hierarchy process for decision making in
engineering applications: some challenges", Vol. 2, No. 1, pp 35-44.
[23] P. Berkhin, (2006), A survey of clustering data mining techniques, in Grouping multidimensional
data, Springer. pp 25-71.
[24] N. Cressie, (1990), "The origins of kriging", Mathematical geology, Vol. 22, No. 3, pp. 239-252.
[25] Oliver, M.A. & R. Webster, (1990), "Kriging: a method of interpolation for geographical information
systems", International Journal of Geographical Information System, Vol. 4, No. 3, pp 313-332.
[26] EarthData availabe from: dwtkns.com/srtm30m.
Author
Gholamreza Farahani received his BSc degree in electrical engineering from Sharif
University of Technology, Tehran, Iran, in 1998 and MSc and PhD degrees in
electrical engineering from Amirkabir University of Technology (Polytechnic),
Tehran, Iran in 2000 and 2006 respectively. Currently, he is an assistant professor in
the Institute of Electrical and Information Technology, Iranian Research Organization
for Science and Technology (IROST), Iran. His research interest is Wireless Network
especially Wireless Sensor network.

NETWORK PERFORMANCE ENHANCEMENT WITH OPTIMIZATION SENSOR PLACEMENT IN WIRELESS SENSOR NETWORK

More Related Content

What's hot (17)

Viewers also liked (16)

Similar to NETWORK PERFORMANCE ENHANCEMENT WITH OPTIMIZATION SENSOR PLACEMENT IN WIRELESS SENSOR NETWORK (20)

Recently uploaded (20)

NETWORK PERFORMANCE ENHANCEMENT WITH OPTIMIZATION SENSOR PLACEMENT IN WIRELESS SENSOR NETWORK