International Journal of Computer Science Trends and Technology (IJCST) Volume 2 Issue 1, Jan-Feb 2014

ISSN: 2347-8578 www.ijcstjournal.org Page 55

A Clustering Protocol for a Non Uniformly Distributed Wireless
Sensor Network
Ashutosh S. Dube
1
, Shailesh A. Shep
2
, Shrikant M. Gavane
3
, Soma N. Ghosh
4
,
Department of Computer Engineering and Information Technology, College of Engineering
Pune, 411 005,
India

ABSTRACT
Considering the deployment of wireless sensor networks in areas of remote access to monitor them for various reasons,
adapting the network for better inter- cluster and intra-cluster communication to prolong its working time is necessitated.
Using optimal parameters for communication taking into account the capabilities of the sensors in the network, Energy
Efficient clustering algorithm for Maximizing Network Lifetime aims to improve the traditional hierarchical protocols. But
since the nodes may not be uniformly deployed over the entire area, the region under consideration also forms an important
determinant of these optimal parameters. The protocol aims at merging the clusters, initially divided to from hierarchical
network, so that in a non-uniform network the values of various network parameters may be adjusted to their most near
optimal values. Moreover, re-clustering and re-electing the cluster heads after a threshold also helps in increasing the network
lifetime.
Keywords- Cluster, Wireless, Sensor, Networks

I. INTRODUCTION
The ability to communicate over distances without any help
of attached cables makes it possible for wireless sensors to
be deployed at remote areas with ease. But the sole source
of energy, its battery, can be charged only once that is
before its deployment and thus limits the network working
time. Once deployed, its almost impossible to recharge
them again. In recent years, many protocols have been
aimed at utilizing this battery power in a way that network
lifetime is maximized and network as a whole remains
connected. Hierarchical protocols like LEACH (low-energy
adaptive clustering hierarchy) algorithm, is based on
gradient cluster reducing the energy consumption of nodes
and enhancing the network lifetime. Clusters formed have
some nodes working as ordinary sensors while others act as
cluster heads working on the behalf of the entire cluster.
LEACH is one of the basic Hierarchical clustering protocol
wherein nodes communicate through a series of other
communicating nodes to the base station. The approach
helps in providing a more fair communication overhead for
the nodes situated far away from the base station in the
network. But even then problems like uneven cluster sizes
exist. An improvement to LEACH was the Hybrid Energy
Efficient Clustering (HEED) algorithm, which made use of
dynamic clustering and cluster head rotations depending
upon the node proximity and the residual energy of the
node. But additional energy is spent in the process.
Energy Efficient Clustering Algorithm for Maximizing
Network Lifetime uses the same logic of re-clustering and
reticulation in a hierarchical form but in a more static way as
compared to HEED. Network is divided into static clusters
based on the network optimal parameters and then in these
uniform clusters the re-clustering mechanism is executed to
find the new cluster heads. Since in previous networks the
network clusters were not uniform, a fixed network
topology was not possible. Thus, keeping the track of
network with the increase in number of dead nodes became
a tedious task. With the static topology of EECML such
concerns were effectively addressed.
But the basic assumption that the network will be made of
nodes uniformly scattered all over the area is a pitfall, as it is
possible the geographical conditions may not permit so.
Thus, the network will be made of some denser and some
rarer areas deviating from the ideal conditions. In such
case merging clusters in way that the new clusters formed
deviate the least from ideal properties might prove useful.
Such adaptability according to varying density across over
network could prove useful maximizing the network lifetime.

II. LITERATURE SURVEY
The potential applications of wireless sensor networks
(WSNs) are highly varied, such as environmental monitoring,
target tracking and military. Sensors in such networks are
equipped with sensing, data processing and radio
transmission while the power is highly limited. This
necessitates devising novel energy-efficient solutions to
some of the conventional wireless networking problems,
such as medium access control, routing, self-organization,
bandwidth allocation, and security. Exploiting the trade-offs
among energy, accuracy, and latency, and using hierarchical
(tiered) architectures are important techniques for prolonging
the network lifetime. Energy consumption in the network
can be either useful or wasteful depending upon the data
being handled while the energy is spent.
A number of protocols aim at reducing useful energy
consumption. They are classified into three categories. The
protocols in the first category control the transmission power
level at each node by increasing the transmission power
level at each node by increasing network capacity thus
International Journal of Computer Science Trends and Technology (IJCST) Volume 2 Issue 1, Jan-Feb 2014
ISSN: 2347-8578 www.ijcstjournal.org Page 56
keeping the network connected (from HEED paper).
Protocols in the second category make routing decisions
based on power optimization goals. Protocols in the third
category control the network topology by determining which
nodes should participate in the network operation (be awake)
and which should not (remain asleep).
In LEACH, Energy-based LEACH and HEED protocols
use a single criterion to elect cluster heads and to form
clusters. LEACH protocol uses a round-robin rotation to
elect its cluster head where nodes elect to become the cluster
heads based on the threshold value. Each node chooses a
random number between 0 and 1 and if the number is lower
than the calculated probability, the node is elected to become
cluster head. Nodes that have not become a cluster head in a
specific round will have higher probability to become the
cluster heads on the next round. Energy-based LEACH
elects nodes to become a cluster head at time t with
probability:


Where Ei is residual energy of node I,
and k is the optimal number of cluster head. Nodes with
higher energy have higher probability of becoming the
cluster heads regardless whether the cluster heads are within
each others range. Energy-based LEACH protocol
demonstrates the same behaviour as LEACH but elects
cluster head based on node residual energy with respect to
the total energy of the network. Energy-based LEACH
shows better cluster heads selection and therefore has
longer network lifetime. Once cluster heads announces
their status, non-cluster head nodes will join a cluster based
on only one criterion which is the strongest received signal
strength. LEACH is a levelled hierarchical routing protocol
which attempts to minimize global energy dissipation and
distribute energy consumption evenly across all nodes. This
is achieved by the formation of clusters with localized
coordination, by rotating the high-energy cluster heads and
by locally compressing data.

The model used makes the following
assumptions:
1. There exists one base station with no energy constraints
and a large number of sensor nodes that are mostly
stationary, homogeneous and energy constrained.
2. The base station is located at some distance from the
sensor nodes and the communication between a sensor node
and the base station is expensive.
3. The purpose of the network is to collect data through
sensing at a fixed rate (i.e. there is always something to
send) and convey it to the base station.
The raw data is too much and must be locally aggregated
into a small set of meaningful information. The nodes self-
organize into local clusters with one node in each cluster
acting as a cluster head. Once a cluster has formed, the
cluster members send their data to the cluster head (low
energy transmission) which in turn combines the data
and sends it to the base station (high energy
transmission). This organization of the nodes creates a 2-
level hierarchy. The operation of the protocol is broken up
into rounds, during which the clusters are dissolved and
recreated. During each round, a node decides
probabilistically whether to become a cluster head. This
decision is based on the suggested percentage of cluster
heads for the network (determined a priori) and the number
of times the node has been a cluster head so far. The
cluster heads advertise their intention and the rest of the
nodes decide which cluster to join, usually based on
signal strength. Once the clusters are formed, the cluster
head creates a TDMA schedule and sends it to its cluster
members.
To reduce interference, each cluster communicates using
different CDMA codes. For their analysis, comparing their
scheme with a direct communication protocol (each sensor
sends data directly to the base station) and the minimum-
energy routing protocol. In the latter, data destined for the
base station is routed through many intermediate nodes that
can each be reached with minimum energy transmission. A
static clustering scheme is also used where cluster heads are
not rotated. Their results indicate that LEACH reduces
communication energy by as much as 8x. Also, the first
node death in LEACH occurs over 8 times later and the last
node dies over 3 times later.
HEED protocol elects the cluster heads based on node
residual energy which is defined as:

Where Residual is the energy of the nodes, Emax is the
initial energy of each node and Cprob is the initial
percentage of cluster heads that is set as 5%. A number of
iterations are performed before a final cluster heads are
elected. HEED protocol ensures uniform distribution of
cluster heads across the network. In its initialization phase,
HEED protocol allows sensors to compute a probability of
becoming cluster heads, proportional to its residual energy
and to a pre-determined percentage of cluster heads. Then,
during a repetition phase, sensor seeks the best cluster head
to connect to. If no cluster heads is found, the sensor doubles
its probability to become cluster head and broadcasts it
again to its neighbours.
This phase stops either when this probability equals to 1 or
when it finds a cluster head to connect to.EECML as
described earlier divides the entire network into static clusters
and then maintains this topology throughout the entire time.
For this purpose it calculates various optimal parameters
such as: optimal clustering angle, the angle at the network
topology should be divided to form static sectors, Optimal
one hop distance, optimal distance for
communication between the cluster heads of the individual
clusters within the sectors. The parameter also helps in
dividing the sectors into clusters situated optimal distance
apart. Then we have the threshold energy at which the
cluster head the given cluster head should stop working and
International Journal of Computer Science Trends and Technology (IJCST) Volume 2 Issue 1, Jan-Feb 2014
ISSN: 2347-8578 www.ijcstjournal.org Page 57
new cluster head is required to be elected. Thus using these
well-defined values the network is put to a start.

III. SYSTEM ARCHITECTURE AND
DESIGN ISSUES
Depending on the application, different architectures and
design goals/constraints have been considered for sensor
networks. Since the performance of a routing protocol is
closely related to the architectural model, in this section we
strive to capture architectural issues and highlight their
implications.
A. Network Dynamics
There are three main components in a sensor network.
These are the sensor nodes, sink and monitored events.
Aside from the very few setups that utilize mobile sensors
most of the network architectures assume that sensor nodes
are stationary. On the other hand, supporting the mobility of
sinks or cluster-heads (gateways) is sometimes deemed
necessary .Routing messages from or to moving nodes is
more challenging since route stability becomes an important
optimization factor, in addition to energy, bandwidth etc.
The sensed event can be either dynamic or static depending
on the application .For instance, in a target election/tracking
application, the event (phenomenon) is dynamic whereas
forest monitoring for early fire prevention is an example of
static events. Monitoring static events allows the network to
work in a reactive mode, simply generating traffic when
reporting. Dynamic events in most applications
require periodic reporting and consequently generate
significant traffic to be routed to the sink.

B. Node Deployment
Another consideration is the topological deployment of
nodes. This is application dependent and affects the
performance of the routing protocol. The deployment is
either deterministic or self-organizing. In deterministic
situations, the sensors are manually placed and data is
routed through pre-determined paths. However in self-
organizing systems, the sensor nodes are scattered randomly
creating an infrastructure in an ad hoc manner. In that
infrastructure, the position of the sink or the cluster-head is
also crucial in terms of energy efficiency and performance.
When the distribution of nodes is not uniform, optimal
clustering becomes a pressing issue to enable energy efficient
network operation.

C. Energy Considerations
During the creation of an infrastructure, the process of
setting up the routes is greatly influenced by energy
considerations. Since the transmission power of a wireless
radio is proportional to distance squared or even higher order
in the presence of obstacles, multi-hop routing will consume
less energy than direct communication. However, multi-
hop routing introduces significant overhead for
topology
management and medium access control. Direct routing
would perform well enough if all the nodes were very close
to the sink. Most of the time sensors are scattered randomly
over an area of interest and multi-hop routing
becomes unavoidable.

D. Data Delivery Models
Depending on the application of the sensor network, the
data delivery model to the sink can be continuous, event-
driven, query-driven and hybrid .In the continuous
delivery model, each sensor sends data periodically. In
event-driven and query-driven models, the transmission of
data is triggered when an event occurs or a query is
generated by the sink. Some networks apply a hybrid model
using a combination of continuous, event-driven and query-
driven data delivery. The routing protocol is highly
influenced by the data delivery model, especially with
regard to the minimization of energy consumption and route
stability. For instance, it has been concluded in that for a
habitat monitoring application where data is
continuously transmitted to the sink, a hierarchical
routing protocol is the most efficient alternative. This is
due to the fact that such an application generates
significant redundant data that can be aggregated on route to
the sink, thus reducing traffic and saving energy.

E. Node Capabilities
In a sensor network, different functionalities can be
associated with the sensor nodes. In earlier works and all
sensor nodes are assumed to be homogenous, having equal
capacity in terms of computation, communication and power.
However, depending on the application a node can be
dedicated to a particular special function such as relaying,
sensing and aggregation since engaging the three
functionalities at the same time on a node might quickly
drain the energy of that node. Some of the hierarchical
protocols proposed in the literature designate a cluster-head
different from the normal sensors. While some networks
have picked cluster-heads from the deployed sensors
other applications a cluster-head is more powerful than the
sensor nodes in terms of energy, bandwidth and memory .In
such cases, the burden of transmission to the sink and
aggregation is handled by the cluster-head.
Inclusion of heterogeneous set of sensors raises multiple
technical issues related to data routing .For instance,
some applications might require a diverse mixture of sensors
for monitoring temperature, pressure and humidity of the
surrounding environment, detecting motion via
acoustic signatures and capturing the image or video
tracking of moving objects. These special sensors either
deployed independently or the functionality can be
included on the normal sensors to be used on demand.
Reading generated from these sensors can be at different
rates, subject to diverse quality of service constraints
International Journal of Computer Science Trends and Technology (IJCST) Volume 2 Issue 1, Jan-Feb 2014
ISSN: 2347-8578 www.ijcstjournal.org Page 58
and following multiple data delivery models, as
explained earlier. Therefore, such a heterogeneous
environment makes data routing more challenging.

F. Data Aggregation/Fusion
Since sensor nodes might generate significant redundant
data, similar packets from multiple nodes can be aggregated
so that the number of transmissions would be reduced.
Data aggregation is the combination of data from different
sources by using functions such as suppression
(eliminating duplicates), min, max and average. Some
of these functions can be performed either partially or fully
in each sensor node, by allowing sensor nodes to conduct in-
network data reduction .Recognizing that computation would
be less energy consuming than communication substantial
energy savings can be obtained through data aggregation.
This technique has been used to achieve energy efficiency
and traffic optimization in a number of routing protocols. In
some network architectures, all aggregation functions are
assigned to more powerful and specialized nodes. Data
aggregation is also feasible through signal processing
techniques. In that case, it is referred as data fusion where a
node is capable of producing a more accurate signal by
reducing the noise and using some techniques such as beam
forming to combine the signals.
IV. THE PROBLEM OUTLINE
In a hierarchical protocol communication between an end
node and base station takes place via a well-defined path of
communicating nodes, which for the major portion is made
of cluster head nodes. Cluster head nodes are no different in
capabilities than any other ordinary node but they do heavy
work of collecting the data from all the nodes and
communication to the higher level base station in hierarchy
thus in most of the protocols role of base station is rotating.
As discussed previously one of efficient algorithms for
WSN communications is HEED (Hybrid Energy Efficient
Distributed Clustering) which uses the dynamic approach
for cluster formation. The dynamic clustering algorithm
benefits HEED by electing cluster head based on its residual
energy as well as on the factor of centrality of cluster head
node in the cluster. But then it suffers from cost of
communication in updating the cluster head information to
each and every node after some period regularly. Moreover,
the network topology doesnt remain stable as nodes are
added and removed from the clusters as the cluster head
changes
So considering a more stable algorithm in the sense of its
topology, EECML (Energy Efficient Clustering for
Maximizing Network Lifetime) uses clusters which
are formed with calculations of several optimal
parameters. EECML does not change its cluster formation
throughout lifetime but it rotates role of cluster head
between nodes within the cluster itself. But the EECML
protocol has major fallacy that it is designed only for
uniform density wireless networks. Wireless sensor
networks have been proving useful in very remote areas of
irregular geographical conditions ranging from hilly terrains
to deep seas. So having the sensors equally distributed is
quite of a job. Also to have such a uniform density is quite
more than expected and sometimes even impossible. But all
optimizing parameters of EECML are calculated
assuming uniform density so the resultant cluster formation
is suitable only for uniform density thus the efficiency
of protocol drops significantly in non-uniform dense
network.
V. PROPOSED APPROACH
The Network lifetime for a Wireless sensor Network is
defined as the time elapsed from the point when the nodes
are being deployed to the time when any sensor node in
the Network dies. So it can be said that time at which the
first node in the network dies, that corner of the
Network is inaccessible thus making the Network unreliable.
As said about previously majority of workload is
trusted upon the Cluster Head (CH) nodes. In addition to the
work of ordinary node, the CH gathers data from its
colleague nodes in cluster, it collects data from the CH of
the cluster neighbouring to it but away from base station, it
communicates this gathered data to the next CH in the
hierarchy. Since there is a lot of redundancy involved,
new protocols perform various aggregation techniques so
as to reduce the energy spent in transmission as well as
receipt. This can be explained with the power model for
network
Energy consumption in a sensor node can
be attributed to either useful or wasteful sources.
Useful energy consumption can be due to
1. transmitting/receiving data,
2. Processing query requests, and
3. Forwarding queries/data to neighbouring nodes.

Wasteful energy consumption can be due to
1. Idle listening to the media,
2. Retransmitting due to packet collisions,
3. Overhearing, and
4. generating/handling control packets.
EECML and HEED deal with reducing the useful
energy consumption. In WSN, the main energy consumption
of the active node is made up of three parts: message
sending,
message receiving and data processing [3, 8]. The simplified
energy consumption model for each part can be defined as:
PT (k) = Eelec k + Eamp d
r

k
PR(k) = Eelec
k
Pcpu(k) = Ecpu
k

Where k is the length (bits) of packets, d is the transmission
International Journal of Computer Science Trends and Technology (IJCST) Volume 2 Issue 1, Jan-Feb 2014
ISSN: 2347-8578 www.ijcstjournal.org Page 59
distance (m). The radio dissipates Eelec (nJ/bit) per bit to run
the radio circuitry. Eamp (nJ/bit/m2) is the power above
Eelec needed by the transmitter for an acceptable Eb/N0
at the receivers demodulator. Ecpu (nJ/bit) is the
energy dissipation for processing per bit and r is the
path loss exponent that is related to the transmission
distance.
As it can be seen from the above equations each of the
energy calculated is directly proportional to the number of
bits being handled, i.e., k. Thus, any strategy that helps
reduce k, keeping intact the meaning of the data, would
suffice the purpose. Data aggregation techniques employ
methodologies that communicate the same data with almost
negligible loss of mnemonics but in a compressed way.
They achieve it mostly by redundancy in the
communications. So now CH will be entrusted with this
additional job. The aggregation techniques will not prove
useful if the cluster doesnt have enough nodes. The CH will
just then dissipate more quickly.
Owing to these problems of network non-uniformity, which
might even render the aggregation techniques useless,
reorganization of sectors originally formed by EECML
protocol so as they can be made to work ideally as far as
possible is necessitated. The approach to be presented aims
at merging the clusters or if possible even dividing the
clusters so as to balance the load on the object of maximum
load. Finding out how much does a cluster deviate from the
ideal characteristics proves a lot useful. EECML protocol
defines an optimal clustering angle for dividing the network
into similar sectors. The property of this clustering angle is
that is independent of the area of the network or the
radius of network. It depends upon the various energy
attributes of the sensor nodes and the number of nodes
deployed. As the energy parameters of the nodes cannot be
changed once deployed they serve as constants. Thus, the
clustering angle is dependent upon the number of nodes,
more accurately the clustering angle varies inversely with
the number of nodes in the area.

Thus, finding out the ideal number of nodes for a given
clustering angle and comparing it with the actual number of
the nodes present in the area we can separate out
dense clusters from rarer clusters. A round for a cluster
may be defined as when all the nodes from the cluster
communicate for once their data with the cluster head. As
we have cluster head re-election logic for maximizing
lifetime after some threshold value we choose another
cluster head from the same cluster. If the cluster is rarer a
lot of rounds will be taking place and the ordinary nodes
will deplete far sooner than the cluster head reaches its
threshold limit. Thus, no re-election takes place and the
cluster dies. So by merging clusters in a way that the value
of number of nodes is pushed closer to the ideal value but
not beyond it, we can utilize the same nodes over larger area
for a longer time.



VI. IMPLEMENTATION
The implementation of protocol is divided into phases as
follows:
A. Initialization
This phase is concerned with calculating the various
optimal parameters for the Network as laid down by
EECML. The parameters are as follows
The optimal one hop distance:
Nodes are spread over the network each one
charged with same power and having same characteristics
of which one hop distance is one. One hop distance is the
distance over which a sensor node can communicate.
That is one hop distance is the range of the node. Now for
the communication to be efficient the nodes should be
situated some optimal distance apart. It is not possible for
all nodes to be situated optimal distance apart. But we can
use this parameter for finding the cluster heads. The cluster
heads can be chosen to be optimal one hop distance apart, as
they are the major source of energy dissipation. The
optimal one hop distance dopt is given by:





The optimal clustering angle:
The clustering angle is a very important parameter,
which directly affects the number of cluster heads. The
optimal clustering angle is calculated taking into
consideration the energy parameters of the nodes and the
number of nodes in the network. It is given by:






For determining the optimal time as to when a cluster
head should be replaced with another node so that cluster
might not die abruptly and suddenly we require some
threshold value. Up until this value a CH is allowed to be
working. Once the limit is crossed the it is replaced with
another probable candidate from the cluster. Its value is
given by:




Base station does have all these values known in advance
and hence calculates these values in the initialization phase.
Then it broadcasts these values over the network. The
International Journal of Computer Science Trends and Technology (IJCST) Volume 2 Issue 1, Jan-Feb 2014
ISSN: 2347-8578 www.ijcstjournal.org Page 60
network is divided into sectors and further into clusters. First
of all the base station divides network with the help of
optimal clustering angle into sectors. Then it finds out nodes
optimal one hop distance away from itself in these sectors.
These found are the first cluster heads for that sector. Then
these cluster heads find out the next cluster heads optimal
one hop distance apart. The process continues till all the
clusters are formed. Now to find out clusters we have one
hop distance to our help. The sectors are divided into clusters
each of length one hop distance. This marks the end of
initialization step as network Is divided into clusters and
sensors into ordinary nodes and cluster heads.

B. Differentiation
Next phase is the finding out the density of a particular
sector and comparing it with the optimal value. The term
density here refers to number of nodes per sector. The
ideal implementation of EECML assumes uniform
network and thus uniform density for sectors to. The ideal
number of nodes can be found out as:


Thus if number of nodes in a sector are lesser than the Nideal
value then its rarer, if it is greater than Nideal then its denser.

C. Merging
After finding the density of each sector, it is required to
find out the sectors which can be merged and then does their
merging actually help out. The criteria for merging
sectors are:

1. They should be neighbours.
2. Their merging should not lead to a density which is
Greater than the ideal value.
Also the pairs chosen for merging should be either rare and
rare or denser and rarer but not denser-denser. After merging
the ideal density of newly formed sector can be found from
the combined clustering angle. The further attributes of the
sector are also found in similar way.
D. Clustering
After finding out the probable candidates for merging and
re-organizing the network as per the newly formed sectors
the actual working of network requires to be started. Since
all the clusters in the network have similar structure, that is
they have one cluster head and other ordinary nodes, they
can work independent of each other. That is individual
clusters can work independently. The nodes sense data and
communicate it to cluster head. As cluster head serves as
single point of collision the use of contention free technique
for communication is required. We use a round structure,
similar to TDMA, wherein a round is composed of
communication by each single node one after other in order.
A node can communicate only once in a round. The cluster
head then aggregates this data and sends it the next cluster
head in the sector hierarchy. For the communication between
the cluster heads the entire network can be considered as a
separate cluster made of base station as head node and
cluster heads of various sectors as its constituent nodes.
Now within a cluster re-election takes place by round
mechanism. A node is allowed to work as a cluster head
until it reaches the threshold value as discussed above.
Once it is reached another node from the cluster with
sufficient energy level acts as cluster head. And all other
cluster heads in the clusters of the same sector are also
updated.

VII. CONCLUSION

Now as defined previously, network lifetime is the time
when the first node in the network dies, whenever any
cluster form any sector in the network reports death of a
node we can say that network has reached its network
lifetime limit.

REFERENCES
[1] Osama Younis and Sonia Fahmy HEED: A Hybrid, Energy-E_cient,
Distributed Clustering Approach for Ad Hoc Sensor Networks
,IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL 3. NO.
4, OCTOBER-DECEMBER 2004.
[2] Heinzelman , W.R. MIT, Cambridge, MA, USA Chandrakasan, A.;
Balakrishnan, Hybrid Energy-efficient communication protocol
for wireless microsensor networksSystem Sciences, 2000. Proceedings of
the33rd Annual Hawaii International Conference on, 4-7 Jan. 2000.
[3] Xiang Mina, ShiWei-rena, Jiang Chang-jianga, Zhang Yinga
Energy efficient clustering algorithm for maximizing lifetime of
wireless sensor networks., Int. j. Electron. COMMUN. (AEU),64,(2010)
[4] Kemal Akkaya, Mohamed Younis A survey on routing protocols
for wireless sensor networks. Ad Hoc Networks xxx (2004) xxx-
xxx, 1 september 2003
[5] Sun Limin, Li Jianzhong, Chen Yu, Wireless Sensor
Networks, Tsinghua publishing company, Beijing, 2005.
[7] SHA Chao1, WANG Ru-chuan, HUANG Hai-ping1, SUN Li-
juan1, Energy efficient Clustering Algorithm for Data Aggregation
Energy for wireless sensor Networks, Procedia Engineering 29 (2012)
1882 1888
[8] Jau-Yang Chang and Pel-Hao Ju, An Efficient Cluster-based power
saving scheme for wireless sensor Networks,Chang and Ju
EURASIP Journal on Wireless Communications and Networking 2012,
2012:172
[9] LI Li, DONG Shu-song, WEN Xiang-ming, An energy efficient Cluster
routing algorithm for wireless sensor Networks,THE JOURNAL
OF CHINA UNIVERSITIESOF POSTS AND
TELECOMMUNICATIONS Volume 13, Issue 3, September 2006
[10] Pratyap Kuila, Prasanta K. Jana, Energy efficient Load-Balanced
Clustering algorithm for wireless sensor Networks, 2nd
International Conference on Communication, Computing and Security
[ICCCS-2012].
[11] Nauman Aslam, William Phillips, William Robertson, Shyamala
Sivakumar, A multi - criterion optimization
technique for energy efficient cluster formation in wireless
sensor networks, Information Fusion 12(2011)202-212
[12] Young Sang Yun and Ye Xia, A Method for Deciding Node Density in
Non-Uniform Deployment of Wireless Sensor Networks.
[13] Archana Bharathidasan, vijay Anand sai ponduru, Sensor Networks: An
overview
[14] Hu Yu, Li Wei, Kang Zhenhua, Study on Energy E_cient
Hierarchical Routing protocols of wireless sensor Network. 2009 WASE
International Conference on Information Engineering.
International Journal of Computer Science Trends and Technology (IJCST) Volume 2 Issue 1, Jan-Feb 2014
ISSN: 2347-8578 www.ijcstjournal.org Page 61
[15] Deepali Virmani, Stair Jain, Clustering Based Topology Control
Protocol for Data Delivery in Wireless Sensor Networks
[16] Mao Ye, Chengfa Li, Guihai Chen and Jie WuEECS: An Energy
Efficient ClusteringScheme in Wireless Sensor Networks s