Copyright © 2013 by American Scientific Publishers Advanced Science Focus

All rights reserved. Vol. 1, pp. 111–119, 2013

Printed in the United States of America (www.aspbs.com/asfo)

Maximum Residual Energy Based Clustering

Scheme for Wireless Sensor Networks
Puneet Azad1, 2, ∗ and Vidushi Sharma2
1
Department of Electronics and Communication, Maharaja Surajmal Institute of Technology,
Janak Puri, New Delhi 110058, India
2
School of Information and Communication Technology, Gautam Buddha University,
Greater Noida, Gautam Budh Nagar 201308, Uttar Pradesh, India

One of the major challenges in wireless sensor networks (WSNs) is uniform energy dissipation and lifetime
of the network. The nodes with higher energy dissipation lead to early death leaving a percentage of area
unattended for the entire lifetime of the network. This result in a shorter stability region as compared to other
protocols designed for extending the lifetime of the network. This paper deals with a new scheme termed the
maximum residual based clustering scheme (MREC) in which the entire network is re-clustered in each cycle
of data transmission. With automatic rotation of cluster head having maximum residual energy, MREC achieves
uniform energy dissipation through the whole network. The algorithm works in static and dynamic mode. In

ARTICLE
former, the cluster remains fixed for entire lifetime and the cluster heads are rotated within the cluster, while
in later, re-clustering is done in each cycle with cluster head having the highest energy. The performance of
the proposed method is compared with low-energy adaptive clustering hierarchy (LEACH) in homogeneous
environment and with energy efficient heterogeneous clustered scheme (EEHC) and distributed energy-efficient
clustering algorithm (DEEC) inDelivered by Publishing
heterogeneous Technology
environment. to: Guest results
Our simulation User demonstrate that MREC is
IP: 180.151.225.6 On: Tue, 18 Aug 2015 07:50:54
more effective for prolonging the network lifetime,
Copyright: enhancing
American stability
Scientific and has better reliability with estimated
Publishers
confidence bounds of network lifetime.

KEYWORDS: MREC, Wireless Sensor Networks, Clustering, Lifetime, Heterogeneous, Optimum.

1. INTRODUCTION for most of the lifetime of the network. In this context,

the clustering technique can be used for designing energy
Wireless Sensor Networks (WSNs) consist of sensor hard- efficient protocols for enhancing the lifetime of entire
ware units called sensor nodes, capable of sensing, data network.6 A few clustering algorithms have been studied
processing and transmitting the information.1 They have in the area of wireless sensor networks.7 8 Sensor nodes
a wide range of applications in monitoring environmen- are grouped into clusters based on their location, resid-
tal conditions,2 surveillance,3 audio and video retrieval,4 ual energy, signal strength or connectivity etc. Clustering
healthcare,5 forest fire detection, flood detection etc. enables bandwidth reuse, saves energy, increases system
WSNs process and transmit data efficiently using an inbuilt capacity, better resource allocation and reduces overhead
processor and consist of four components: sensor, proces- communication. Each cluster has a CH which is responsi-
sor, transceiver and a power unit.1 Additional components ble for gathering the data from all the nodes in its cluster
such as a GPS unit, power generator and mobiliser may and subsequently transfers it to the base-station. A vari-
be added to enhance their performance. The major issue ety of promising protocols such as LEACH,9 HEED,10
in WSN technology is the efficient management of energy DHAC11 and ADRP12 have been reported in literature.
between randomly deployed sensor nodes so that their util- These protocols operate in a homogeneous environment
ity function can be maximised. The aim is to develop where all the nodes are initially assigned equal energy.
an energy efficient protocol that avoids the gradual death Whereas protocols like EEHC,13 DEEC,14 SEP15 are suit-
of nodes, thereby creating a uniform delivery of packets able for a heterogeneous environment where a fixed per-
centage of nodes have more energy then others; however,
∗
further work is required in order to improve the efficiency
Author to whom correspondence should be addressed.
Email: [email protected] of WSNs for better lifetime. In this paper, a new algorithm
Received: 27 September 2012 based on the residual energy of the nodes is developed,
Accepted: 1 December 2012 which initially uses fuzzy c-means clustering16 17 and then

Adv. Sci. Focus, Vol. 1, No. 2, 2013 2330-0760/2013/1/111/009 doi:10.1166/asfo.2013.1021 111

 Maximum Residual Energy Based Clustering Scheme for Wireless Sensor Networks Azad and Sharma

develops a new approach for the selection of cluster heads Clusters were formed using various hierarchical cluster-
(CHs). ing methods which include SLINK, CLINK, UPGMA, and
The outline of this work is as follows. Section 2 dis- WPGAM.6 The simulation of homogeneous sensor nodes
cusses the related literature; Section 3 discusses the pro- reveals the improved lifetime of the network as compared
posed methodology in detail; Section 4 demonstrates the to LEACH and LEACH-C. Another important protocol is
simulations and performance results; and finally conclu- Adaptive Decentralised Re-Clustering Protocol (ADRP)12
sions are summarised in Section 5. in which the cluster heads and next heads are elected on
the basis of the residual energy of each node and the aver-
age energy of each cluster. The selection of cluster heads
2. BACKGROUND AND RELATED WORK and next heads are weighted by the remaining energy
2.1. Homogeneous Environment Protocols of sensor nodes and the average energy of each cluster.
The sensor nodes with the highest energy in the clusters
LEACH9 is the most popular distributed cluster-based can be cluster heads at different cycles of time. Thus the
routing protocol in wireless sensor networks for a homo- role of cluster heads can be switched dynamically. Zhu
geneous environment. In this paper, a central control algo- et al.18 presented a new clustering protocol based on haus-
rithm is designed to form the clusters by dispersing the dorff distance and minimum energy routing for WSNs to
cluster head (CH) nodes throughout the network. Each maximise the lifetime. Clusters once formed are based
node chooses to become cluster head based on a prede- on node locations, communication efficiency and network
fined threshold given whereby connectivity and the role of the cluster head is optimally
⎧ popt rotated among the cluster members. After cluster heads
⎨ are selected, they form a network to periodically collect,
T S = 1 − popt ∗ c ∗ mod 1/popt  if S ∈ G (1)
ARTICLE

⎩0 aggregate, and forward data to the base station using min-

imum energy (cost) routing. This method can significantly
popt is the percentage of cluster heads, c is the current increase the network lifetime as compared to other known
cycle and G is the set of nodes that have not been clus- methods. Since the clusters are static, the selection of clus-
Delivered
ter heads in the last 1/popt cycles. by head
The cluster Publishing ter heads is
nodes Technology to:limited
Guest toUsernodes present in that cluster, which
fuse and aggregate data arrivingIP: 180.151.225.6
from On: Tue, 18 Aug 2015 07:50:54 have maximum residual energy.
nodes from every may not choose nodes that
Copyright: American Scientific
Min et al.Publishers
19
presented an energy efficient clustering algo-
cluster and send the aggregated data to the base station
(sink) in order to reduce the amount of data and avoid rithm for maximising lifetime of wireless sensor networks
duplicate entry. Data collection is centralised to sink and with optimum parameters for reducing energy consump-
perform periodically. In LEACH, the number of clusters tion and prolonging the system lifetime. All nodes are
chosen in an optimum manner is found to be about 5 per- divided into static clusters with optimum parameters with
cent of the total number of nodes. The nodes select a smaller clusters close to the base station and larger clusters
cluster head based on the received signal strength of the away from it. Thus the cluster head near the base station
advertisement message and send data in their time slot in should be kept alive avoiding the energy hole problem in
each round. Leach-C an improvement over Leach, finds multi-hop communication. The disadvantage of such an
clusters using the simulated annealing algorithm and find algorithm is the incomplete coverage due to consideration
k optimum clusters with minimum average energy dissi- of sector shape clusters.
pation per round. Whereas the hybrid energy-efficient dis-
tributed (HEED)10 protocol is an energy-aware hierarchical 2.2. Heterogeneous Environment Protocols
approach that is an improvement on LEACH. This proto-
col uses two radio transmission power levels for intra-and EEHC13 adopt the heterogeneity of the nodes in terms of
inter-cluster communication. Cluster heads are probabilis- their initial energy i.e., a percentage of nodes are equipped
tically selected based on their residual energy. Another with more energy than others. The nodes play the role
parameter called neighbour proximity or node degree is of a cluster head based on weighted election probabilities
used to find the best cluster head for a normal node to according to the residual energy given by
join. If a CH is far from the sink, it tries to send the popt
aggregate data to another CH instead of sending to the pn = (2)
1 + m ∗  + mo 
sink directly. Zhou et al.11 reported taking an entirely dif-
ferent approach called distributed hierarchical agglomera- where popt is the predetermined percentage of the cluster
tive clustering (DHAC) algorithm in which the clusters are head, m is the fraction of the total number of nodes, mo is
formed before the selection of cluster head unlike in other the percentage of m having  times more energy than oth-
protocols. The clusters are formed on the basis of quanti- ers. The probability of the selection of a node (as a cluster
tative data such as location of nodes and received signal head) is more when the initial energy of the node is higher.
strength as well as qualitative data such as connectivity. It is reported that the lifetime of the network increased

112 Adv. Sci. Focus, 1, 111–119, 2013

Azad and Sharma Maximum Residual Energy Based Clustering Scheme for Wireless Sensor Networks

by 10 percent as compared with the results obtained by clusters remain static throughout the lifetime of the net-
using the LEACH method. However, Zhou et al.20 have work, but the role of CH is rotated among other nodes
further analysed computational heterogeneity in which few of the cluster. This strategy is further modified (explained
nodes with powerful processing capability, more energy in the next section), where clusters are re-formed in each
and storage can act as cluster heads for data aggregation cycle of data transmission unlike in s-MREC. This re-
and transmission in heterogeneous WSNs. There are three clustering approach creates a new method called MREC
node types; 0 type, 1 type and few management nodes. (Maximum Residual Energy based Clustering) wherein an
The cluster heads are selected based on energy dissipation optimum number of cluster heads are selected in each
forecast and clustering management (EDFCM), in which cycle of data transmission based on the maximum residual
the energy consumption of the cluster heads in last round energy in the whole network. Based on the selected CH,
is used as the forecast value for the next round. The cluster the nodes join a particular cluster based on the nearest dis-
head selection algorithm is based on the method of energy tance (Euclidean distance). The performance of these two
dissipation forecast and clustering management (EDFCM). techniques (s-MREC and MREC) is compared with that
DEEC14 is an energy efficient clustering protocol in which of the results obtained form other reported protocols. We
the cluster-heads are selected by a probability based on the study these two techniques (s-MREC and MREC) with the
ratio between residual energy of each node and the aver- following presumptions.
age energy of the network. The nodes with high initial and 1. Nodes are placed randomly in a 100 × 100 region in
residual energy will have more chances to be the cluster- uniform distribution. They are location aware and can get
heads than the nodes with low energy. Another method information through GPS. The role of nodes is to sense
considers strategic deployment2 for selecting the cluster the environment and send the data to their respective clus-
head. The clusters are formed in the form of multiple- ter head.

ARTICLE
sized fixed grids while taking into account the arbitrary- 2. The base station is also static and is located in the
shaped area sensed by the sensor nodes. Since deployment centre of the field with no energy constraints.
is strategic, no separate algorithm is designed for the CH 3. The initial number of clusters is fixed by taking opti-
selection. This kind of scheme is only suitable for civil mum value (Section 4.2) and keeps on varying with the
purpose and cannot be applied Delivered
for militarybyorPublishing
vigilance Technology to: Guest User
node density.
purposes. IP: 180.151.225.6 On: Tue, 4. 18To
Aug 2015heterogeneity,
ensure 07:50:54 two type of nodes are defined
Copyright: American Scientific Publishers
with energies Eo and Eo (1+, where  is the percentage
3. PROPOSED METHODOLOGY increase in energy.
5. The nodes in a cluster can only communicate with
The present methodology is based on selecting the opti- their respective cluster head and not with the base station
mum number of cluster heads (described in Section 4.2), directly.
forming the clusters and then transmitting the data sensed 6. The transmission and processing capability of all the
from the environment. The base station divides the net- nodes is same.
work into clusters and selects optimum number of clus- The following messages are used in the formation of
ter heads. This is done initially by a well known fuzzy clusters during the set-up phase:
c-means (FCM) clustering technique, whereby centre of a. Hello Message: This message is sent by all the nodes to
each cluster become the cluster head. Further after few the base-station in the beginning. It contains the location
rounds, the cluster heads are selected based on their resid- information obtained through GPS.
ual energy and clusters are formed based on the shortest b. Broadcast Message: Each elected CH broadcasts its
distance between CHs and nodes. This algorithm consists node-id to all the nodes present in the network in such a
of set-up and steady state phases. In the set-up phase, way that they should reach the farthest node in the net-
clusters are formed using FCM clustering for few initial work.
cycles and then, a new method is adopted for the for- c. CH_Join Message: Each non-CH node then measures
mation of clusters based on nodes residual energy. In the the distance with the available list of elected CHs and
steady state, CHs are responsible for aggregating and send- chooses the one at minimum distance. Then a CH_Join
ing the data to the sink. Thus static clusters are formed message is sent from each node to the respective cluster
with the cluster head being located nearest to the centroid head to be joined.
of the cluster. After few cycles of data transmission (about
100 cycles), the clusters formation adopts a new method-
ology in which the position of CH is optimally sched- 3.1. Static Maximum Residual Energy Based
uled among the nodes present in the cluster based on the Clustering (s-MREC)
residual energy discarding the traditional FCM method.
This method is hereafter referred to as s-MREC (Static s-MREC consists of the following steps:
Maximum Residual Energy based Clustering) in which the Step 1: Obtain an input data set

Adv. Sci. Focus, 1, 111–119, 2013 113

 Maximum Residual Energy Based Clustering Scheme for Wireless Sensor Networks Azad and Sharma

k+1 k
An input data set for s-MREC is a data matrix that These iterations stop, when maxij uij − uij  < ,
consists of nodes placed randomly in an area following where is a termination criterion between 0 and 1,
a uniform distribution method. Figure 1 shows a typical whereas k signifies the iteration steps. In the present study,
100 node network in a 100 × 100 m2 area. Each node sensor nodes are grouped into the ten clusters (Fig. 2(a)),
sends its location information to the base station through which is the optimum value in a scenario of 100 nodes as
a “Hello” message containing the node id and physical discussed later in Section 4.2. Each cluster contains a clus-
location. ter head (red square in Fig. 2(a)) nearest to the centroid
Step 2: Initialize the number of clusters and assign clus- of the cluster. Each selected CH broadcasts its node id in
ter heads its respective cluster in such a way that it should reach
The basic purpose of the s-MREC is to generate an opti- the farthest node in the cluster. The nodes then send the
mum number of clusters (described in Section 4.2) using join-request message to their CH.
the FCM technique. The cluster heads are chosen as the Step 3: Data transmission and CH re-assignment
node nearest to the centre as determined by this technique. After the cluster head is allocated and clusters are
In this method, each data point has a degree of belonging formed, all the nodes are assigned a Time Division
to clusters rather than belonging completely to just one Multiple Access (TDMA) schedule in each cycle of data
cluster. It was developed in 1973 by Dunn15 and improved transmission. There is no collision between neighbour-
in 1981 by Bezdek.16 This method obtains the best loca- ing clusters as they do not have a specified boundary
tion of clusters in an optimum manner by minimising an instead the nodes present in a cluster are connected to
objective function as given by their respective cluster head as shown in Figure 2(a). The
nodes continuously monitor the environment and send the

N 
C
data to their respective cluster head in the assigned slot-
J= ij xi − cj 
um 2
(3)
ARTICLE

i=1 j=1
subsequently the data can be transferred to the base sta-
tion from all the cluster heads. After few cycles of data
where xi is the set of data points, N is the number of data
set, cj is the center of the clusters, C is the number of clus- (a) 100
ters, uij is the degree of membership of xi by
Delivered in Publishing
the cluster Technology to: Guest User
90
j,  ∗  is any norm expressing the IP:similarity
180.151.225.6
betweenOn:
anyTue, 18 Aug 2015 07:50:54
measured data and the centre (which is Copyright: American
the distance here) Scientific
80 Publishers

and m is the weighing exponent on each fuzzy member- 70

ship. Fuzzy partitioning is carried out through an iterative 60

Base Station
optimisation of the objective function shown above, with 50
Y

the update of membership uij and the cluster centres cj by 40

30
1
uij = C  20
k=1 x i − c j xi − ck 
2/m−1

N m 10

i=1 uij · xi 0
where cj = N m (4) 0 10 20 30 40 50 60 70 80 90 100
i=1 uij X

(b) 100
100
90
90 80
80 70
70 60
60 Base Station
50
Y
Y

50 40
40 30
30 20
20 10
10 0
0 10 20 30 40 50 60 70 80 90 100
0
0 10 20 30 40 50 60 70 80 90 100 X
X
Fig. 2. Cluster formation with (a) CH nearest to the centre using FCM,
Fig. 1. A 100-node network. (b) Maximum residual energy node (in the same cluster) as a CH.

114 Adv. Sci. Focus, 1, 111–119, 2013

Azad and Sharma Maximum Residual Energy Based Clustering Scheme for Wireless Sensor Networks

transmission, it is found that cluster heads deplete their selected cluster head using a distance matrix, DM (m × n)
energy drastically and lose their efficiency to act as a clus- given as follows;
ter head. In this situation, new cluster heads are selected ⎡ ⎤
based on maximum residual energy among all the nodes dCH1 x1 dCH1 x2 dCH 1 xn
⎢ dCH2 x1 dCH2 x2 dCH2xn ⎥
in respective clusters (Fig. 2(b)). It is to be noted that ini- DM = ⎢ ⎣
⎥
⎦ (5)
tially the number of clusters remains same during the data
transmission. As the energy of nodes become less than a dCHm x1 dCHm x2 dCHm x3 dCHm xn
threshold value required for sending the data packet, they
where d is the Euclidean distance between CH and a node
are declared dead and no longer take part in the data trans-
based on its location information. If y and z represent the
mission. If any cluster contains fewer nodes than defined
location of two nodes p and q, then the Euclidean dis-
by a threshold value, then it merges into the neighbour-
tance is
ing cluster. Also the two CHs in different clusters may be
very close to each other and the nodes present in the cor-
dpq = px − qx 2 + py − qy 2 1/2
(6)
responding clusters have to choose a farther distance for
the transmission resulting in higher energy expenditure. Each element di j in the distance matrix represents the
This drawback is avoided in MREC (next section) as the distance between the ith clusterhead and j node. The col-
clusters are re-formed each having selected CHs having umn containing the minimum value represents the cluster
minimum Euclidean distance in every cycle. In this study, number to be joined by the corresponding node. For exam-
simulations are performed in both a homogeneous and a ple, if dCH2 x1 is the minimum value in the first column,
heterogeneous environment. in this situation the node x1 gets associate with the second
cluster where CH2 is cluster head.

ARTICLE
3.2. Maximum Residual Energy Based The operation of re-clustering and data transmission
Clustering (MREC) continues for many cycles (as discussed above) until the
death of first node. If the size of the cluster is smaller then
In this section, the above mentioned protocol (s-MREC) the predefined threshold, the cluster merges with the neigh-
is modified, however the methodology Deliveredforbycluster head Technology
Publishing to: Guest
bouring clusters shownUser
in the dotted lines in Figure 3. With
selection is same as used in s-MREC. IP: 180.151.225.6
The differenceOn:lies Tue,the
18 Aug 2015 07:50:54
start of the death of nodes, it is found that there are a
in the formation of clusters, which Copyright:
are reformed American
after Scientific Publishers
lesser number of nodes present in each cluster now. Thus
each cycle of data transmission. It is to be noted that as the number of alive nodes starts decreasing with cycles,
the MREC method is investigated in heterogeneous mode, the number of clusters also decreases and the decrease in
where fewer nodes are assigned more initial energy (in a the number of alive nodes eventually results in the reduc-
predefined ratio) than other nodes. The explicit methodol- tion in number of clusters as in Figure 3. The amount of
ogy of MREC is as follows: information also decreases with the fewer nodes left in
Step 1: Initial cluster formation and selection of cluster the physical area. The typical schematics of MREC and
head s-MREC are shown in flow chart (Fig. 4). The technique
The input dataset is obtained in the same manner as for data transmission is similar to that used in s-MREC
discussed in step 1 of Section 3.1. The methodology for technique.
electing the cluster head is same as followed in s-MREC
(using the FCM technique and designating the node near-
est to the centre of the cluster as cluster head). 100

Step 2: Re-selection of cluster head and data 90

transmission 80
This step mainly differentiates MREC from s-MREC in 70
which the clusters are formed dynamically after the selec- 60
tion of CH in each cycle. After receiving the data from Base Station
50
Y

the nodes for a few cycles, the node with the maximum
40
residual energy in its respective clusters is elected as a CH
at the end of each cycle. Subsequently, new clusters are 30

formed around all elected CHs using Euclidean distance. 20

Thus nodes lying in the vicinity of any CH form a new 10
cluster in each cycle. It is to be noted that the re-clustering 0
methodology is also adopted in LEACH protocol, where 0 10 20 30 40 50 60 70 80 90 100

CHs are elected by using the probabilistic approach rather X

then based maximum residual energy criterion (MREC). Fig. 3. Merging of small clusters (as shown in dotted lines) and reduc-
The nodes are re-clustered based on the distance with the tion in clusters with 24 nodes left.

Adv. Sci. Focus, 1, 111–119, 2013 115

 Maximum Residual Energy Based Clustering Scheme for Wireless Sensor Networks Azad and Sharma

Cycle, C =1

Initialize Generate Clusters Find number of

number of with center using nodes in cluster,
clusters FCM nodes_cluster

N Is
nodes_cluster
Find nodes with maximum >1
residual energy in all the
clusters for the next cycle Y
Find nearest
cluster and
merge
Broadcast CH ids to
all the nodes
Is there a node
coinciding with
the center?
Re-generate clusters by
building distance matrix N

Y
Find node
Invite nodes to join nearest to the
CH with min distance center
ARTICLE

to build new Clusters

Is nodes_cluster > 1 Assign it as CH for

Find node with the corresponding
max residual cluster
Y Delivered by Publishing Technology to:
energy in all
Guest User
IP: 180.151.225.6N
On: Tue, 18 Augclusters
existing 2015 07:50:54
Copyright: American Scientific Publishers
Is nodes_cluster
<10 Transmit data
using TDMA
Y
Find nearest
Split
cluster and MREC s-MREC
merge clusters
Start sensing data
N for next cycle

Reduce Incr C Y
Is C>100 ?
number of
clusters by 1 Incr C

Transmit data Start sensing data

using TDMA for next cycle

Fig. 4. Flowchart for s-MREC and MREC.

4. RESULTS AND DISCUSSION 2. Confidence bounds: It is an interval estimate of the life-

time of the nodes, which is used to indicate the reliability.
The simulated results in MATLAB of the proposed s- The network lifetime lies in the range of the confidence
MREC and MREC are compared with previously reported interval.
protocols, which include LEACH, DHAC, EEHC and 3. Stability region: It is the total time before the death of
DEEC. The protocols are executed 500 times for differ- the first node occurs.
ent deployment of nodes and mean results are used for 4. Number of packets received: It shows the number of
comparison. The performance of the proposed protocols is packets received by the base station from all the CHs.
measured in terms of following metrics:
4.1. Setup
1. Network lifetime: It shows the time interval from the
start of operation until the death of all the nodes present The WSN consists of 100 sensor nodes placed in a
in the network. 100 × 100 area. The simulation is conducted for both a

116 Adv. Sci. Focus, 1, 111–119, 2013

Azad and Sharma Maximum Residual Energy Based Clustering Scheme for Wireless Sensor Networks

homogeneous and a heterogeneous environment. A simple 75

Average energy dissipation per round (mJ)

radio energy dissipation model1 in transmitting a l bit mes- 70
sage over a distance d is presented in this section using
65
the simulation parameters of practical interest shown in
Table I. Free space and multi path fading model21 are used 60

for the data transmission of packets from nodes to cluster 55

head and further to base station. The threshold value of 50
distance of transmission is given by
45

dt = fs / mp (7) 40

35
Where fs and mp are the radio energy parameters
30
(Table I) for free space and multi path fading model. 2 3 4 5 6 7 8 9 10 11 12
If the transmission distance, d is less then the dt thresh- Number of Clusters
old, free space model is adopted, otherwise the multi-path Fig. 5. Average energy dissipated per round.
model is used. The energy used to send a l bit data at a
distance d is given by secondly far away from the field: 9 m < disttoBS < 63 m
 and 53 m < disttoBS < 151 m. An estimate of the optimum
l ∗ Eelec + l ∗ fs ∗ d 2 if d ≤ dt number of clusters, kopt 22 is given by
ETX = (8)
l ∗ Eelec + l ∗ mp ∗ d 4 if d ≥ dt 
fs
√
For receiving a l bit message, the energy consumed is kopt = · M N (10)
4
− Eelec 

ARTICLE
 mp dtoBS
ERX = l ∗ Eelec (9) Using the above equation, we calculate the optimum
number of clusters to be 8 < kopt < 10. These analytical
where Eelec is the energy consumed in the electronic circuit results are verified using simulations on a 100 node net-
to transmit or receive the signal, Delivered
fs and mp by
are Publishing
the energy Technology to: Guest User
work by considering the average energy dissipation per
consumed by the amplifier for transmissionIP: 180.151.225.6
at shorterOn: or Tue,round
18 Aug 2015 07:50:54
as shown
Copyright: American Scientific Publishers in Figure 5. This graph shows that the
longer distance. optimum number of clusters are around 9–12 for the 100-
node network with minimum energy dissipation. It is seen
4.2. Optimum Number of Cluster Heads that the simulation results of energy dissipation cover the
range as obtained through analytical results of Eq. (10).
The selection of the number of clusters is a critical task Thus for this algorithms, we set the number of clusters to
in any clustering algorithm in a general size network as be 10.
it depends upon how many cluster heads are chosen in an
optimum manner. Each cluster has a cluster head that is
4.3. Confidence Bounds
responsible for the data aggregation of the data received
from its cluster members and does not take part in the Each time the simulations are conducted for a different
sensing operation. In this algorithm, the average energy per placement of nodes, the distance between nodes and CHs
round is plotted versus the number of clusters as shown is different, which results in different network lifetime.
in Figure 5. For our experiment, two ranges of distances The data of network lifetime is collected which varies with
between nodes and base station are observed when the the number of cycles. In this context it is necessary to
base station is placed firstly in the centre of the field and estimate the confidence bounds for network lifetime. The
bounds measure the confidence that the obtained network
Table I. Simulation parameters for transmission. lifetime lies within a range regardless of the random dis-
tribution of nodes.
Description Symbol Value The bounds are defined with a level of desired certainty.
Number of nodes in the system N 100 The level of certainty is often 95%, but it can be any value
Initial energy EInit 2 J/1.5 J such as 99%, 99.9%, and so on. The method for finding
Size of the data packet — 500 bytes
the confidence bound is based on the mean of the data
Hello/Broadcast/CH_Join message — 25 bytes
Energy consumed by the amplifier to fs 10 pJ/bit/m 2 which is given by n
x
transmit at a short distance x = i=1 i (11)
Energy consumed by the amplifier to mp 0.0013 pJ/bit/m 4
n
transmit at a longer distance The standard deviation of the data values is
Energy consumed in the electronics circuit Eelec 50 nJ/bit  n
i=1 xi − x
2
to transmit or receive the signal
s= (12)
n−1
Adv. Sci. Focus, 1, 111–119, 2013 117
 Maximum Residual Energy Based Clustering Scheme for Wireless Sensor Networks Azad and Sharma

Table II. Network lifetime within 95% and 99% confidence bounds. 100

90 MREC
s-MREC (cycles) MREC (cycles) s-MREC
80 DEEC
Homogeneous Heterogeneous Homogeneous Heterogeneous

Number of Alive Nodes

EEHC
70
UCL99% 3622 4388 3827 5306
60
LCL99% 3597 4297 3719 5055
UCL95% 3619 4377 3814 5276 50
LCL95% 3600 4307 3732 5085 40

30

This estimated standard deviation is used to create the 20

lower confidence limit (LCL) and upper confidence limit 10

(UCL) about the mean given by 0
0 1000 2000 3000 4000 5000 6000 7000 8000
s
x ± t n−1 √ (13) Time (Cycles)
n
Fig. 7. Network lifetime for heterogeneous environment.
where t in the above equation represents the student −t
distribution and  is 0.05 for 95% confidence and 0.01 for of each advanced node). The stability region, in which all
99% confidence. The calculated values of LCL and UCL the sensor nodes are sensing the environment, is found to
for overall network lifetime in cycles in our experiments be significantly higher for MREC and s-MREC than for
are shown in Table II. other protocols (DEEC and EEHC). It is observed that
almost 88% of the nodes in MREC die quickly within a
ARTICLE

4.4. Performance time span of 26 cycles, which shows that the dissipation
of energy in the majority of nodes is uniform. Particu-
For comparison, computed WSN lifetimes for s-MREC, larly, the death of the first node occurs after 3561 cycles
MREC (99 % confidence bound) and LEACH protocols in MREC; 3239 cycles in s-MREC, 1500 cycles in DEEC
Delivered
are depicted in Figure 6 (for 2 joules by Publishing
of initial energy of Technology
and 1008 to: Guest
cycles in User
EEHC. In EEHC, the death of the
IP: 180.151.225.6
the nodes) in a homogeneous environment. It is clear fromOn: Tue, 18 Aug 2015 07:50:54
Copyright: American Scientific Publishers in the 1000th cycle, due to which
node starts very early
the figure (Fig. 6) that the nodes remain alive for longer there is a less number of nodes collecting the information
duration for s-MREC and MREC as compared to LEACH. most of the time as compared to other protocols (MREC,
Also MREC has the longest stability region compared to s-MREC and DEEC). Interestingly, the longest lifetime
the other two protocols. (5264 cycles) is associated with MREC protocol, which
While for the heterogeneous environment, an equal is almost 22% higher than s-MREC (4325 cycles), 32%
number of advanced and normal nodes are randomly dis- than EEHC (4000 cycles) and 42% higher than DEEC
tributed over the field with the sink placed in the centre. (3714 cycles).
The initial energy of normal nodes is 1.5 J and that of Figure 8 shows the total number of data packets received
advanced nodes is 3 J. The results for proposed MREC at the base station from different cluster heads which is
and s-MREC for an average cycle within the 99% con- found to be higher for MREC and s-MREC protocols
fidence bound (as shown in Table II) are compared with
EEHC and DEEC (for similar environment) as shown in
100
Figure 7 where the total initial energy of the system is
90 MREC
1.5nEinit (n is the total number of nodes, Einit is the initial s-MREC
energy of each normal node and Einit is the initial energy 80 DHAC
Number of Nodes Alive

70
100 60
90
Number of Alive Nodes

MREC 50
80 s-MREC
70 LEACH 40
60
30
50
40 20
30
10
20
10 0
0 0.5 1 1.5 2 2.5 3 3.5 4
0
0 1000 2000 3000 4000 5000 6000 7000 Number of data packets received at Base Station ×105
Time (Cycles)
Fig. 8. Number of nodes alive versus number of data packets received
Fig. 6. Network lifetime for homogeneous environment. at base station.

118 Adv. Sci. Focus, 1, 111–119, 2013

Azad and Sharma Maximum Residual Energy Based Clustering Scheme for Wireless Sensor Networks

14 MREC has highest stability region and a 32% longer life-

MREC time than EEHC and 42% more than DEEC.
12 EEHC

10 References and Notes

Number of Clusters

8 1. I. F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, Com-

put. Networks 38, 393 (2002).
6 2. A. Cerpa, J. Elson, M. Hamilton, and J. Zhao, Habitat monitoring:
Application driver for wireless communications technology, ACM
4 SIGCOMM’2000, Costa Rica, April (2001).
3. E. P. de Freitas, T. Heimfarth, and C. E. Pereira, Evaluation of
2 coordination strategies for heterogeneous sensor networks aiming at
surveillance applications, Proceedings of IEEE Sensors (SENSORS),
0 Christchurch, New Zealand (2009), pp. 591–596.
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 4. M. Rahimi, R. Baer, O. Iroezi, J. Garcia, J. Warrior, D. Estrin, and
Cycles M. Srivastava, Cyclops: In situ image sensing and interpretation in
wireless sensor networks, Proceedings of the ACM Conference on
Fig. 9. Comparison between EEHC and MREC: Number of clusters per
Embedded Networked Sensor Systems San Diego, CA (2005).
round.
5. J. M. Corchado, J. Bajo, D. I. Tapia, and A. Abraham, IEEE Trans.
Inf. Technol. Biomed. 14, 234 (2010).
as compared to DHAC in an homogeneous environment. 6. H. C. Romesburg, Cluster Analysis for Researchers, Lifetime Learn-
ing Publications, Belmont, California (1990).
More specifically, the number of data packets received 7. J. Y. Yu and P. H. J. Chong, IEEE Commun. Surv. Tutor. 7, 32 (2005).
(till the death of all nodes) at the base station is 1,49,309 8. A. A. Abbasi and M. Younis, Comput. Commun. 30, 2826 (2007).
in MREC, 1,48,242 for s-MREC; and 26,500 in DHAC 9. W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrishnan, IEEE

ARTICLE
protocol. Figure 9 shows the variation in the number of Trans. Wirel. Commun. 1, 660 (2002).
10. O. Younis and S. Fahmy, IEEE Trans. Mob. Comput. 3, 366 (2004).
clusters for the entire life of the network. It is found that
11. C. H. Lung and C. Zhou, Ad Hoc Networks. 8, 328 (2010).
number of clusters is invariant up to 1000 cycles using 12. F. Bajaber and I. Awan, J. Comput. Syst. Sci. 77, 282 (2011).
MREC protocol. Also the stability region is significantly 13. D. Kumar,
Delivered by Publishing Technology to:T.Guest
C. Aseri, and R. B. Patel, Comput. Commun. 32, 662
User
higher in comparison to EEHC which indicates a higher (2009).
IP: 180.151.225.6 On: Tue, 18 Aug 2015 07:50:54
lifetime in MREC. 14. L. Qing, Q. Zhu, and M. Wang, Comput. Commun. 29, 2230 (2006).
Copyright: American Scientific Publishers
15. G. Smaragdakis, I. Matta, and A. Bestavros, SEP: A stable elec-
tion protocol for clustered heterogeneous wireless sensor networks,
Proceedings of Second International Workshop on Sensor and Actor
5. CONCLUSIONS Network Protocols and Applications (SANPA) (2004).
16. J. Dunn, Cybernet. Syst. 3, 32 (1973).
In this paper, we have investigated the problem of non- 17. J. C. Bezdek, Pattern Recognition with Fuzzy Objective Function
uniform energy dissipation and the limited lifetime of Algorithms, Plenum Press, New York (1981).
the network by developing an energy efficient clustering 18. X. Zhu, L. Shen, and T. S. P. Yum, IEEE Trans. Veh. Technol. 58, 990
scheme based on maximum residual energy. The initial (2009).
19. X. Min, S. W. Ren, J. C. Jiang, and Z. Ying, Int. J. Elect. Commun.
number of clusters is fixed and optimum and varies in 64, 289 (2010).
accordance with the node density in each cluster with the 20. H. Zhou, Y. Wu, Y. Hu, and G. Xie, Comput. Commun. 33, 1843
start of reduction in number of nodes. The simulations in (2010).
static and dynamic mode show that the proposed protocol 21. T. Rappaport, Wireless Communications: Principles and Practice,
is able to extend the lifetime of the network as compared IEEE Press, Piscataway, NJ, USA (1996).
22. F. Comeau, S. C. Sivakumar, W. Robertson, and W. J. Phillips,
to the LEACH method in a homogeneous environment and Energy conserving architectures and algorithms for wireless sensor
DEEC and EEHC protocols in a heterogeneous environ- networks, Proceedings of the 39th Annual Hawaii International Con-
ment for same input parameters. It is also found that the ference on System Sciences (2006), Vol. 9, p. 1.

Adv. Sci. Focus, 1, 111–119, 2013 119