Optimization of Energy-efficient Protocols with Energy-heterogeneity for Coverage
Preservation in Wireless Sensor Networks: An Empirical Study
Femi A. Aderohunmu, Jeremiah D. Deng, Martin K. Purvis
Department of Information Science, University of Otago
PO Box 56, Dunedin 9056, New Zealand{afemi,ddeng,mpurvis}@infoscience.otago.ac.nz
Abstract—The advances of wireless and sensing technologieshave opened up new doors for wide application of sensornetworks. To fully achieve the potentials of wireless sensornetworks, however, a few challenging issues have to be solved.While a majority of the research has focused on energypreservation, little attention has been paid to the coveragepreservation as a quality of services requirements. Anotherissue that is often overlooked is the inherent energy heterogene-ity among sensor nodes in the network. In this paper, we willconduct an empirical study on some improved heterogeneity-aware clustering protocols, and compare their performance onnetwork lifetime as well as coverage preservation under energyheterogeneity. We define a “full coverage time”, as well asa spatial uniformity measurement. Simulation results demon-strate the superiority of the new protocols in this respect. Ourconclusion is that the energy heterogeneity can be harnessedand it is promising to find optimized solutions for prolongednetwork lifetime and coverage preservation simultaneously.
Keywords-wireless sensor networks; clustering; coveragepreservation
I. INTRODUCTION
In recent years wireless sensor networks (WSN) have
gained much attention in the research communities due to
the fast advances in wireless communication technologies
that have paved the way for developing diverse applications
and services. The capabilities of sensor devices offers great
potential for use in different application domains ranging
from monitoring of our immediate environment to more
sophisticated industrial surroundings or production systems.
There is however a growing need for WSN nodes to handle
more complex tasks such as video streaming and biometric
recognition, and energy-saving remains a major requirement
for these battery-powered sensor nodes even though en-
ergy harvesting techniques are becoming available. Much
research attention has been given to energy conservation in
WSNs [3].
To provide robust functionalities in WSN applications,
another concern closely connected with energy consumption
is coverage maintenance. If the spatial distribution of live
nodes is not even during operation, the network can quickly
lose coverage to certain regions for monitoring. This is
highly undesirable for surveillance applications where a per-
sistent coverage is necessary for an application that requires
continuous surveillance in order to satisfy a specific quality
of service (QoS) constraint. Coverage maintenance has re-
ceived attention in recent studies [8], where a few coverage-
aware cost metrics are defined. In these approaches, sensors
that are important to the coverage task are less likely to be
selected as cluster leaders.
On the other hand, existing studies on clustered WSNs
have mainly focused on energy-homogeneous networks
where sensor nodes have the same energy capacities. In
realistic application scenarios, sensor nodes are prone to
failures and damages; on other occasions, nodes can be
replaced during network operation, and new nodes could
be deployed. These lead to energy heterogeneity among the
network nodes. So far, the effect of energy heterogeneity in
WSNs has not received much discussion. Overlooking such
a factor may result in not only reduced network lifetime, but
also speedier failure in maintaining network coverage.
Motivated by earlier works [7], [8], we investigate the
effect of energy heterogeneity in clustering protocol that
favors both energy preservation and coverage maintenance.
In a nutshell, we model the energy imbalance existing in
the network with a three-tier hierarchy and we analyze
the performances of the protocols with respect to coverage
preservation. We also investigate the spatial uniformity of
live node distribution and define a quantitative metrics for
performance evaluation.
The remainder of this paper is organized as follows.
We briefly review related work in Section II. New sensor
network models are introduced in Section III, followed by
our metrics defined for performance evaluation. Section IV
presents the simulation settings and the results obtained. Fi-
nally, in Section V we conclude the paper with a discussion
on future work.
II. RELATED WORK
The existing work related to clustering approach is so vast
that it would deserve an extensive literature review. Hence
we will only discuss a few related issues that is directly
2012 IEEE 14th International Conference on High Performance Computing and Communications
978-0-7695-4749-7/12 $26.00 © 2012 IEEE
DOI 10.1109/HPCC.2012.172
1165
2012 IEEE 14th International Conference on High Performance Computing and Communications
978-0-7695-4749-7/12 $26.00 © 2012 IEEE
DOI 10.1109/HPCC.2012.172
1173
connected with our work. Clustering is an effective approach
in handling energy management in WSNs. Low Energy
Adaptive Clustering Hierarchy (LEACH) [6] is a pioneering
work in this respect. In HEED [9] a hybrid criterion for
cluster-head selection was proposed based on residual energy
and a node’s proximity to its neighbors. Both LEACH and
HEED solutions assumed a homogeneous setup, which does
not conform to the real-world operation of a typical WSN.
Most likely, nodes will often have different energy levels as a
result of the terrain, recharging or new nodes deployment for
replacing dead ones. In Stable Election Protocol (SEP) [7],
it is demonstrated how energy heterogeneity can be better
utilized by adapting the cluster-head election probabilities
according to the sensors’ energy settings. The SEP protocol
uses two energy levels among sensor nodes to characterize
the heterogeneity problem. This is extended in the SEP-E
model [2], which proposed a three-node setup and further
adapted the cluster-head election probabilities to prolong the
network lifetime.
All the above works focused on a stochastic model to
improve the lifetime of WSN performance. Recently, a deter-
ministic model was proposed in [1], achieving a better per-
formance in network lifetime than the previous approaches
in both homogeneous and heterogeneous settings. This paper
is an extension to [1], by considering coverage preservation
and spatial distribution of energy as performance criteria.
On the other hand, the importance of network coverage
has been discussed along with network lifetime and energy
efficiency in [4]. Soro and Heinzelman proposed a set
of coverage-aware cost metrics and presented a cluster-
based network organization in [8]. They further proposed an
unequal clustering to force a balanced energy consumption.
A popular definition of functional network lifetime of
WSN is the time until the first node dies out [6], [7]. Others
defined the network lifetime to be the time in which the net-
work is able to perform sensing functions and transmit data
to the base station [4]. To our knowledge there is no network
lifetime definition related to coverage maintenance. Indeed,
spatial distribution of energy has hardly been quantitatively
investigated in the literature. In [6], the performance of the
LEACH protocol is visually compared with that of other
protocols, showing snapshots of live node distributions.
The contribution of this work is two-fold: 1) We define
new performance metrics that measure the coverage preser-
vance and coverage uniformity of WSNs; 2) we present an
empirical study that includes a number of recent clustering
protocols and evaluating their coverage performance under
energy heterogeneity settings.
III. THE WSN MODELS
A. The energy dissipation model
We consider the energy dissipation model as used in
a number of previous studies [6], [7]. Depending on the
transmission distance, two propagation modes are used: free
space and two-way. Details are given in [1].
B. Cluster formation in SEP-E and DEC
LEACH is based on the idea of using a random number
thresold to randomly decide whether a network node should
be elected as the cluster-head. SEP and SEP-E extended
this idea by adapting the threshold for different energy
capacities: normal, intermediate (not in SEP), and advanced
nodes. Details about the thresholding and random election
process are clearly outlined in [6], [7], [2].
The cluster formation process in DEC [1] is different
from the approaches used by LEACH, SEP and SEP-E.
DEC uses a deterministic model to elect cluster-heads (CH)
in each cluster, strictly depending on the residual energy
of each node as the network evolves. At the beginning
of the network, i.e., Round 1, the BS sets up the clusters
by choosing the nodes with the highest residual energy to
be the cluster-heads. In the subsequent rounds the clusters
self-elect the next cluster-head base on the residual energy
information which is piggy-backed during the exchange of
JOIN REQUEST sent by non-CHs to their respective CHs
at Round 1. This process continues until the sensor nodes
die out.
Once the cluster is formed, the CH node sets up a
TDMA schedule that governs the data transmission in the
cluster. The logical assignment of each non-CH node to a
cluster forms a logical Voronoi graph across the network.
The TDMA schedule ensures that there are no collisions
among data packets and also allows turning off the radio
components of each non-CH node at all time except dur-
ing their transmission [6]. In order to reduce interference
among multiple cluster-heads, the communication with BS
is achieved using CDMA codes.
More details of DEC are available in [1].
C. Network Coverage
In previous work [4], [8], there are two basic coverage
problems associated with clustered WSN: point coverage
and area coverage. The former is normally used in appli-
cations where the deployment is deterministic, while the
latter is used more often in random deployments. In our
case we deal with area coverage, and we define the network
coverage as the summation of coverage of each individual
node modeled as a circular area of radius R with the sensor
node sitting at the center. Without any specific applications
in mind, we assume a scenario which is delay tolerant and
requires continuous data delivery, but it is mission critical
to maintain the full coverage. Here we further assume that
each sensor node does not move once deployed, and is
equipped with an omni-directional antenna that monitors
its own range. Note that due to the random deployment of
sensor nodes in the simulation, it is possible for the network
to start with a coverage less than 100%, but normally the
11661174
uniform distribution of the nodes allow the coverage to be
above 95% at start in our simulations.
Our coverage definition has two implications. First, it
deviates from the assumption in LEACH that sensor nodes
of the same cluster detect the same event. In LEACH it was
assumed that the distance between nodes within a cluster
is small compared with the distance from which events are
sensed. The robustness of coverage is inherent under this
condition: one or two nodes’ death will not cause coverage
loss because other nodes in the same cluster can still sense
the same event. Under the new definition, however, even
though there may be overlapped areas between sensor nodes,
the death of one node will potentially cause coverage loss.
It is therefore a more stringent definition. And secondly, it
is likely that the sensor nodes will sense different events,
which therefore makes the new definition more appropriate
for modeling monitoring applications that demand localized
information, e.g., wild-fire monitoring or motion detection
in a vast open space.
This leads us to define a new metric called ‘full coverage
time’ (FCT), which integrates the coverage maintenance
with network lifetime: FCT is the time from the beginning of
the network operation till the network suffers from coverage
loss. This is different from the previous definition of network
lifetime (NLT), which is only indirectly related to coverage
preservation. As initial node death may or may not cause
coverage loss, we have NLT ≤ FCT.
D. Spatial Uniformity
In some application scenarios that tolerate sporadic sensor
deaths, strict coverage preservation is unnecessary. Live
nodes evenly distributed across the entire monitoring area
can roughly maintain an effective sensing coverage. It is
however possible for a large amount of nodes of a certain
monitoring region to all suddenly die out, and that region
becomes lost in coverage. Along with energy consumption
over time and therefore potential node deaths, how evenly
the network manages to maintain its sensors’ energy across
the monitoring area is also relevant to maintaining effective
coverage. Hence, this translates to measuring the spatial
uniformity of coverage as the network evolves. A good
protocol design is expected to maintain spatial uniformity
of coverage, which implies that energy consumption is more
even, and it is less likely for the network to suffer from
coverage loss due to early node deaths.
To evaluate how energy consumption occurs spatially over
time, we choose a cutting point during the simulation where
the amount of dead nodes reaches 50%. The entire sensing
area is split by a s × s grid, and live nodes within each of
the (M/s)2 cells can be counted. We use the variance of the
number of live nodes across all cells as a measurement of
the spatial uniformity of coverage. Obviously the smaller the
variance is, the more uniformly energy is consumed across
the network, resulting in live nodes more evenly spread
over the sensing area. This idea is shown in Fig.1. We can
then carry out F-test to verify whether the difference on the
variance measurement is significant or not across different
protocols.
0 20 40 60 80 1000
20
40
60
80
100A 5x5 grid at 50% network operation
livenode
deadnode
Figure 1: Live node distribution can be examined, e.g., by
using a 5× 5 grid.
IV. SIMULATION
A. Experiment Setup
Following the scenario used in [6], [7], we assume a
100m × 100m region with 100 sensor nodes scattered
randomly within. MATLAB is used to implement the sim-
ulations of four protocols: LEACH, SEP, SEP-E and DEC.
Their simulation results are then compared. Common pa-
rameters used in the simulation are shown in Table I. The
Table I: Common parameter settings for all experiments.
Parameter Values
Eelec 50nJ/bit
EDA 5nJ/bit/message
Eo 0.5J
k 4000
Popt 0.1
εfs 10pJ/bit/m2
εmp 0.0013pJ/bit/m4
R 10/15mn 100
heterogeneity configurations that is tested for the protocols is
m = 0.2, b = 0.3, α = 3 and μ = 1.5. Different settings will
affect the operation of the network. For example, we let 20%and 30% of the nodes be advanced nodes and intermediate
nodes with additional energy levels (3 and 1.5 times more
energy than the normal nodes, respectively).
LEACH was proposed for homogeneous energy settings.
For fair comparison, we initialize the network with the same
energy configuration and run LEACH so as to assess how
it copes with the heterogeneous setting. To run SEP simu-
lations, the setting is configured in a way that it maintains
the same total energy as that of other protocols. Despite the
difference on the internal settings of the four protocols, it is
reasonable to compare their performance as they start with
the same or similar initial conditions.
11671175
B. Simulation Results
1) Performance metrics: The metrics below are used to
assess the performance of all clustering protocols involved:
1) Stability period: the period from the start of the
network operation to the first node death.
2) Instability period: the period between the first node
death and the last node death.
3) Full Coverage-Time (FCT): the period from the start
of the network operation till full coverage is no longer
maintained.
4) Number of live (and dead) nodes per round.
5) Half-life: the period from the start of the network to
time when any node uses up half of its energy.
6) Spatial uniformity of energy spread in the network.
As explained in [7] the larger the stability period and the
smaller the instability period are, the better the reliability of
the clustering process of the network is. However, we need to
note the trade-off between the reliability and lifetime of the
network system. In some cases the last alive node can still
provide feedback, but this could in most cases be unreliable.
So we expect a balance between these trade-offs, which is
reflected in both the stability and instability periods.
Multiple simulation runs are carried out for all four
protocols: LEACH, SEP, SEP-E and DEC. We first examine
network coverage performance using FCT under energy
heterogeneity conditions, then we investigate the network
lifetime with different load of extra energy, and finally
examine the spatial uniformity by using visual comparison
of node heat-maps, and using the evaluation of spatial
uniformity statistics.
Table II: Summary of Average lifetime of the sensors for 10
trials with Etotal = 102.5J , when the BS is located inside
the sensing region.
Protocols FCT Stability Half-life Instability
LEACH 1241 995 478 4585
SEP 1398 1385 631 5050
SEP-E 1624 1450 704 3751
DEC 2075 1839 945 640
2) Network lifetime under heterogeneity: Fig.2 shows the
coverage evolution over time measured in rounds during
one set of the simulations. Obviously DEC keeps the full
coverage for the longest time. This is very important for a
surveillance application with strict coverage requirements,
e.g., situation management in disaster response [5]. If the
coverage requirement is relaxed down to 80% coverage,
DEC still performs the best. In same manner, SEP-E per-
formed better than both SEP and LEACH up till about 75%.
Furthermore, LEACH seems to take over SEP-E after around
2000 rounds, while SEP stands out for coverage as low as
60% i.e at about 3000 rounds. One may note that LEACH
and SEP manages to keep around 20% coverage as long as
4500 rounds into simulation, however, the reliability of the
sensing result is already in question. When less coverage is
required for some application, down to 60% for instance,
it seems that SEP and LEACH then outruns both DEC and
SEP-E.
1000 2000 3000 4000 5000 6000 70000
10
20
30
40
50
60
70
80
90
100
Rounds
Cov
erag
e(%
)
LEACHSEP−EDECSEP
Figure 2: Comparing coverage performance of DEC, SEP-E,
LEACH (m = 0.2, b = 0.3, α = 3 and μ = 1.5) and SEP
(m = 0.3, b = 0, α = 3.5 and μ = 0), Etotal = 102.5J .
Previous studies have shown that LEACH performs most
competitively when nodes are homogeneous [7]. This be-
havior is expected according to the optimized cluster-head
election in the homogeneous setup. The performance data of
the four protocols are summarized in Table II. On average,
the stability period as well as the half-life time of DEC and
SEP-E compared with LEACH and SEP are significantly
longer. Likewise the stability period of SEP is significantly
better than that of LEACH. The ranking of instability period
length is SEP, LEACH, SEP-E and DEC.
We further observe the performance of the protocols to
different load of heterogeneity as shown in Fig.3 where
the network lifetime (stability period) is plotted against the
percentage of total extra energy added to the network. It
is apparent that DEC significantly and increasingly outper-
forms the other protocols. Even though both SEP-E and SEP
are close in performance, SEP-E consistently outperforms
SEP. DEC, SEP-E and SEP continuously perform better as
energy heterogeneity increases in the network. On the other
hand, LEACH is majorly oblivious to the load of extra
energy heterogeneity. One of the advantages of the DEC
protocol is that it elects the cluster-heads based on their
respective residual energies hence it is able to cope well
with energy heterogeneity.
Further, FCT values are measured over ten simulation
runs respectively for the four protocols, and the statistics are
reported in Table II. Both DEC and SEP-E have a clear-cut
advantage over SEP and LEACH.
3) Spatial Uniformity: To visualize the energy distribu-
tion within the network over time, we plot out the ‘heat-
map’ of the sensor node energy at three points in time:
Round 1500 and Round 2200, roughly located at the stable
and instable periods respectively. The heat-maps are shown
in Fig.4 and Fig.5. A red pixel on the map indicates the
11681176
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9800
1000
1200
1400
1600
1800
2000
2200
Net
wor
k Li
fetim
e
Percentage of total extra energy (a*m +μ*b)
DECLEACHSEP−ESEP
Figure 3: Performance of DEC, SEP-E, LEACH and SEP as
the percentage energy load increases.
location is close to a live node. While the red fades away,
meaning the spot is still covered by a live node despite an
increase in distance. The blue color indicates the spot is not
covered by any live nodes.
From these figures, we observe that both DEC and SEP-
E offer better coverage compared with LEACH and SEP at
Round 1500, while SEP also outperforms LEACH. When
we examine the coverage in the early instability period at
Round 2200, DEC still shows a robust performance than
the other protocols. Further on, DEC is able to evenly
balance the energy consumption among the nodes resulting
to a prolonged FCT and hence the nodes die out closely at
the same time. In SEP-E, SEP and LEACH, the majority
of the normal nodes are dead but advanced nodes are
well preserved, and hence we see the prolonged instability
period, most obviously in SEP. It should be noted that these
behaviors of SEP-E, SEP and LEACH (in terms of coverage)
could be good for some applications with minimal coverage
requirements. But for a mission-critical application with full
coverage requirements, DEC is superior to SEP-E, SEP and
LEACH.
For quantitative analysis we use the spatial uniformity
defined in Section III-D. Grids of 4×4 and 5×5 granularities
are used to get the variance of live nodes in cells as the
spatial uniformity measure. For completeness, we have used
the same seed for the node distribution in order for good
comparisons among the protocols. The smaller the variances
are, the more uniform the live nodes distribute themselves
across the monitored region. To assess the difference on the
variance values obtained by different protocols, we resort to
the F-test, where the F-ratio (i.e., variance-ratio) indicates
the equality of two variances. If the the F-ratio is 1.00, it
indicates no statistical difference between the two values.
The bigger the ratio is, the more likely the two variances are
different. Usually it is expected that the p-value should have
less than or equal to 0.05 for a high statistical significance
level. Because of the pair-wise nature of the test, we conduct
the pairwise comparisons of DEC versus other protocols. As
it makes little sense to compare the early stages of network
operation, we choose the 50% network operation time, i.e.,
10 20 30 40 50 60 70 80 90 100
10
20
30
40
50
60
70
80
90
100
(a)
20 40 60 80 100
10
20
30
40
50
60
70
80
90
100
(b)
20 40 60 80 100
10
20
30
40
50
60
70
80
90
100
(c)
20 40 60 80 100
10
20
30
40
50
60
70
80
90
100
(d)
Figure 4: The heat-map at Round=1500 for (a) DEC
coverage, (b) SEP-E coverage, (c) SEP coverage, and (d)
LEACH coverage.
10 20 30 40 50 60 70 80 90 100
10
20
30
40
50
60
70
80
90
100
(a)
10 20 30 40 50 60 70 80 90 100
10
20
30
40
50
60
70
80
90
100
(b)
10 20 30 40 50 60 70 80 90 100
10
20
30
40
50
60
70
80
90
100
(c)
10 20 30 40 50 60 70 80 90 100
10
20
30
40
50
60
70
80
90
100
(d)
Figure 5: The heat-map at Round=2200 for (a) DEC
coverage, (b) SEP-E coverage, (c) SEP coverage and (d)
LEACH coverage.
when the network reaches the point with only 50% nodes
still alive, as the cutting point of the simulation.
We conduct the F-test on results from five independent
experiments for each protocol and collect the pair-wise F-
ratios, giving the box-plot in Fig.6. DEC has the highest vari-
ance ratios compared with the other protocols with medium
sitting well above 2. Overall the difference is however only
statistically significant on the p = 0.10 level.
To summarize, we have used two methods to assess the
spatial uniformity of coverage: visual assessment of cover-
age heat-maps, and F-test on the defined spatial uniformity
measurement. Overall, DEC and SEP-E report better spatial
uniformity among the sensor nodes than both LEACH and
SEP. DEC is able to best tackle the effect of energy-
11691177
2.0
2.2
2.4
2.6
Protocols
Var
ianc
e R
atio
DEC vs. SEP DEC vs. LEACH DEC vs. SEP−E
2.0
2.2
2.4
2.6
2.0
2.2
2.4
2.6
Figure 6: Comparison of spatial uniformity of the protocols.
The boxplot is generated from pairwise F-ratios averaged
over 5 independent experiments.
heterogeneity. This is reasonable, as the DEC model guar-
antees that the nodes with the highest residual energy gets
elected as cluster-head.
V. DISCUSSIONS AND CONCLUSION
The WSN clustering scheme, originally inspired by [6]
and further enhanced by [7], was proposed to cope with
energy consumption in WSN. Following this approach, this
paper introduced an empirical study on energy consumption
and coverage preservation of two new protocols SEP-E
and DEC in heterogeneous settings. Nodes in SEP-E elect
themselves as cluster-heads stochastically, based on their
energy levels, retaining more uniformly distributed energy
among sensor nodes. Nodes in DEC use a more deterministic
approach to elect themselves as cluster-head, thus showing
a better performance than the other protocols. Our results
show that DEC and SEP-E is more robust with respect to
network lifetime and the full coverage time. In most of
the cases, DEC and SEP-E maintain a balanced and more
uniform spread of energy among nodes.
Both the DEC and SEP-E model design can be very
important for applications that require continuous re-
energization of nodes throughout the data retrieval process,
by deploying new nodes to replace dead ones, or it could
be useful when testing the effect of introducing updated
sensor nodes into an existing network. This provides a
justification for the different energy levels we have used in
our simulations.
It should be noted that we have not specifically attempted
any mathematical framework for coverage optimization. We
have however demonstrated empirically that it is possible to
take advantages of the energy heterogeneity in WSNs for
both energy and coverage preservation. We have not either
tried out different heterogeneity settings exhaustively and the
simulations are conducted in a limited application scenario.
Furthermore, we have focused on an architecture with single-
hop routing (via the cluster-head), but for larger networks a
more reasonable solution is to use relayed, multi-hop routing
among the CHs. This would have impact on the energy
consumption patterns and the coverage uniformity, and these
will be interesting directions for further investigation in light
of energy heterogeneity.
ACKNOWLEDGMENT
This work has been supported in part by the University
of Otago Postgraduate Scholarship.
REFERENCES
[1] F. A. Aderohunmu, J. D. Deng, and M. K. Purvis. Adeterministic energy-efficient clustering protocol for wirelesssensor networks. In Seventh International Conference on In-telligent Sensors, Sensor Networks and Information Processing(ISSNIP), pages 341 –346, Dec. 2011.
[2] F. A. Aderohunmu, J. D. Deng, and M. K. Purvis. Enhancingclustering in wireless sensor networks with energy heterogene-ity. IJBDCN, 7(4):18–31, 2011.
[3] G. Anastasi, M. Conti, M. Di Francesco, and A. Passarella.Energy conservation in wireless sensor networks: A survey.Ad Hoc Netw., 7(3):537–568, May 2009.
[4] M. Cardei and J. Wu. Energy-efficient coverage problems inwireless ad-hoc sensor networks. Computer Communications,29(4):413 – 420, 2006.
[5] S. George, W. Zhou, H. Chenji, M. Won, Y. O. Lee, A. Pazar-loglou, R. Stoleru, and P. Barooah. DistressNet: a wirelessad hoc and sensor network architecture for situation manage-ment in disaster response. Communications Magazine, IEEE,48(3):128 –136, mar. 2010.
[6] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan.An Application-Specific Protocol Architectures for WirelessNetworks. IEEE Transactions on Wireless Communications,1:660–670, 2002.
[7] G. Smaragdakis, I. Matta, and A. Bestavros. SEP: A StableElection Protocol for clustered heterogeneous wireless sensornetworks. In Proceeding of the International Workshop onSANPA, 2004.
[8] S. Soro and W. B. Heinzelman. Cluster head election tech-niques for coverage preservation in wireless sensor networks.Ad Hoc Networks, 7(5):955 – 972, 2009.
[9] O. Younis and S. Fahmy. HEED: A Hybrid, Energy-Efficient,Distributed Clustering Approach for Ad Hoc Sensor Networks.IEEE Transactions on Mobile Computing, 3:366–379, 2004.
11701178