Content uploaded by Sabrine Khriji
Author content
All content in this area was uploaded by Sabrine Khriji on Aug 30, 2020
Content may be subject to copyright.
Content uploaded by Sabrine Khriji
Author content
All content in this area was uploaded by Sabrine Khriji on Aug 30, 2020
Content may be subject to copyright.
Content uploaded by Sabrine Khriji
Author content
All content in this area was uploaded by Sabrine Khriji on Aug 31, 2019
Content may be subject to copyright.
A Fuzzy Based Energy Aware Unequal
Clustering for Wireless Sensor Networks
Sabrine Khriji1,2(B
), Dhouha El Houssaini3, Ines Kammoun2,
and Olfa Kanoun1
1Chemnitz University of Technology, Measurement and Sensor Technology,
Chemnitz, Germany
sabrine.kheriji@etit.tu-chemnitz.de, kanoun@ieee.org
2LETI Laboratory, National School of Engineers of Sfax,
University of Sfax, Sfax, Tunisia
3Centre for Research on Microelectronics and Nanotechnology,
Technopark of Sousse, University of Sousse, Sousse, Tunisia
Abstract. One of the most important issues in wireless sensor networks
is energy autonomy. Thereby wireless communication leads to excessive
demands of energy. In this paper, a fuzzy based energy aware unequal
clustering algorithm is presented. The network is partitioned into certain
number of rings. An energy analysis model is proposed to measure the
optimal radius of each cluster. This enables to vary the size of clusters
from one ring to another, which ensures the load balance of the network.
Then, a fuzzy logic system is employed to select suitable cluster head.
The fuzzy set relies only on three parameters; residual energy, number of
neighboring nodes and centrality of node among its neighbors. The pro-
posed algorithm outperforms other clustering approaches, like LEACH,
DUCF and MCFL in terms of energy efficiency and network lifetime.
Keywords: Wireless sensor networks ·Unequal clustering
Multi-hop communication ·Optimal cluster radius ·Fuzzy logic
1 Introduction
Wireless Sensor Networks (WSNs) is an interesting axis of research in the last
decade for its wide range of applications such as industry, automation, smart
buildings and smart agriculture [1]. Wireless communication can lead in general
to excessive demands of energy. As a result, different communication strategies
for energy consumption reduction are proposed to improve the efficiency of the
network [2]. Many studies have shown that hierarchical routing protocols based
on clustering schemes can reduce significantly the energy consumption [3].
A Low-Energy Adaptive Clustering Hierarchy [4] (LEACH) is the first hier-
archical based protocol for WSN. Nodes create clusters and the cluster heads
(CHs) act as routers communicating directly with the base station (BS) which
will save energy during data transmission. A DUCF [5] algorithm is a distributed
load balancing unequal clustering based on fuzzy approach. In this algorithm,
c
Springer Nature Switzerland AG 2018
N. Montavont and G. Z. Papadopoulos (Eds.): ADHOC-NOW 2018, LNCS 11104, pp. 1–6, 2018.
https://doi.org/10.1007/978-3-030-00247-3_12
2 S.Khrijietal.
the smaller cluster size is assigned to CH nearer to BS since it acts as a router for
other distant CHs. DUCF ensures load balancing among the clusters by varying
the cluster size of its CH nodes. A multi-clustering algorithm based on fuzzy
logic (MCFL) [6] reduces the number of CH selections which enables to reduce
the number of sent messages while guaranteeing energy efficiency.
According to these related works, the calculation of the optimal radius of each
cluster is not considered. However, having the appropriate radius can lead to load
balancing. This paper is addressed for investigation of clustering and network
organization of WSNs for low energy consumption. It focuses on an energy aware
unequal clustering algorithm, which is characterized with a circular partitioning
network model. The aim of this proposal is to balance the energy consumption
among clusters by forming adequate sized clusters to solve the hotspot problem
created by multi hop communication and increase the overall network lifetime.
To determine the radius of each cluster, an energy analysis model is used, which
enables to vary the size of clusters from one ring to another. Thus, the energy of
the network is balanced. To select the CHs, the proposed clustering algorithm
is based on fuzzy logic which uses three parameters; node residual energy, node
density and node centrality.
In Sect. 2, a detailed description of the proposed data routing algorithm is
performed including mainly the system model, assumptions, work contributions
and the main characteristics. In Sect. 3, the proposed algorithm is compared with
existing algorithms.
2 Proposed Fuzzy Based Unequal Clustering
Four our proposal, the network area is assumed as a circle with the BS in the
center. It is partitioned into a certain number of rings with a specific radius. The
choice of a circular partitioning scheme is based on the comparison performed in
[7], which prove that circular network model have better accuracy in the energy
consumption analysis than the rectangular one. In [7], authors prove also that
the use of rectangular network is inaccurate for the varied dimension compared
to circular partitioning scheme. All sensor nodes are randomly distributed with
a uniform distribution of a density λ. All nodes are stationary with an unequal
initial energy level. Each data packet will be transmitted from one cluster head
to another in different rings.
The radio model used is based on the model introduced in [4,5,8,9]. To
transmit an lbit message to a receiver over a distance d, the energy consumed
by the transmitter ETx(l, d) and by the receiver ERx(l)are:
ETx(l, d)=l×Eelect +l×εfs ×d2if d<d
0
l×Eelect +l×εamp ×d4if dd0
(1)
ERx(l)=lEelect (2)
Eelect presents the energy required for one-bit long time span to run the trans-
mitter’s or receiver’s circuitry. The parameters εfs and εamp are defined, respec-
tively, the energy consumption factor for free space and multi path radio models.
d0is a threshold distance.
A Fuzzy Based Energy Aware Unequal Clustering for WSNs 3
The proposed routing is divided into two phases; an off-line phase and cluster
formation phase. The first phase introduces an energy analysis to compute the
optimal cluster radius for different rings, while the second phase consists on
cluster building and CH selection.
2.1 Off-Line Phase
Firstly, we assume that the nodes are aware of their position and the ring where
they are located. After network deployment, the optimal radius of each cluster
will be determined. Assuming that both sensor nodes and CHs are uniformly
distributed in the area. The network area with radius Ris partitioned into L
rings. R=L×δ, with δis the width of a ring. Due to multi-hop communication,
distance used by single-hop schemes is relatively short. We assume, that the
distance between all sensor nodes is less than the critical distance d0,sothe
total energy of the whole network is expressed as in Eq.3:
EClusterk =lERx (Nk
mk
−1) + lEDA
Nk
mk
+lL
i=k+1 mi
mk
(ERx +ETx +εfsE[d2
CHk,CH(k−1) ]) (3)
+l(ETx +εfsE[d2
CHk,CH(k−1) ])
The total energy spent by all nodes in a ring is
ETotalk =mkEClusterk (4)
The optimal radius of cluster in a ring kis therefore described in Eq. 5
R(CHk)opt =δ2k−1
mk
k=2,...,L (5)
2.2 Cluster Formation Phase
All nodes within same ring have same radius of cluster. This radius differs from
one ring to another. After finishing the off-line phase, each node broadcasts a
discovery neighboring message within the cluster radius to create its routing
table containing list of neighbors, residual energy and current position.
Each node may calculate independently its chance to be a CH based on
three parameters; residual energy, density and centrality which are extending
the network lifetime. Node density presents the number of neighbor nodes in
a tentative CH’s neighbor nodes set. The centrality represents how central the
node is among its neighbors. For the three parameters, a trapezoidal membership
function is used for describing “Low” and “high” functions, while the “medium”
function has a triangular membership function (see Fig. 1).
Node will transmit its input parameters to its fuzzy deduction engine which
will calculate its chance to be a CH. A message enclosing node’s chance is
4 S.Khrijietal.
00.1 0.2 0.3 0.4 0.5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Residual energy
Membership degree
low medim high
00.1 0.2 0.3 0.4 0.5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Node density
Membership degree
low medim high
00.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Node centrality
Membership degree
low medium high
020 40 60 80 100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Chance
Membership degree
vw w lw m lh h vh
Fig. 1. Membership function for: residual energy, node density, node centrality, and
chance to be a CH
exchanged between nodes within the same cluster. Then, the node will com-
pare its chance to the chance of its neighbors within the same cluster. The node
having the highest chance will select itself as a CH and transmit an announce-
ment message to its node members informing them with its status. The nodes
receiving this message will transmit a join message to the respective CH. If the
node is not a final CH and has received more than one final CH message, it will
select the final CH with the high cost to join it. If a node finished the clustering
process and has not yet received any final CH message, it announces itself as a
final CH.
After the selection of CH, each node transmits its data packet to its adequate
CH. Only CHs from the first ring will transmit their aggregated packet directly
to BS. Other CHs will transmit their packets to CH from previous ring, which
plays the role of a relay node.
3 Performance Evaluation
To evaluate the performance of the proposed routing algorithm different metrics
are respected which are the total remaining energy of the network, the number
of first dead node in each round (FND), the half of dead nodes (HND) and
last dead node (LND) are used to achieve this comparison. The simulations are
performed using Matlab where 100 nodes are randomly deployed. The BS is
located at center of the field as seen in Fig. 2. All detailed configuration is illus-
trated in Table 1. Simulations (Figs. 3a and b) show that the proposed algorithm
outperforms LEACH, DUCF and MCFL in terms of FND, HND and LND met-
rics. Considering the FND metric, the proposed algorithm is more efficient than
A Fuzzy Based Energy Aware Unequal Clustering for WSNs 5
Fig. 2. Node dispersion and clus-
tering for the first round
Table 1. Configuration parameters
Parameter Value
Network size 100 m ×100 m
Number of sensor nodes 100
Initial energy 0.5 J
Data packet size 4000 bits
Eelect 50 nJ/bit
εfs 0.0013 pJ/bit/m4
εamp 10 pJ/bit/m2
Fig. 3. (a) Total remaining energy of the methods in each round, (b) Number of nodes
dead over rounds
LEACH by 49%, DUCF by 13% and MCFL by 5%. According to HND met-
ric, the proposed algorithm is more efficient than LEACH about 21%, DUCF
about 12% and MCFL about 2%. For the LND metric, FEAUC is more efficient
than LEACH about 12%, DUCF about 2% and MCFL about 1%. The obtained
results (Fig. 3a and b) show that the energy consumption of LEACH is maxi-
mum and the network lifetime is minimum because in each round, each node
in the network has to compute the threshold and generates a random number,
which consumes many CPU cycles. Cluster head selection ensures that nodes
with high residual energy and maximum number of neighboring nodes decrease
the overall energy consumption of network. Further, LEACH algorithm does
not use unequal clustering which generates “hot spots” and reduces the number
of data packets received at base station. Varying the cluster size of CH nodes,
DUCF ensures good results in term of energy consumption and also network life-
time. Reducing the number of cluster head selection and reducing the repeated
sending of messages, MCFL increases the network lifetime.
6 S.Khrijietal.
4 Conclusion
In this paper an unequal size clustering algorithm is proposed. The algorithm
enables to balance the energy consumption among all sensor nodes which will
increase the network lifetime. The proposed solution is split into two phases, off-
line phase and cluster formation phase. The first phase is performed to compute
the optimal radius of each cluster. The second phase uses fuzzy logic based on
residual energy, number of neighboring nodes and centrality of node among its
neighbors. Simulation results show that the proposed algorithm provides a better
performance compared to LEACH, DUCF and MCFL clustering algorithms.
It outperforms the others and realizes acceptable results in terms of energy
management and network lifetime.
In future works, a new strategy based on the use of relay node will be
employed during the CH selection which will be responsible for data packet
transfer.
References
1. Khriji, S., El Houssaini, D., Jmal, M.W., Viehweger, C., Abid, M., Kanoun, O.:
Precision irrigation based on wireless sensor network. IET Sci. Meas. Technol. 8(3),
98–106 (2014)
2. Leite, J.R.E., Ursini, E.L., Martins, P.S.: Simulation of AdHoc networks includ-
ing clustering and mobility. In: Puliafito, A., Bruneo, D., Distefano, S., Longo, F.
(eds.) ADHOC-NOW 2017. LNCS, vol. 10517, pp. 199–209. Springer, Cham (2017).
https://doi.org/10.1007/978-3-319- 67910-5 16
3. Ogundile, O.O., Alfa, A.S.: A survey on an energy-efficient and energy-balanced
routing protocol for wireless sensor networks. Sensors 17(5), 1084 (2017)
4. Heinzelman, W.R., Chandrakasan, A., Balakrishnan, H.: Energy-efficient commu-
nication protocol for wireless microsensor networks. In: Proceedings of the 33rd
Annual Hawaii International Conference on System Sciences, p. 10. IEEE (2000)
5. Baranidharan, B., Santhi, B.: DUCF: distributed load balancing unequal clustering
in wireless sensor networks using fuzzy approach. Appl. Soft Comput. 40, 495–506
(2016)
6. Mirzaie, M., Mazinani, S.M.: MCFL: an energy efficient multi-clustering algorithm
using fuzzy logic in wireless sensor network. Wirel. Netw. 24(6), 2251–2266 (2018)
7. Suharjono, A., Hendrantoro, G., et al.: A new area partitioning strategy for
unequal clustering of wireless sensor networks. In: 2012 International Conference
on Advanced Computer Science and Information Systems (ICACSIS), pp. 71–75.
IEEE (2012)
8. Bagci, H., Yazici, A.: An energy aware fuzzy approach to unequal clustering in
wireless sensor networks. Appl. Soft Comput. 13(4), 1741–1749 (2013)
9. Kim, J.M., Park, S.H., Han, Y.J., Chung, T.M.: CHEF: cluster head election mech-
anism using fuzzy logic in wireless sensor networks. In: 2008 10th International
Conference on Advanced Communication Technology, ICACT 2008, vol. 1, pp. 654–
659. IEEE (2008)
