Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317

Contents lists available at ScienceDirect

Journal of King Saud University –

Computer and Information Sciences
journal homepage: www.sciencedirect.com

A survey on unequal clustering protocols in Wireless Sensor Networks

Sariga Arjunan ⇑, Pothula Sujatha
Department of Computer Science, Pondicherry University, Puducherry, India

a r t i c l e i n f o a b s t r a c t

Article history: WSN became a key technology for ubiquitous living and remains an active research due to wide range of
Received 14 September 2016 applications. Energy awareness is a critical design issue in WSN. Clustering is the most popular energy
Revised 28 March 2017 efficient technique and provides various advantages like energy efficiency, lifetime, scalability and less
Accepted 29 March 2017
delay; but it leads to hot spot problem. To overcome this, unequal clustering is proposed. In unequal clus-
Available online 31 March 2017
tering, the cluster size varies proportionally to the distance to Base Station (BS). Absence of recent and
detailed survey papers in unequal clustering approaches stimulated us to perform this study. In this
Keywords:
paper, a comprehensive survey of various unequal clustering approaches with their objectives, character-
Wireless Sensor Networks
Data aggregation
istics etc., is presented. Also, the classifications of unequal clustering approaches are made and compared
Clustering based on various cluster properties, Cluster Head (CH) properties and clustering process.
Hot spot problem Ó 2017 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University. This is an
Unequal clustering open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Energy-efficiency

1. Introduction (Sohraby et al., 2007). WSN is usually deployed in regions where

Rapid growth in the field of Information Technology (IT) and
integrated circuits (IC) leads to the development of cheap and com-
pact size sensor nodes. WSN is an integral part of IoT; it makes bil-
lions of devices to share data for improving the environmental user
control. WSN is composed of a set of large number of sensor nodes
arranged in adhoc fashion to observe and interact with the physical
world. Each sensor node consist of four components: sensors,
microcontrollers, power supply and transceivers. The sensors in
the sensing unit measures the physical parameters in the real
world such as temperature, pressure, humidity, vibration, acoustic
signal, infrared, vehicular movement, etc., (Estrin et al., 1999). The
sensed value is processed by the processing unit and forwarded to
the Base Station (BS) through single hop or via intermediate nodes
by the communicating unit (Anastasi et al., 2009). WSN is com-
monly used in real time monitoring and tracking applications such
as military surveillance, agriculture, disaster management, health-
care monitoring, industry automation, inventory control, etc.
⇑ Corresponding author.
E-mail addresses: sarikaaut@gmail.com (S. Arjunan), spothula@gmail.com
(S. Pothula).
Peer review under responsibility of King Saud University.
Production and hosting by Elsevier
human intervention is difficult or not possible. Energy consump-
tion, bandwidth and memory are considered as a major issue in
WSN design. As the sensors are deployed in harsh environment,
it is very difficult or not possible to replace or recharge batteries
(Akyildiz et al., 2002). The transmission cost is more than sensing
and processing cost in WSN (Raghavendra et al., 2004) Therefore,
an energy efficient data transmission strategy is required to for-
ward data from sensor nodes to BS to lengthen the network
lifetime.
Clustering is the most important energy efficient technique. In
this technique, the sensor nodes are organized into groups termed
as clusters. The regular nodes in the cluster are called as cluster
members and a Cluster Head (CH) is selected among them
(Heinzelman et al., 2000). The architecture of clustering in WSN
is shown in Fig. 1. There are two types of traffic in clustered
WSN: data transmission within a cluster defined as intra-cluster
traffic and data transmission between clusters which is termed
as inter-cluster traffic. The cluster members sense the real world
parameters and transmit the sensed value to its CH. The CH
receives and aggregates data to remove redundant data and trans-
mit aggregated data to CH directly or via intermediate CHs. The
cluster members cannot send the data directly to BS, it sends only
to the CH and CH forwards it to BS. The advantages of clustering
are: energy consumption is reduced by improving bandwidth uti-
lization, reduced overhead, increased connectivity, stabilized net-
work topology, decreased delay, effective load balancing and
reduction in the size of the routing table. The CHs near the BS

http://dx.doi.org/10.1016/j.jksuci.2017.03.006
1319-1578/Ó 2017 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
Fig. 1. Architecture of clustering in WSN.
consumes more energy and drains out energy more quickly than
the CHs farther from BS. CHs closer to BS are loaded with heavy
traffic due to intra-cluster traffic from its own cluster members,
data aggregation, and inter-cluster traffic from other CHs for relay-
ing data to BS. This result in disrupting the network connectivity
and coverage issues are created in the clusters closer to BS. This
issue is referred as hot spot problem.
In order to prevent the network from hot spot issue, unequal
clustering techniques can be utilized for load balancing between
the CHs (Soro and Heinzelman, 2005a). The architecture of unequal
clustering in WSN is shown in Fig. 2. Unequal clustering reduces
the size of the clusters closer to BS and the cluster size increases
as the distance between the BS and CH increases. The cluster size
is directly proportional to the distance of CHs from BS. When the
distance to BS increases, cluster size also increases. Smaller cluster
near the BS indicates less number of cluster members and less intra
-cluster traffic. So, the smaller clusters consume less energy for
intra-cluster traffic and concentrates more on inter cluster traffic.
Similarly, larger clusters farther from BS indicate more cluster
members and spend more energy on intra-cluster traffic. As a
result, it spends less energy for inter-cluster traffic hence no need
of spending more energy for inter cluster routing. Unequal cluster-
ing permits all CHs to spend same amount of energy so that the
CHs near BS spend equal energy as CHs farther from BS. So, unequal
clustering eliminates hot spot problem by balancing the load
efficiently.
Cluster size
increases as
the distance
between CH
and BS
increases
Fig. 2. Architecture of unequal clustering in WSN.
305
For constructing unequal clusters, each node needs to deter-
mine its own competition radius Rc. The formula for Rc (Yu et al.,
2011b) is shown in Eq. (1)
 
dmax  dðsi ; DSÞ 0
RC ¼ 1  c RC ð1Þ
dmax  dmin
where dmax and dmin are the maximum and minimum distance from
the nodes in the network to the BS, d(si,DS) is the distance from
node si to the BS, c is a weighted factor whose value is in [0, 1]
and R0C is the maximum value of competition radius. As the node
have various energy levels in heterogeneous network, the competi-
tion radius is calculated by considering distance between nodes and
the BS and the residual energy of nodes which is shown in Eq. (2).
  
dmax  dðsi ; DSÞ Er
RC ¼ 1  a b 1 Rmax ð2Þ
dmax  dmin Emax
where d(si, DS) is the distance from node si to the BS, a is a weighted
factor whose value is in [0, 1], Er is the residual energy of node si and
Rmax is the maximum value of competition radius. For intra-cluster
communication, cluster members send data to CH directly. Inter-
cluster communication involves aggregation of data from cluster
members and then sends the aggregated data to the next hop nodes.
A node si selects some nodes from its neighbor nodes to construct
the candidate forwarding nodes set and then it selects one node
from the set as the final forwarding node according to a parameter
‘‘Erelay”. the formula to compute the energy consumption Erelay
when cluster head si chooses sj as its next hop is given in Eq. (3).
So si will select the node with the highest residual energy from
the two nodes with the smallest ‘‘Erelay” in the candidate forward-
ing nodes set.
2 2
Erelay ¼ d ðsi ; sj Þ þ d ðsi ; DSÞ ð3Þ
In the last decade, various clustering and unequal clustering
techniques are presented for energy efficient WSN. Although sev-
eral comprehensive surveys are presented for clustering
approaches, very few have performed a survey on unequal cluster-
ing approaches. Probabilistic based unequal clustering protocols
are reviewed and the protocols are compared in terms of node
deployment region (square or circle), mobility, location awareness
and data aggregation (Kaur and Varsha., 2016). Another review of
Probabilistic based unequal clustering protocols is presented. It
compares various protocols based on number of nodes, energy effi-
ciency, balanced cluster, location awareness and heterogeneity
level (Vennira Selvi and Manoharan, 2013). Absence of recent
and elaborate survey of unequal clustering protocols inspired us
to perform this work. In this paper, unequal clustering objectives,
characteristics, classification, merits and demerits of each
Smaller size cluster (near to BS)
Medium size cluster
Larger size cluster (far from BS)
306 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
technique is discussed in detail and comparison is also made based
on various cluster properties, Cluster Head (CH) properties and
clustering process is given. The remainder of the paper is organized
as follows. The overview of unequal clustering which includes the
objectives and characteristics are explained in Section 2. Unequal
clustering algorithms are classified and explained in Section 3. Var-
ious classified algorithms are compared and tabulated in Section 4
and the paper is concluded in Section 5.
2. Overview of unequal clustering
In the section, an overview of unequal clustering is discussed in
detail with objectives and characteristics.
2.1. Unequal clustering objectives
The objectives of unequal clustering are same as equal cluster-
ing with some additional functions. The nodes are clustered in
WSN with different objectives which are based on application
requirement. Energy conservation and eliminating hot spot prob-
lem are the most common objectives of unequal clustering. Some
of the additional objectives are explained as below.
2.1.1. Scalability
Sensor nodes are deployed in large numbers ranging from hun-
dreds to thousands based on the application requirement in the
real scenario. The design of routing techniques should consider
the ability to work with these huge number of sensor nodes. When
a node in the cluster needs to transmit data to a node in another
cluster, the nodes should know the details of the associated receiv-
ing cluster head (CH). Hierarchical architecture provides scalability
in large scale WSN by dividing the sensing field into various layers
and each layer is again divided into number of clusters (Kleinrock
and Kamoun, 1977). This leads to increased scalability and reduces
the size of the routing table.
2.1.2. Fault-Tolerant
In several application, sensor are deployed in harsh environ-
ment (e.g., sensors are dropped from helicopter) and these nodes
have an increase risk of physical damage, malfunction of nodes.
Fault-tolerant nodes are important in critical application where
the loss of some sensor data results to catastrophe. Clustering is
an efficient way to make a fault tolerant and secured WSN (Zhou
et al., 2008). The self organized WSN manages the fault by the pro-
cess of re-clustering the network. Re-clustering process not only
increases the resource burden but also disrupts the current opera-
tion. Re-clustering, assigning backup CH, depute CH or rotating CH
results to fault-tolerance with an advantage of proper load balanc-
ing (Heinzelman et al., 2002).
2.1.3. Data Aggregation/ fusion
Since large number of sensor sense the same data in the phys-
ical environment, there is a greater chance of data redundancy.
Data aggregation is an effective way of avoiding redundant
data transmission and also it reduces the number of transmission.
This technique is a signal processing method, which aggregates all
received packets into an output packet. This technique amplifies
the common data and suppresses the unwanted noise
(Krishnamachari et al., 2002). In WSN, CH performs data aggrega-
tion of all data received from its cluster members and forwards
the aggregated data to BS (BS) via single hop or multi-hop.
Consequently, the number of transmission and the total load of
the network are also significantly reduced.
2.1.4. Load balancing
Load balancing plays an important role to prolong the network
lifetime. Load balancing is a crucial issue where CHs are selected
from available nodes in the network (Na and Vergados, 2007). Uni-
form load distribution among the CHs is essential to avoid hot spot
problem. Unequal clustering guarantees uniform load distribution
in which all the CHs consumes approximately same amount of
energy. As a result, more energy efficient network can be easily
achieved.
2.1.5. Stabilized network topology
Nodes are organized into clusters and CH is selected from each
cluster; CH is responsible for any topology changes at the cluster
level. CH contains the information of its cluster members like node
id, location and energy level. Managing the network topology in
hierarchical architecture is better than flat architecture. When a
node dies or moves to other cluster, these changes are immediately
registered and informed by CH to BS and reclustering will be done
to maintain the network topology effectively.
2.1.6. Increased lifetime
The major aim of unequal clustering is improve the network
lifetime as a long as possible. As the sensors are energy-
constrained, maximizing the network lifetime is very important
for real time applications. Intra-cluster communication can be
reduced by selecting nodes as CHs with more neighbor nodes
(Younis et al., 2003). Clustering and routing process can also be
combined to maximize lifetime (Hou et al., 2005). Unequal cluster-
ing prolongs the lifetime of WSN by rotating CHs properly among
the cluster members, sleep modes and cluster maintenance tech-
niques can be properly utilized to increase the network lifetime.
2.2. Clustering characteristics
To classify different clustering approaches, various clustering
characteristics are used. In this section, the three characteristics
of unequal clustering are discussed in detail.
 Cluster properties
 CH properties
 Clustering process properties
2.2.1. Cluster properties
The specifications of the cluster are defined as cluster properties
which include: cluster count, cluster size, intra- cluster communi-
cation and inter-cluster communication.
2.2.1.1. Cluster count. The number of cluster formed is predefined
or variable based on the application requirement. In some cases,
the number of cluster is 5% of total number of the nodes deployed.
In many applications, the number of clusters is variable when the
CHs are randomly selected.
2.2.1.2. Cluster size. The cluster size can be classified into equal and
unequal size cluster. In equal clustering, the size of the cluster is
same throughout the network. In unequal clustering, the cluster
size is determined based on the distance to BS. The cluster size is
smaller when the distance to BS is small and the size increases
as the distance to BS increases.
2.2.1.3. Intra – cluster communication. Intra – cluster communica-
tion involves the data transmission between CH and cluster mem-
ber within a cluster. Based on the clustering approaches, the
communication can be direct or multi-hop. For large scale WSN,
multi-hop communication is needed for data transmission within
a cluster.
 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
2.2.1.4. Inter – cluster communication. Inter – cluster communica-
tion can be direct or multi hop communication. Usually, multi-
hop mechanism is preferred for energy efficient data transmissions
from CHs to BS through intermediate CHs in large scale WSN. In
some applications of small scale WSN, the communication
between the CH and BS is single hop transmission.
2.2.2. CH properties
CH performs the following operations: Collecting data from
cluster members, aggregating data and forwarding data to BS
through direct or multi-hop communication.
2.2.2.1. Role. The CH receives data from its cluster members, per-
formance data aggregation of the collected sensor data, relays the
aggregated data to BS.
2.2.3. Clustering process
The characteristic of Clustering process is listed below.
2.2.3.1. Clustering methods. There are two methods of clustering:
centralized and distributed. In centralized approaches, a central
authority like BS or super nodes controls the entire operation (clus-
ter formation, CH selection etc.,) while distributed approaches
have no central authority and widely employed in large scale WSN.
2.2.3.2. Objective of Node grouping. Various objectives of nodes
grouping are already discussed in this study. e.g., Fault tolerance,
load balancing, etc.
2.2.3.3. Nature. The clustering process can be proactive, reactive or
hybrid in nature. The node continuously senses the data and for-
wards it to CH. In proactive type, the CH transmits the data to BS
continuously. In reactive type, CH transmits the data whenever
the sensed value crosses the predefined threshold. In hybrid cases,
CH transmits the data to BS at longer regular time intervals and
also when the value crosses the threshold value.
2.2.3.4. CH selection. There are three ways to select CH in WSN:
probabilistic methods, attribute based method and preset type. In
probabilistic approaches, CHs are selected randomly without any
previous consideration. In attribute based method, various metrics
are used to select CHs like residual energy, node degree, node cen-
trality, expected residual energy, distance to BS, etc. In preset type,
CHs are predetermined before placing the sensors in the sensing
field.
3. Classification unequal clustering algorithm
Generally, WSN is built of large number of sensor nodes ranging
from hundreds to thousands. Unequal clustering is an effective
way of organizing huge number of nodes uniformly distributing
the load and eliminating hot spot problem. In this section, exten-
sive literature survey of published unequal clustering algorithms
is discussed. These algorithms are classified in three ways: proba-
bilistic, deterministic and preset clustering algorithm and the hier-
archical structure is shown in Fig. 3.
3.1. Probabilistic clustering algorithm
A major aim of probabilistic clustering algorithm is to maximize
the network lifetime. In this approach, the algorithm selects the CH
randomly. This is found to be simple, near optimal overhead, faster
convergence and energy efficient clustering method. For energy
efficient clustering protocol, time complexity and message com-
plexity should be low. Probabilistic approaches are divided into
307
random methods and hybrid methods. Random approaches are
simple which selects CHs and form randomly and achieve near
optimal overhead. Hybrid methods are used to properly balance
the clusters by combining random methods with some parameters
like residual energy or distance to BS. Hybrid approaches are iter-
ative based or competition based which increases the complexity
in terms of message and time. In the next section, random
approaches are discussed.
3.1.1. Random approaches
3.1.1.1. Probability Driven Unequal Clustering Mechanism for WSN
(PRODUCE). PRODUCE (Kim et al., 2008) is a randomized unequal
clustering algorithm which eliminates the hot spot problem. It
maximizes the network lifetime and coverage time in WSN where
the density of nodes is high. Coverage time is the time when the
first CH runs out of battery resulting in coverage issues and lifetime
indicates the time when all the nodes in the network die. It uses
localized probabilities for constructing clusters of unequal size
and stochastic geometry for inter-cluster routing. Unequal cluster-
ing organizes clusters of smaller size near BS and larger size farther
from BS. This makes the CHs near the BS to focus more on inter-
cluster data relaying and CHs farther from BS can focus more on
intra-cluster communication which results in the elimination of
hot spot problem. This probability driven unequal clustering
scheme balances the energy consumption results and maximized
the network lifetime and coverage time especially in the network
where the density of the nodes is high. It gives better results in
terms of network lifetime when compared to EEUC.
3.1.1.2. Energy Driven Unequal Clustering (EDUC). EDUC is a dis-
tributed algorithm which reduces the energy consumption and
avoids hot spot problem in heterogeneous WSN (Yu et al.,
2011a). This algorithm effectively manages the energy consump-
tion of nodes within a cluster to reduce the energy depletion. It
involves an unequal clustering algorithm and an energy driven
CH rotation method. There are two phases in EDUC: cluster con-
struction and data collection phase. The cluster construction phase
includes CH competition stage and cluster formation stage. Each
node serves as CH only once in the entire network lifetime. The
CHs prepares TDMA schedule for its cluster members to avoid
intra-traffic collision. CHs are elected randomly and the energy
level is computed accurately for CH rotation. CHs forward the data
directly to BS. This assumption of single-hop communication is not
possible in many real time applications. This method is not useful
for multi-hop networks because energy threshold should be very
precise. Compared with LEACH and HEED, EDUC can prolong the
network lifetime.
3.1.1.3. Location Based Unequal Clustering Algorithm (LUCA).
Another probability based distributed scheme to prevent hot spot
problem is LUCA (Lee et al., 2011), where the cluster size varies
proportionally to the distance from BS. To eliminate hot spot prob-
lem, LUCA forms smaller clusters near the BS and larger clusters
farther from BS. In the initialization stage, each node has a back
off timer with some random value. When the nodes receive any
CH advertisement message within the time interval, it joins to
the cluster. When none of the CH advertisement messages are
received, it selects itself as CH and advertises as CH to its neigh-
bors. As LUCA organizes unequal clusters based on the location
from BS, it uses GPS to determine its location. In LUCA, the nodes
are location aware which makes it unsuitable for many real time
application and increases energy overhead.
3.1.2. Hybrid approaches in unequal clustering algorithm
3.1.2.1. Energy Efficient Unequal Clustering (EEUC). EEUC is a hybrid
approach widely used for periodical data gathering application in
308 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317

Unequal clustering algorithms

Probabilistic Deterministic Preset

PRODUCE MRPUC
Random EDUC PEBECS UCS
LUCA EADUC
ACT
Weight EBCAG
EEUC EUCS
EEDUC UCMR
UCR EPUC
Unequal LEACH CUCA
Hybrid EC Imp. EADUC
EB-UCP
Unequal HEED
COCA
FUCP
EAUCF
Fuzzy IFUC
FBUC
DUCF

EBUC
GAEEP
Heuristic IPSO
SMEBUC
FAMACROW
nCRO- UCRA
UMBIC

EDDUCA
Compound
UCCGRA

Fig. 3. Classification of unequal clustering algorithms.

WSN (Li et al., 2005). To avoid hotspot problem, the network is
divided into clusters of various unequal sizes and multi-hop rout-
ing is involved in forwarding data to BS. The cluster size closer to
BS is smaller which reduces the energy consumption due to
intra-cluster traffic and save more energy for inter-cluster
multi-hop routing. The cluster size is directly proportional to
the distance to BS. EEUC is a distributed method where the
CHs are elected by localized competition and final CHs are cho-
sen based on the residual energy of the sensor node. In multi-
hop data transmission, CHs chooses the relay node using two
metrics: residual energy and distance of relay CH from BS. EEUC
significantly reduces the energy consumption and lengthens the
network lifetime compared to LEACH and HEED and more practi-
cal than UCS.
3.1.2.2. Energy Efficient Distributed Unequal Clustering protocol
(EEDUC). EEDUC (Yu et al., 2011b) is a distributed approach can
be used for periodical data gathering applications. The drawback
of EEUC is overcome by EEDUC. In EEDUC, each node broadcasts
the hello message based on the waiting time. The weight (waiting
time) metric is computed at each node based on the residual
energy and node degree; nodes with maximum weight are selected
as CHs. EEDUC avoids the hot spot problem and balances the
energy consumption 24% better than EEUC. EEDUC successfully
maximizes the network lifetime. However, the CH chooses the
adjacent CH as relay node and without considering the residual
energy and distance to BS.
3.1.2.3. Unequal Cluster based Routing (UCR). UCR (Chen et al., 2009)
divides the network into clusters of unequal sizes to avoid hot spot
problem. UCR protocol incorporates two components: energy effi-
cient unequal clustering algorithm (EEUC) for clustering process,
inter-cluster greedy geographic and energy aware routing protocol
for multi-hop routing process. BS broadcasts the beacon signal to
all sensors to calculate its distance from every node based on the
received signal strength. This is helpful for selecting appropriate
power for data transmission to BS and building unequal clusters.
Tentative CHs are chosen randomly and they compete for final
selection of CH. Each tentative CH has a competitive range and it
is used to construct unequal size clusters. After the selection of
CHs, CH broadcasts advertisement message to the network. The
node join as cluster member to the CH with higher received signal
strength and the voronoi region of sensor node is also constructed.
For multi-hop inter-cluster routing, the relay nodes are chosen
based on the ratio of residual energy and energy cost of the relay
paths. It achieves maximum lifetime over HEED but prone to error
and less robust because of the noise in the real environment.
3.1.2.4. Unequal LEACH. LEACH is a first developed clustering algo-
rithm which elects CH based on probability. Due to practical diffi-
culties in LEACH, improved LEACH with more practical setup phase
is proposed. Improved LEACH (Ren et al., 2010) constructs clusters
based on Adaptive On-demand Weighting (AOW) which is a trade-
off between residual energy to total energy and competition range.
The clusters are unequal in size and the CHs directly forward the
 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
data to BS without any intermediate nodes. Round robin method is
used for cluster head selection and the rotated time slot is prede-
termined. In steady state phase, cluster members uses TDMA
schedule to transmit data to CH and CH uses CSMA schedule to
transmit the received and aggregated data to BS. After certain time
period, re-clustering takes place. Improved LEACH results to more
stabilized network topology and maximum lifetime compared to
LEACH. This protocol avoids hot spot problem and is well suited
for large scale WSN where the density of nodes is high.
3.1.2.5. Energy-Efficient Clustering (EC). EC is a simple, scalable and
energy aware clustering algorithm which calculates the required
cluster size based on the hop distance to sink (Wei et al., 2011).
It attains approximate equalization of network lifetime and
reduces energy consumption. In EC, tentative CHs are selected ran-
domly and final CHs are elected depending on their residual
energy. An energy-efficient multi-hop data collection protocol is
proposed to determine the amount of energy consumption and
the performance of EC. The distributed inter-cluster routing algo-
rithm performs uniform energy distribution and produces minimal
overhead due to route discovery process. EC is not dependent on
the energy- efficient data collection protocol. EC performs well
and achieves more equalization of energy than UCR and HEED.
3.1.2.6. Energy – Balancing Unequal Clustering Protocol for WSN (EB-
UCP). EB-UCP is presented to attain maximum lifetime and uni-
form load balancing among all the nodes in WSN (Yang, 2009).
To eliminate hot spot problem, probabilistic approach is used to
organize clusters. Clusters near the BS are smaller in size for more
inter- cluster routing and less intra-cluster routing. The sensing
field is divided into various layers and each layer is assigned a dif-
ferent probability relative to the distance to BS. The nodes nearer
to BS have higher probability which means more CH occupies
smaller cluster area and less number of cluster members. This
results to less energy dissipation for intra-cluster traffic and save
energy for inter-cluster routing. Tentative CHs are chosen based
on the node’s residual energy to equalize the energy consumption.
The multi-hop data transmission depends on unequal clustering
algorithm and energy balancing layering algorithm. CH selects
the relay node based on the residual energy. When two nodes have
equal residual energy, CH chooses the relay node randomly from
these nodes. EB-UCP achieves maximum lifetime compared to
LEACH and EEUC.
3.1.2.7. Unequal Hierarchal Energy Efficient Distributed Clustering
(UHEED). UHEED is a distributed approach which resolves the hot-
spot problem and maintains same amount of residual energy in the
network and maximizes the network lifetime (Ever et al., 2012). It
is the improved version of HEED (Younis and Fahmy,, 2004). HEED
uses two parameters for clustering: residual energy and node
degree. The CHs use the same competition radius and distance to
BS is not considered. This leads to hot spot problem. UHEED con-
structs various size clusters depending upon the distance of CHs
to BS and smaller clusters are formed for CHs nearer to BS. This
method also finds the approximate cluster size using the distance
information. A competition radius formula is used to create smaller
size cluster near BS. This result to reduced energy consumption in
smaller clusters due to intra-cluster traffic and preserve more
energy to focus on more inter cluster traffic. This algorithm results
to increased network lifetime compared to LEACH, HEED and
Unequal LEACH.
3.1.2.8.Constructing Optimal Clustering Architecture (COCA). COCA is
a scalable, distributed unequal clustering scheme which investi-
gates the logical difficulties in unequal clustering methods in
homogenous sensor network (Li et al., 2013). COCA constructs
309
optimal clustering architecture, energy-aware CH rotation and
routing for approximate equalization of energy consumption in
the whole network. COCA uses a strategy that the number of clus-
ter in the unit area increases when the distance to BS decreases. It
saves more energy for multi hop routing between clusters which
results in the elimination of hot spot problem. For CH selection,
all nodes exchange residual energy information with its neighbors.
The node with maximum residual energy declared them as CH.
Each CH randomly choose some CH in the neighbor cluster as rout-
ing candidates and CH with highest residual energy is elected as
final routing CH. It is a simple and effective protocol with reduced
energy consumption and maximum lifetime twice or thrice than
UCR.
3.2. Deterministic approaches
On contrast to probabilistic approaches, deterministic
approaches use standard metrics for selecting CHs. The commonly
used traditional metrics are residual energy, node degree,
expected residual energy, distance to BS, node centrality, etc.
which are attained locally. This information is usually updated
by exchanging message between its neighbors. This approach is
called as deterministic approaches because the clusters with
elected CHs are more controllable. This is again is classified into
four categories: Weight based, Fuzzy based, heuristic based and
Compound unequal clustering algorithm. In weight based
approach, a weight is calculated at each node based on some met-
rics such as residual energy, node degree, distance to BS, etc. The
node with minimal weight is elected as a cluster head. Fuzzy
logic is used to elect CHs in situations where uncertainties are
more. The cluster head is chosen based on fuzzy input parame-
ters. The input parameters can be residual energy, node degree,
distance to BS, node centrality, etc. and the output fuzzy param-
eters are cluster size and probability of becoming CHs. The clus-
tering problem in WSN is considered to be a NP hard problem
and evolutionary algorithms are suitable to produce optimal solu-
tions for NP hard problems. In the recent years, heuristic based
clustering algorithms provide optimal solution in the process of
selecting CHs and cluster size. Many optimization algorithms
such as Genetic Algorithm (GA), Ant Colony Optimization (ACO),
Artificial Bee Colony Optimization (ABC) Optimization, Particle
Swarm Optimization (PSO), Bacterial Foliage Algorithm (BFA), Dif-
ferential Evolution (DE), Simulated Annealing, etc are used in
WSN. Each algorithm defines various metrics in fitness function
to achieve better performance. Heuristic approaches are central-
ized and a central authority like BS controls all operations in
the network. In exceptional cases, some approaches works in dis-
tributed manner using agent nodes. Compound algorithm uses
different metric like connected graph, Sierpinski triangle, etc. in
clustering methods.
3.2.1. Weight based unequal clustering algorithms
3.2.1.1. Multi-hop Routing Protocol with Unequal Clustering (MRPUC).
MRPUC (Gong et al., 2008) is a distributed approach aims to equal-
ize the energy consumption in all nodes for hot spot problem and
forward the data to BS with relays to reduce energy consumption.
MRPUC selects CH with more residual energy and the regular
nodes joins the cluster as cluster members in which the CH con-
tains maximum residual energy and lesser distance to BS. Relay
nodes are chosen based on minimal energy consumption to relay
data and the CHs with more residual energy. Inter–cluster routing
tree is formed as a network backbone and multi-hop routing is
done for effectively transmitting data from CH to BS. The network
lifetime is increased by 34.4% when compared to equal clustering
schemes.
310 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
3.2.1.2. Partition Energy Balanced and Efficient Clustering
Scheme (PEBECS). PEBECS (Wang and Yang, 2009) mitigates the
hot spot problem and uniformly distributes the load throughout
the network. It divides the sensing field into equal partitions and
groups the nodes in the partition to clusters of unequal sizes. CH
selection is done by a weight based heuristic algorithm based on
residual energy, node degree and distance to BS. A CH competition
radius algorithm is introduced to assign number of cluster mem-
bers in each cluster. The CH in the partition near to BS is smaller
in size to reduce intra-cluster traffic and save energy for inter-
cluster traffic. It effectively balances the energy consumption,
improves scalability and lengthens the network lifetime
significantly.
3.2.1.3. Energy-Aware Distributed Unequal Clustering Protocol
(EADUC). EADUC is a distributed multi-hop protocol for heteroge-
neous WSN to provide balanced energy, uniform load distribution
and continuous coverage throughout the network (Yu et al.,
2011b). In EADUC, a tradeoff between the average residual energy
of neighboring nodes with residual energy of the node itself is
computed. Based on the ratio value, CHs are selected. Using the
residual energy and distance to BS, different cluster radius is deter-
mined and no isolated points are present in EADUC. Smaller clus-
ters are formed near BS and larger clusters are formed when the
CH is located away from BS; it makes all the CHs to consume equal
amount energy which automatically avoids hot spot problem and
maximizes the lifetime significantly. For inter-cluster routing, a
new parameter called threshold distance is introduced. When the
distance of BS from CH is does not exceed the threshold value, it
sends the data directly. When the distance is more, relay nodes
are chosen based on residual energy. Time complexity and mes-
sage complexity is low in EADUC. It achieves maximum lifetime
compared to LEAD, LEACH, HEED.
3.2.1.4. Arranging cluster size and data transmission WSN (ACT). ACT
is a distributed routing protocol which eliminates the hot spot
problem by reducing the cluster size closer to BS, where the CHs
spend more energy for inter-cluster multi-hop routing (Lai et al.,
2012). By this method, every CH spends approximately same
amount of energy and CHs next to BS does not drain out of energy
quickly. To prolong the network lifetime, ACT partitions the net-
work into a number of hierarchical levels. ACT calculates the cluster
radius mathematically to construct unequal size clusters and cross
level data transmission is proposed to improve the network life-
time. ACT successfully shows the energy consumptions that CHs
has same amount of energy dissipation throughout the network.
3.2.1.5. Energy Balancing Unequal Clustering Approach for Gradient
based routing (EBCAG). EBCAG is a distributed clustering approach
which balances and reduces the energy consumption of all CHs
which prevents the network from hot spot problem (Liu et al.,
2012). In EBCAG, each sensor node holds a gradient value, which
represents the minimum number of hop count to BS. It gives the
optimal cluster radii based on the gradient value. Tentative CHs
are randomly selected with a probability T (predefined threshold).
Tentative CHs whose residual energy is maximum is selected as
final CHs. Based on the CHs gradient value, the unequal cluster
radius is constructed. The CH collects the data from its cluster
members and forwards the aggregated data to BS based on the
descending gradients of CH. EBCAG successfully attains energy
equalization among CHs and improves the network lifetime signif-
icantly compared to EDUC and HEED.
3.2.1.6. Enhanced Unequal Clustering Scheme (EUCS). Existing clus-
tering approaches selects CHs with maximum residual energy
and rotates the CH periodically for uniform load distribution in
the cluster. As a result, more energy is wasted for CH re-election
and gathering data from its cluster members. To preserve energy
in the process of re-clustering and eliminating hot spot issue in
multihop WSN, EUCS is proposed (Messai, 2013). It organizes the
nodes into unequal size clusters. In EUCS, CHs are selected based
on two metrics namely, residual energy and distance to BS. Re-
election takes place when the energy level of the current CH
exceeds the threshold value. This minimizes the frequent re-
election of CH and eliminates overhead. For inter-cluster routing,
a new parameter called threshold distance is introduced. When
the distance of BS from CH is below threshold value, it sends the
data directly. In some situations where the distance is high, relay
nodes are chosen based on residual energy. EUCS has improved
network lifetime when compared to HEED.
3.2.1.7. Unequally Clustered Multi-hop Routing protocol (UCMR).
UCMR is a simple protocol where the cluster size depends on the
distance to BS (Jinwala et al., 2008). For energy, efficiency, Dijk-
stra’s algorithm is used to find the shortest path route for both
inter-cluster and intra-cluster communication. UCMR configures
the clusters in each round. The deployed region is divided into
unequal clusters; clusters closer to BS are smaller in size and the
size increases when the distance to BS increases. UCMR uses vari-
ous parameters like residual energy, number of neighboring and
distance to BS to choose cluster heads. In the first round, the node
centrality is used to select CH because of the nodes at the centre
contains more neighbors resulting to less intra-cluster traffic and
reduces the chance of hot spot problem. CH of the next rounds is
decided by CH in the previous round. Dijkstra’s algorithm finds
the shortest route to relay traffic from CH to BS. UCMR improves
the error rate and data rate. The network lifetime is increased
due to better positioning of CHs. UCMR improves lifetime 40% over
UCR and 75% over LEACH.
3.2.1.8. Energy and proximity based unequal clustering algorithm
(EPUC). EPUC balances the energy consumption of all CHs by spa-
tially distributing the clusters (Bagci et al., 2010). In EPUC, the area
is divided into tracks around the BS. Nodes located at the same
track create clusters of same size. In EPUC, small cluster are formed
in the proximity of the BS. EPUC uses two CH distance metrics: dis-
tance to BS and distance to its neighboring nodes within a cluster.
In CH selection process, nods with maximum residual energy is
chosen as candidates CH (CCHs) and distance metric rules are
applied to select final CHs from CCHs. EPUC eliminates hot spot
problem achieves maximum lifetime compared to UCR and EEDC.
3.2.1.9. Coverage aware and Unequal Clustering Algorithm (CUCA).
CUCA is a first coverage based distributed scheme which uses
unequal clustering and single hop communication (Mazumdar
and Om, 2015). The major aim of CUCA is uniform load distribution
among the CHs to prevent the network from hot spot problem. For
maximum lifetime, the node whose sensing range is completely
covered by the sensing range by its neighbors is given higher pri-
ority to be selected as CHs over the partially overlapped sensor
nodes. The cluster radius is computed by the distance of the node
from CH where the cluster radius is inversely proportional to the
distance from CH. Unequal clustering in CUCA eliminates the hot-
spot problem. The cluster member whose sensing area is fully cov-
ered by its neighbors is moved to sleep mode to avoid redundant
data and minimizes energy consumption significantly. When the
overlapped nodes are not present, CHs are selected based on par-
tially overlapped area and residual energy. Hence, the result of
the node death does not disrupt the coverage of the deployed
region. Moreover, the nodes having less residual energy and the
covering area is completely covered by its neighbor are moved to
 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
sleep state to save energy. CUCR algorithm achieves maximum life-
time compares to CA and HEED.
3.2.1.10. Improved Energy Aware Distributed Unequal Clustering for
heterogeneous WSN (Improved EADUC). Improved EADUC aims to
lengthen the lifetime of EADUC and avoid hot spot problem in
multi-hop heterogeneous WSN (Gupta and Pandey, 2016). It is
widely used in continuous data gathering applications. It differs
from EADUC in the consideration of node degree while computing
competition radius in addition to residual energy and distance to
BS. Node degree is included for proper energy balancing in the net-
work. In Improved EADUC, CHs are selected based on the ratio of
average energy of neighbor nodes and residual energy of the node
itself. The competition radius is calculated based on three metrics:
residual energy, distance to BS and node degree. Improved EADUC
uses energy as a relay metric for selecting relay nodes while
EADUC uses distance to BS as a relay metrics. For several rounds,
same cluster setup is used which eliminates re-clustering overhead
and minimizes energy consumption. Improved EADUC achieves
maximum lifetime than EADUC and HUCL.
3.2.2. Fuzzy based unequal clustering algorithm
Because of uncertainties occurring in WSN environment, num-
ber of protocol uses fuzzy logic for making decisions effectively.
Fuzzy has many advantages over classical approaches which
include: low computational complexity, more flexibility, less
development cost, less memory, less design time and fault tolerant.
In WSN, Fuzzy logic is used for selecting CHs effectively. Fuzzy
logic for CH selection uses input parameters like residual energy,
distance to BS, distance from neighbors, node degree, centrality,
excepted residual energy and the output parameters are CH selec-
tion probability and cluster size. Various fuzzy based approaches
are proposed to attain energy efficiency in unequal clustering
algorithm.
3.2.2.1. Energy Aware Fuzzy Unequal Clustering algorithm (EAUCF).
EAUCF (Bagci and Yazici, 2010) is also a distributed approach used
to lengthen the network lifetime like EEUC. In EAUCF, tentative
CHs are randomly selected and residual energy is used to elect final
CHs. EAUCF is mainly focused to determine the competition radius
of CHs to evenly distribute the load, This results to the elimination
of the hot spot problem. EEUC uses residual energy to determine
the competition radius while EAUCF uses fuzzy logic to compute
the competition radius based on residual energy and distance to
BS. When the tentative CH residual energy and distance to BS is
high, then the competition radius is large and vice versa. It is more
stable and attains maximum lifetime when compared to LEACH
CHEF and EEUC.
3.2.2.2. Improved Fuzzy Unequal Clustering algorithm (IFUC). IFUC
(Mao et al., 2012) is a distributed approach which reduces the
energy consumption. It avoids hot spot problem and also lengthens
the network lifetime. Fuzzy logic is used for electing CHs and
determining the radius of the cluster. The fuzzy input parameters
are residual energy, distance to BS and node density. The output
parameters are the probability of becoming CH and cluster radius.
After each tentative CH determines the chance of becoming final
CHs and competition radius, they compete for final CHs by
exchanging messages. If a tentative CH finds other CH within its
range and higher chance, it is selected as final CH. Ant Colony Opti-
mization (ACO) finds the shortest path for inter-cluster routing
(Dorigo and Gambardella, 1997). ACO is used for energy aware
inter-cluster routing to provide an optimal path from CH to BS.
In ACO, the next hop route node is selected based on the commu-
nication cost and degree of energy consumption along the path.
IFUC provides improved network lifetime over LEACH and EEUC.
311
3.2.2.3. Fuzzy based Unequal Clustering Protocol (FUCP). FUCP is a
novel clustering algorithm which incorporates a CH selection algo-
rithm and relay traffic distributed algorithm to eliminate hot spot
problem (Gajjar and Talati, 2015). Fuzzy Inference System (FIS) is
used for selecting CHs in a distributed manner. Mamdani method
is used for fuzzification. FUCP uses three input parameters for CH
selection: residual energy, centrality of the node and Link Quality
Indicator (LQI). Usage of LQI in CH selection increases the reliability
of WSN. The fuzzy output parameter is probability of becoming CH.
To avoid hot spot problem, FUCP determine the number of CHs in a
layered hexagon using scatter factor and distance of hexagon from
BS. Node scatter factor defines the average distance of each node to
its neighbor node in the same hexagon. High scatter factor repre-
sents that the node are scattered from each other and requires
more CHs to concentrate intra-cluster communication. FUCP
achieves 40% more energy-efficient, 30% maximum lifetime and
57% more packet transmission to Distributed Energy Efficient Hier-
archical Clustering.
3.2.2.4. Fuzzy logic Based Unequal Clustering (FBUC). FBUC is also a
distributed clustering algorithm which concentrates on the
method of joining cluster members with the CH (Logambigai and
Kannan, 2016). FBUC is the improved version of EAUCF. Tentative
CHs are selected based on a probabilistic method. After the selec-
tion of tentative CHs, fuzzy logic is used to determine competition
radius. The fuzzy input parameters for calculating competition
radius are: residual energy, distance to BS and node degree. Resid-
ual energy and node degree is used to elect final CHs. The nodes
join the CH based on the CH degree and distance to CH to effec-
tively utilize the energy and maximize the network lifetime. It
achieves maximum lifetime when compared to LEACH and EAUCF.
3.2.2.5. Distributed Load Balancing Unequal Clustering in Wireless
Sensor Networks using Fuzzy approach (DUCF). DUCF is a distributed
approach for uniform load distribution which balances the load
among the clusters for eliminating hot spot problem which in turn
extends the network lifetime (Baranidharan and Santhi, 2016). It
uses fuzzy for CH selection and to determine the cluster size. To
avoid hot spot problem, smaller clusters are formed for CHs closer
to BS and larger cluster size for CHs farther from BS. The three
input parameters to FIS are residual energy, node degree and dis-
tance to BS. Probability of becoming CH and cluster size are the
two fuzzy output parameters. Mamdani method is used for fuzzifi-
cation and centroid method is used for defuzzification. The output
parameter size restricts the number of cluster members in a cluster
resulting in proper load balancing. Multi-hop data transmission for
inter-cluster routing which also reduces the energy consumption.
DUCF performs well when compared to LEACH, CHEF and EAUCF.
3.2.3. Heuristic based unequal clustering algorithm
3.2.3.1. Energy Balanced Unequal Clustering for Wireless Sensor
Networks (EBUC):. EBUC is a centralized unequal clustering proto-
col commonly used in periodical data gathering applications
(Jiang et al., 2010). It uses PSO algorithm to select candidate CHs
and divide the network into various sized clusters. The multi-hop
inter-cluster routing protocol is based on greedy algorithm; resid-
ual energy and distance to BS is used to choose a relay node. This
algorithm reduces the rate of dead nodes and lengthens the net-
work lifetime.
3.2.3.2. Genetic Algorithm based Energy-Efficient Adaptive Clustering
Hierarchical Protocol (GAEEP). GAEEP uses Genetic Algorithm (GA)
to determine the number and position of CHs to reduce the energy
consumption (Abo-zahhad et al., 2014). The entire operation
undergoes many rounds; each round consists of two phases
namely setup phase and steady state phase. In setup phase, BS runs
312 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
GA and determines the optimal number of CHs and position of CH.
The inter-cluster routing from CH to BS takes place in steady state
phase. When a node is located much closer to BS than any CH, the
node directly transmits the data to BS. Each CH uses TDMA sched-
ule and assign slots to its cluster members to avoid collision due to
intra-cluster communication. To reduce inter-cluster collision, CHs
uses CDMA code to reduce energy consumption. GAEEP is more
energy-efficient and reliable than LEACH, SEP, ERP, LEACH-GA
and DEU in both homogeneous and heterogeneous networks.
3.2.3.3. Unequal Clustering by Improved Particle Swarm Optimiza-
tion. Improved Particle Swarm Optimization (IPSO) is used in
Energy Balanced Unequal Clustering (EBUC) to improve the net-
work lifetime (Salehian and Subraminiam, 2015). This algorithm
is a distributed scheme which eliminates hot spot problem and
also overcomes the standard PSO issues. EBUC algorithm operates
in several rounds and each round starts with a setup phase fol-
lowed by a steady state phase. Each node sends its information
(node id, energy level, location, etc.) to BS and. BS runs IPSO algo-
rithm to select best candidate CHs. Cluster formation and CH selec-
tion takes place in setup phase. TDMA schedule is prepared by CHs
during setup phase to avoid collision due to intra-cluster commu-
nication. Cluster members send data in their respective slot and it
forwards the data to BS. IPSO eliminates the standard PSO issues
and number of dead node is significantly reduced in several
rounds.
3.2.3.4. Sink Mobility based Energy Balancing Unequal Clustering
Protocol (SMEBUC). SMEBUC (Fan and Du, 2015) is proposed to bal-
ance the energy consumption and it uses Shuffled Frog Leaping
Algorithm (SFLA) to elect CHs and form clusters of various sizes
according to residual energy (Zhu, 2009). To minimize the re-
clustering overhead, CH continuously works to find the exchange
time and node weight. There are two stages in SMEBUC: cluster
establishment and data transmission. N cluster heads to be
selected from M nodes leads to the optimization problem.
Improved SFLA algorithm is used to partition clusters of various
sizes. Greedy algorithm finds the relay node for multi-hop inter-
cluster routing. To eliminate hot spot problem, sink location
mobile algorithm is used. This algorithm achieves less energy dis-
sipation and effective balancing than LEACH and EBUCP.
3.2.3.5. Fuzzy and ant colony optimization based combined MAC,
routing and unequal clustering cross-layer protocol for wireless sensor
networks (FAMACROW). FAMACROW is a cross layer hierarchical
protocol which includes: CH selection, unequal cluster formation
and inter-cluster multi-hop routing protocol (Gajjar et al., 2014).
FAMACROW consists of three phases: setup phase, neighbor find-
ing phase and steady state phase. During setup phase, the nodes
are built into layers. In neighbor finding phase, each node broad-
cast its details using non-persistent CSMA MAC protocol. The
steady state phase includes: CH selection, clustering and data
delivery. CHs are selected by fuzzy logic and ACO is used for
inter-cluster multi-hop routing. Fuzzy logic uses three input
parameters to select CHs which includes residual energy, number
of neighboring nodes and quality of communication link. The reli-
ability is increased by the inclusion of link quality in CH selection.
ACO chooses the relay node based on four parameters: residual
energy, distance to BS, queue length and delivery likelihood.
FAMACROW achieves 41% more energy-efficient, 63% more net-
work lifetime, 15% more throughput than IFUC.
3.2.3.6. Novel Chemical reaction optimization based unequal cluster-
ing and routing algorithm for Wireless Sensor Networks (nCRO-
UCRA). nCRO-UCRA (Srinivasa and Banka, 2016) is a distributed
approach eliminates the hot spot problem by combining novel
chemical reaction optimization (nCRO) paradigm with unequal
clustering and routing algorithm. nCRO algorithm is used for
selecting CHs and cluster members joins CH based on derived cost
function. The CH forwards the data to BS using UCRA algorithm.
The algorithm is developed with efficient methods of molecular
structure encoding and novel potential energy functions. nCRO-
UCRA achieves better performance in terms of residual energy, life-
time, number of alive nodes and convergence rate compared to
CRO-UCRA.
3.2.3.7. An Unequal Multi-hop Balanced Immune Clustering protocol
for wireless sensor networks (UMBIC). UMBIC uses Unequal Cluster-
ing Mechanism (UCM) and Multi-Objective Immune Algorithm
(MOIA) for adjusting the energy dissipation of inter-clusters and
intra-clusters (Nabil and Abo-Sahhad, 2016). UCM organizes the
clusters of various size based on the distance to BS and residual
energy. MOIA produces optimal clusters and routing tree is con-
structed for covering the entire region (Lu, 2009). It ensures con-
nectivity and low communication cost among all nodes. CH
rotation takes place when the present CH residual energy becomes
lesser than the threshold energy value. UMBIC avoids the hot spot
problem with less overhead and computational complexity.
3.2.4. Compound unequal clustering algorithm
3.2.4.1. Energy Degree Distance Unequal Clustering Algorithm
(EDDUCA). EEDUCA algorithm is proposed to approximate the
equalization of energy consumption. It improves the network life-
time and also eliminates the hot spot problem (Guiloufi et al.,
2016). There are three phases in EDDUCA: cluster formation phase,
CH selection phase and data transmission phase. In cluster forma-
tion phase, ‘Sierpinski’ triangle divides the network into unequal
clusters. The size of the clusters is same when the distance from
CH to the BS is same. In CH selection phase, CHs are selected based
on residual energy, node degree and distance to BS. A node weight
is calculated based on the above three parameters in each cluster
and the node with minimal weight is elected as CH. Each node for-
wards data to CH and CH transmits to its upper cluster until it
reaches the BS. It performs well and lengthens the lifetime com-
pared to ECLEACH.
3.2.4.2. Energy-Efficient Routing Algorithm Based on Unequal Cluster-
ing and Connected Graph in Wireless Sensor Networks (UCCGRA).
UCCGRA (Xia et al., 2016) is a distributed approach which improves
the energy efficiency in two methods: cluster head election and
cluster routing. Voting scheme is used to construct unequal size
clusters and smaller clusters are constructed near BS to reduce
intra-cluster traffic and results to the elimination of hot spot prob-
lem. CH selection is mainly based on topology, residual energy and
transmission power. For inter-cluster multi-hop communication,
Connected graph based routing utilizes the geographic location of
the nodes. UCCGRA distributes the load effectively, reduces energy
consumption. It achieves maximum network lifetime than UCRA,
EEUC and HEED.
3.3. Preset clustering algorithm
In Preset clustering algorithm, the clusters or CHs and the loca-
tion of CHs are predetermined before deployment into the real
world. Major drawbacks of this approach are: static and network
conditions are not considered. In WSN, wireless link failures/ node
failures leads to network topology changes frequently. These algo-
rithms are not suitable for real time application.
3.3.1. Unequal clustering scheme (UCS)
The aim of UCS is maximizing the network lifetime by equal
amount of energy dissipation in all CHs (Soro and Heinzelman,
 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
2005b). The position of CHs are predetermined before deployment
and arranged symmetrically in concentric circles with BS at the
centre. Hence, it comes under the preset algorithm. The cluster
consists of nodes in Voronoi region with BS at the centre. UCS is
a two layered network with the clusters in same layer are equal
in shape and size and differ in other layer. The CH transmits the
data to BS via two hops. UCS can be more effective for networks
which involves large amount of data transmission. It achieves
10–30% improvement when compared to equal clustering schemes
based on data aggregation at the CHs. The drawback is the diffi-
culty of using in real time application due to the predefined
assumptions made.
4. Comparison and discussion
A comparison of various reviewed unequal clustering
approaches is tabulated in the Table 1. In the beginning, various
algorithms are compared based on cluster properties which
include cluster size (equal or unequal), cluster count (fixed or vari-
able), intra-cluster communication and inter-cluster communica-
tion (direct or multi-hop). Next, various algorithms are compared
based on cluster head properties incorporating mobility (station-
ary or mobile), node type (homogeneous or heterogeneous), role
(relay or aggregating data), method (probabilistic, deterministic
and preset) and objectives (load balancing, avoid hot spot problem,
lengthen network lifetime, etc.) of the respective algorithms.
Finally, various algorithms are compared based on clustering pro-
cess which includes the way of CH selection, nature (proactive or
reactive) and location awareness.
Basically, unequal clustering protocols are classified in three
ways: Probabilistic, deterministic and preset type. The probabilis-
tic algorithms are more popular because of its simplicity, energy-
efficient and fast convergence. It is divided into two types namely
random and hybrid. Random approaches are simple and achieved
near optimal overhead but fail to conserve energy. Hybrid methods
use random methods with some parameters like residual energy or
distance to BS. Hybrid approaches are iterative based or competi-
tion based which increases the complexity in terms of message
and time. Deterministic approaches are non-probability based
methods and uses some metrics for CH selection. It is more reliable
and controllable than probabilistic methods. It is not suitable for
large scale WSN because of its complexity and slower convergence.
It is classified into four categories: weight based fuzzy based,
heuristic based and compound methods. Weight based protocols
are iterative based which increases the message complexity. Fuzzy
approaches consume more energy for algorithm execution and
message exchange. Heuristic approaches needs global information
of the network and controlled by the BS. The centralized and
lengthy process makes heuristic methods impractical for many
applications. As a result, there is a trade-off between provisioning
simplicity and Quality of Service (QoS) requirements of the net-
work. The probabilistic methods are simple and fast convergence;
it performs well in large scale WSNs like environmental monitor-
ing. For more reliable and robust applications, deterministic algo-
rithms can be used. Achieving optimal solution in application
specific environment, heuristic approach is the better option.
Although the existing approaches are reviewed well, some parts
of clustering are not investigated yet.
Most of the clustering approaches are static and they do not
have ability to adapt network changes. So, dynamic clustering
can be concentrated in future. Next, the mobility of the network
is not considered. In clustered WSN, three parts can be mobile:
cluster members, CH and BS. WSN with mobility leads to topology
changes frequently and increases overhead. As WSN are generally
data-centric, data-centric routing techniques are needed. Most of
313
the algorithms are developed for proactive networks and less num-
ber of techniques is available for reactive networks. In future, clus-
tering methods for reactive networks can be investigated more.
Various unequal clustering protocols are explained in this
paper. For the sake of simplicity, results of few protocols are dis-
cussed here. The comparative results for various protocols in terms
of energy consumption and network lifetime are shown in Figs 4–
9. Fig 4 shows the network lifetime of PRODUCE protocol com-
pared with EEUC for a network size of 100 m. When the number
of sensor nodes is 400, both the coverage time and lifetime for
PRODUCE are slightly longer than EEUC. The coverage time and
the lifetime for EEUC and PRODUCE are 806, 819 and 838, 905
respectively. PRODUCE achieves longer lifetime because of less
communication overhead and the CH assignment is driven by not
only the distance but also probability and therefore, is powerful
for densely deployed network.
Fig 5a, b shows the energy consumption and network lifetime of
EEUC when compared with HEED and LEACH. First, the amount of
energy spent by cluster heads in three algorithms is compared. 15
rounds of simulations are sampled and the amount of total energy
spent by all cluster heads is shown in Fig. 5a. The energy consumed
by cluster heads per round in EEUC is much lower than that in
LEACH, and is about the same as that in HEED. Because cluster
heads send their packets to the BS directly in LEACH, the energy
consumption is much higher. In EEUC and HEED, cluster heads
transmit their data to the BS via multi-hop, thus a considerable
amount of energy is saved. Next, the energy efficiency of three
algorithms is analyzed by examining the network lifetime. Fig. 5b
shows the number of sensor nodes still alive over the simulation
time.
EEUC clearly improves the network lifetime (both the time until
the first node dies and the time until the last node dies) over
LEACH and HEED. In HEED, tentative cluster heads are randomly
selected based on their residual energy. Therefore, sensors with
low residual energy can still become cluster heads since it uses
the intra-cluster communication cost to select the final cluster
heads. Thus some nodes die too earlier in HEED. This is avoided
in EEUC because energy consumption is well balanced among
nodes. The small interval between the time until the first node dies
and the time until the last node dies implies that EEUC has success-
fully solved the hot spots problem.
Fig. 6a gives the amount of the energy consumed by three algo-
rithms for 15 randomly selected rounds. It shows that EBCAG bal-
ances the energy consumption among cluster heads best, and
HEED performs worst. The variance of EBCAG is very steady due
to the stability of the unequal clustering mechanism based on
the gradient routing. EEUC is also an unequal clustering mecha-
nism, but it does not give the optimal value for balancing the
energy consumption. It performs worse than EBCAG. Fig. 6b shows
the number of alive sensor nodes over the simulation time. With
the given time interval, the number of alive sensor nodes in EBCAG
is more than those in EEUC and HEED. It testifies that the energy
consumed by all nodes in EBCAG is much lower than HEED and
EEUC.
Fig. 7a gives the standard deviation of residual energy among all
nodes, which reflects how well the three protocols balance energy
dissipation of network. It shows that MRPUC performs best, and
HEED is worst. Fig. 7b illuminates total number of nodes alive over
the time, which indicates the lifetime of network. MRPUC performs
much better than the other two protocols. In HEED, cluster heads
farther away from BS have to transmit data over longer distances
to BS, which consumes the large energy and lead to the earlier
death of the nodes farther away from BS. In MRPEC, the cluster
heads closer to BS will die much faster than the other cluster heads
farther away from BS. MRPUC avoids the hot spots problem per-
fectly, so it can extend the network lifetime. In addition, smaller
 314
Table 1
Comparison of various unequal clustering approaches.

Protocol Cluster properties CH properties Clustering process

Unequal Variable Single hop Multi hop Mobility Node Role Relay/ Method Objective Load CH election Proactive Loca.

S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
Cluster Cluster Intra Inter typeHomogeneous aggregation Distributed balancing & Nature aware
size count comm. comm. maximum
lifetime
PRODUCE (Kim et al., 2008) U U U U  U U U U Random U 
EDUC (Yu et al., 2011a) U U U U   U U U Random U 
LUCA (Lee et al., 2011) U U U U  U U U U Random U U
EEUC (Li et al., 2005) U U U U  U U U U Hybrid U 
EEDUC (Yu et al., 2011b) U U U U  U U U U Hybrid U 
UCR (Chen et al., 2009) U U U U  U U U U Hybrid U 
Unequal LEACH (Ren et al., 2010) U U U   U U U U Hybrid U 
EC (Wei et al., 2011) U U U U  U U U U Hybrid U 
EB-UCP (Yang, 2009) U U U U  U U U U Hybrid U 
UHEED (Ever et al., 2012) U U U U  U U U U Hybrid U 
COCA (Li et al., 2013) U U U U  U U U U Hybrid U 
MRPUC (Gong et al., 2008) U U U U  U U U U Deterministic U 
PEBECS (Wang and Yang, 2009) U U U U  U U U U Deterministic U 
EADUC (Yu et al., 2011b) U U U U  U U U U Deterministic U 
ACT (Lai et al., 2012) U U U U  U U U U Deterministic U 
EBCAG (Liu et al., 2012) U U U U  U U U U Deterministic U 
EUCS (Messai, 2013) U U U U  U U U U Deterministic U 
UCMR (Jinwala et al., 2008) U U UU U  U U U U Deterministic U 
EPUC (Bagci et al., 2010) U U U U  U U U U Deterministic U 
CUCA (Mazumdar and Om, 2015) U U U   U U U U Deterministic U U
Improved EADUC (Gupta and Pandey, U U U U  U U U U Deterministic U 
2016)
EAUCF (Bagci and Yazici, 2010) U U U U  U U U U Fuzzy U 
IFUC (Mao et al., 2012) U U U U  U U U U Fuzzy U 
FUCP (Gajjar and Talati, 2015) U U U U  U U U U Fuzzy U 
FBUC (Logambigai and Kannan, 2016) U U U U  U U U U Fuzzy U 
DUCF (Baranidharan and Santhi, 2016) U U U U  U U U U Fuzzy U 
EBUC (Jiang et al., 2010) U U U U  U U  U Heuristic U 
GAEEP (Abo-zahhad et al., 2014) U U U U  U U  U Heuristic U U
SMEBUC (Fan and Du, 2015) U U U U  U U  U Heuristic U 
IPSO (Salehian and Subraminiam, 2015) U U U U  U U  U Heuristic U 
FAMACROW (Gajjar et al., 2014) U U U U  U U  U Heuristic U 
nCRO-UCRA (Srinivasa and Banka, 2016) U U U U   U U U Heuristic U 
UMBIC (Sabil et al., 2016) U U U U  U U U U Heuristic U 
EDDUCA (Guiloufi et al., 2016) U U U U  U U U U Compound U 
UCCGRA (Xia et al., 2016) U U U U  U U U U Compound U 
UCS (Soro and Heinzelman, 2005b) U U U U  Both U U U Preset U U
 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317 315

450 the interval between the dead time of the first node and the dead
No of Alive Nodes 400 EEUC time of the last node, the performance of a protocol will be better.
350 PRODUCE It shows MRPUC achieves 34.4% improvement over MRPEC and
300 251.7% improvement over HEED in the dead time of the first node.
250 The performance of fuzzy based clustering algorithm IFUC is
200
compared with LEACH and EEUC. In Fig. 8a, 20 rounds are ran-
150
100 domly taken and compared the amount of energy consumed by
50 all cluster heads in the three algorithms, The energy consumed
0 by cluster heads in IFUC per round is much lower than in LEACH,
0 200 400 800 880 1000 and is a little lower than in EEUC. The experiment results indicate
Time (s) that the election of cluster heads and the decision of cluster size
are more reasonable due to the use of fuzzy logic.
Fig. 4. Network lifetime of probability based clustering protocol PRODUCE and
EEUC.

450 LEACH
0.3 EEUC
Energy consumption of CHs

400
EEUC
HEED
0.25 350

Number of alive nodes

LEACH HEED
300
0.2
250
0.15 200
150
0.1
100
0.05 50
0
0 0 200 400 600 800
0 2.5 5 7.5 10 12.5 15 Rounds
Rounds
(a) (b)

Fig. 5. Comparative results of Hybrid Clustering protocol EEUC with HEED and LEACH. Energy consumption of EEUC, HEED and LEACH (b) Network lifetime of EEUC, HEED
and LEACH.

0.25 700

HEED 600
Number of alive sensor nodes
Number of alive nodes

0.2 EEUC
EEUC
500 HEED
EBCAG
0.15 EBCAG
400

0.1 300

200
0.05
100

0 0
0 2.5 5 7.5 10 12.5 15 0 100 200 300 400 500 600 700 800

Rounds Rounds
(a) (b)

Fig. 6. Comparative results of deterministic weight based clustering protocol EBCAG with EEUC and HEED (a) Energy consumption of EBCAG, EEUC and HEED (b) Network
lifetime of EBCAG, EEUC and HEED.

0.45 MRPUC 600

MRPUC
Standard deviation of residual energy

0.4 MRPEC MRPEC

500
Number of alive nodes

0.35 HEED HEED

0.3 400

0.25
300
0.2

0.15 200

0.1
100
0.05

0 0
0 50 100 150 200 250 300 350
20 40 60 80 100 120 140 160 180 200
Rounds Rounds
(a) (b)

Fig. 7. Comparative results of deterministic weight based clustering protocol MRPUC with MRPEC and HEED (a) Energy consumption of MRPUC, MRPEC and HEED (b)
Network lifetime of MRPUC, MRPEC and HEED.
316 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317

0.35
LEACH
400

Energy consumption of CHs/ J

EEUC
0.3 EEUC
IFUC 350

Number of alive nodes

IFUC
0.25 300
LEACH
0.2
250
200
0.15
150
0.1 100
0.05
50
0
0 0 200 400 600 800 100012001400
0 1 2 3 4 5 6 7 8 9 1011121314151617181920
Rounds
Rounds
(a) (b)

Fig. 8. Comparative results of fuzzy based clustering protocol IFUC with EEUC and LEACH a) Energy consumption of IFUC, EEUC and LEACH (b) Network lifetime of IFUC, EEUC
and LEACH.

60
LEACH LEACH
120
50 EBUCP EBUCP
Remaining energy /J

Number of alive nodes

SMEBUC 100 SMEBUC
40
80
30
60
20
40

10 20

0 0
0 100 200 300 400 500 600 700 800 0 50 100 150 200
Time (Rounds) Time ( Rounds)
(a) (b)

Fig. 9. Comparative results of heuristic based clustering protocol SMEBUC with EBUCP and LEACH a) Energy consumption of SMEBUC, EBUCP and LEACH (b) Lifetime of
SMEBUC, EBUCP and LEACH.

The simulation result in Fig. 9a shows that the remaining References

energy of SMEBUC is higher than LEACH and EBUCP. Fig. 9b shows
that the network lifetime of SMEBUC is much longer than that of Abo-zahhad, M., Ahmed, S.M., Sabor, N., 2014. A new energy-efficient adaptive
clustering protocol based on genetic algorithm for improving the lifetime and
the other two protocols. This is due that SMEBUC adopts the the stable period of wireless sensor networks. Int. J. Energy Inf. Commun. 5, 47–
method of centralized control. The sink node performs clustering 72. http://dx.doi.org/10.14257/ijeic.2014.5.3.05.
and determines the cluster head. SMEBUC uses the unequal clus- Akyildiz, I.F., Su, W., Sankarasubramaniam, Y., Cayirci, E., 2002. Wireless sensor
networks: a survey. Comput. Networks 38, 393–422. http://dx.doi.org/10.1016/
tering and multi-hop routing, and the Sink node is mobile in a pre- S1389-1286(01)00302-4.
determined region. SMEBUC can effectively balance network Anastasi, G., Conti, M., Di Francesco, M., Passarella, A., 2009. Energy conservation in
energy consumption which makes the balance of energy and better wireless sensor networks: a survey. Ad Hoc Netw. 7, 537–568. http://dx.doi.org/
10.1016/j.adhoc.2008.06.003.
network lifetime of WSN.
Bagci, H., Yazici, A., Member, S., 2010. An energy and proximity-based unequal
clustering algorithm for wireless sensor networks. In: 39th Annu. IEEE Conf.
5. Conclusion Local Comput. Networks. 18–23.
Bagci, H., Yazici, A., 2010. An energy aware fuzzy unequal clustering algorithm for
wireless sensor networks. IEEE Int. Conf. Fuzzy Syst., 1–8 (FUZZ)
Clustering is the popular energy efficient technique but it suf- Baranidharan, B., Santhi, B., 2016. DUCF: distributed load balancing unequal
fers from hot spot problem and minimizes the network lifetime clustering in wireless sensor networks using fuzzy approach. Appl. Soft
Comput. J. 40, 495–506. http://dx.doi.org/10.1016/j.asoc.2015.11.044.
significantly. Unequal clustering evenly distributes the load, elim- Chen, G., Li, C., Ye, M., Wu, J., 2009. An unequal cluster-based routing protocol in
inates the hot spot problem and maximizes the network lifetime. wireless sensor networks. Wirel. Networks 15, 193–207. http://dx.doi.org/
In this paper, various unequal clustering protocols are classified 10.1007/s11276-007-0035-8.
Dorigo, M., Gambardella, L.M., 1997. Ant colony system: a cooperative learning
into three main categories: probabilistic, deterministic and preset
approach to the traveling salesman problem. IEEE Trans. Evol. Comput. 1, 53–
clustering algorithms. These protocols are explained with their 66. http://dx.doi.org/10.1109/4235.585892.
objectives, characteristics, classification, merits and demerits. The Estrin, D., Heidemann, J., Kumar, S., Rey, M., 1999. Next Century Challenges: Scalale
Coordination in Sesnor Networks. Admiralty Way. Inf. Sci. (Ny).
probabilistic methods are simple and faster convergence; it per-
Ever, E., Luchmun, R., Mostarda, L., Navarra, A., Shah, P., 2012. UHEED – an unequal
forms well in large scale WSNs like environmental monitoring. clustering algorithm for wireless sensor networks. Sensornets.
For more reliable and robust applications, deterministic algorithms Fan, X., Du, F., 2015. Shuffled frog leaping algorithm based unequal clustering
can be used. Achieving optimal solution in application specific strategy for wireless sensor networks. Appl. Math. Inf. Sci. 1426, 1415–1426.
Gajjar, S., Sarkar, M., Dasgupta, K., 2014. ScienceDirect FAMACRO : fuzzy and ant
environment, heuristic approach is the better option. All the colony optimization BASED MAC/routing cross-layer protocol for wireless
reviewed protocols are compared based on different cluster prop- sensor networks. Appl. Soft Comput. J., 235–247
erties, Cluster Head (CH) properties and clustering process. Gajjar, S., Talati, A., 2015. FUCP: fuzzy based unequal clustering protocol for
wireless sensor networks. In: 39th Natl. Syst. Conf. 1–6.
 S. Arjunan, S. Pothula / Journal of King Saud University – Computer and Information Sciences 31 (2019) 304–317
Kaur, Gurmeet, Varsha, 2016. Review on hierarchical unequal clustering based
protocols in Wireless Sensor Network. Int. J. Rec. Innov. Comput. Commun. 4,
96–99.
Gong, B., Li, L., Wang, S., Zhou, X., 2008. Multihop routing protocol with unequal
clustering for wireless sensor networks. In: Proc. – ISECS Int. Colloq. Comput.
Commun. Control. Manag. CCCM 2008 2, 552–556. doi:http://dx.doi.org/10.
1109/CCCM.2008.99.
Guiloufi, A.B.F., Nasri, N., Kachouri, A., 2016. An energy-efficient unequal clustering
algorithm using ‘‘Sierpinski Triangle” for WSNs. Wirel. Pers. Commun. 88, 449–
465. http://dx.doi.org/10.1007/s11277-015-3137-0.
Gupta, V., Pandey, R., 2016. An improved energy aware distributed
unequal clustering protocol for heterogeneous wireless sensor networks.
Eng. Sci. Technol. Int. J. 19, 1050–1058. http://dx.doi.org/10.1016/j.jestch.
2015.12.015.
Heinzelman, W.B., Chandrakasan, A.P., Balakrishnan, H., 2002. An application-
specific protocol architecture for wireless microsensor networks. IEEE Trans.
Wirel. Commun. 1, 660–670. http://dx.doi.org/10.1109/TWC.2002.804190.
Heinzelman, W.R., Chandrakasan, A., Balakrishnan, H., 2000. Energy-efficient
communication protocol for wireless microsensor networks. In: Proc. 33rd
Annu. Hawaii Int. Conf. Syst. Sci. 0, 3005–3014. doi:http://dx.doi.org/10.1109/
HICSS.2000.926982.
Hou, Y.T., Shi, Y., Sherali, H.D., Midkiff, S.F., 2005. On energy provisioning and relay
node placement for wireless sensor networks. IEEE Trans. Wirel. Commun. 4,
2579–2590. http://dx.doi.org/10.1109/TWC.2005.853969.
Jiang, C.J., Shi, W.R., Xiang, M., Tang, X.L., 2010. Energy-balanced unequal clustering
protocol for wireless sensor networks. J. China Univ. Posts Telecommun. 17, 94–
99. http://dx.doi.org/10.1016/S1005-8885(09)60494-5.
Jinwala, D.C., Patel, D.R., Dasgupta, K.S., 2008. for Wireless Sensor Networks I, 2–6.
doi:http://dx.doi.org/10.1007/s11277-013-1052-9.
Kim, J.H., Hussain, C.S., Yang, W.C., Kim, D.S., Park, M.S., 2008. Produce. A
probability-driven unequal clustering mechanism for wireless sensor
networks. Proc. Int. Conf. Adv. Inf. Netw. Appl. http://dx.doi.org/10.1109/
WAINA.2008.116.c. AINA 928-933.
Kleinrock, L., Kamoun, F., 1977. Hierarchical routing for large networks:
performance evaluation and optimization. Comput. Networks 1, 155–174.
http://dx.doi.org/10.1016/0376-5075(77)90002-2.
Krishnamachari, L., Krishnamachari, L., Estrin, D., Estrin, D., Wicker, S., Wicker, S.,
2002. The impact of data aggregation in wireless sensor networks. In: Proc.
22nd Int. Conf. Distrib. Comput. Syst. Work. 575578, 575–578. doi:http://dx.doi.
org/10.1109/ICDCSW.2002.1030829.
Lai, W.K., Fan, C.S., Lin, L.Y., 2012. Arranging cluster sizes and transmission ranges
for wireless sensor networks. Inf. Sci. (Ny) 183, 117–131. http://dx.doi.org/
10.1016/j.ins.2011.08.029.
Lee, S., Choe, H., Park, B., Song, Y., Kim, C.K., 2011. LUCA: an energy-efficient unequal
clustering algorithm using location information for wireless sensor networks.
Wirel. Pers. Commun. 56, 715–731. http://dx.doi.org/10.1007/s11277-009-
9842-9.
Li, C., Ye, M., Chen, G., Wu, J., 2005. An energy-efficient unequal clustering
mechanism for wireless sensor networks. In: 2nd IEEE Int. Conf. Mob. Ad-hoc
Sens. Syst. MASS 2005 2005, 597–604. doi:http://dx.doi.org/10.1109/MAHSS.
2005.1542849.
Li, H., Liu, Y., Chen, W., Jia, W., Li, B., Xiong, J., 2013. COCA: constructing
optimal clustering architecture to maximize sensor network lifetime.
Comput. Commun. 36, 256–268. http://dx.doi.org/10.1016/j.comcom.
2012.10.006.
Liu, T., Li, Q., Liang, P., 2012. An energy-balancing clustering approach for gradient-
based routing in wireless sensor networks. Comput. Commun. 35, 2150–2161.
http://dx.doi.org/10.1016/j.comcom.2012.06.013.
Logambigai, R., Kannan, A., 2016. Fuzzy logic based unequal clustering for wireless
sensor networks. Wirel. Networks 22, 945–957. http://dx.doi.org/10.1007/
s11276-015-1013-1.
Lu, H., 2009. An adaptive multi-objective immune optimization algorithm. Int. Conf.
Control Autom. Syst. Eng., 140–143
317
Mao, S., Zhao, C., Zhou, Z., Ye, Y., 2012. An improved fuzzy unequal clustering
algorithm for Wireless Sensor Network. Mob. Networks Appl. 206–214. http://
dx.doi.org/10.1007/s11036-012-0356-4.
Mazumdar, N., Om, H., 2015. Coverage-aware unequal clustering algorithm for
Wireless Sensor Networks. Procedia Comput. Sci. 57, 660–669. http://dx.doi.
org/10.1016/j.procs.2015.07.437.
Messai, M., 2013. New Clustering Scheme for Wireless Sensor Networks. 75, 487–
491.
Na, P., Vergados, D.D., 2007. A survey on power control issues in wireless sensor
networks. IEEE Commun. Surv. Tutorials 9, 89–107.
Nabil, Sabor, Abo-Sahhad, Mohammad, 2016. An Unequal Multi-hop BAlanced
Immune Clustering protocol for wireless sensor networks. Appl. Soft Comput. J.
43, 372–389. http://dx.doi.org/10.1016/jasc.2016.02.016.
Raghavendra, C.S., Sivalingam, K.M., Znati, T., 2004. Wireless Sensor Networks.
Springer, US.
Ren, P., Qian, J., Li, L., Zhao, Z., Li, X., 2010. Unequal clustering scheme based LEACH
for wireless sensor networks. In: Proc. – 4th Int. Conf. Genet. Evol. Comput.
ICGEC 2010 90–93. doi:http://dx.doi.org/10.1109/ICGEC.2010.30.
Salehian, S.K., Subraminiam, S., 2015. Unequal clustering by improved particle
swarm optimization in Wireless Sensor Network. Procedia Comput. Sci. 62,
403–409. http://dx.doi.org/10.1016/j.procs.2015.08.433.
Sohraby, K., Minoli, D., Znati, T., 2007. Wireless Sensor Networks. Booksgooglecom.
http://dx.doi.org/10.1002/047011276X.
Soro, S., Heinzelman, W.B., 2005a. Prolonging the Lifetime of Wireless Sensor
Networks via Prolonging the Lifetime of Wireless Sensor Networks via. In: 19th
IEEE Int. Parallel Distrib. Process. Symp. doi:http://dx.doi.org/10.1109/IPDPS.
2005.365.
Soro, S., Heinzelman, W.B., 2005b. Prolonging the lifetime of wireless sensor
networks via unequal clustering. In: Proc. – 19th IEEE Int. Parallel Distrib.
Process. Symp. IPDPS 2005 2005. doi:http://dx.doi.org/10.1109/IPDPS.2005.365.
Srinivasa Rao, P.C., Banka, H., 2016. Novel chemical reaction optimization based
unequal clustering and routing algorithm for Wireless Sensor Networks (nCRO-
UCRA). Wirel. Network. http://dx.doi.org/10.1007/s11276-015-1148-0.
Vennira Selvi, G., Manoharan, R., 2013. A survey of energy efficient unequal
clustering algorithms for wireless sensor networks. Int. J. Comput. Appl. 79.
0975-8887.
Wang, Y., Yang, T.Z.D., 2009. An energy efficient and balance hierarchical unequal
clustering algorithm for large-scale sensor networks. Inf. Technol. J. 8, 28–38.
Wei, D., Jin, Y., Vural, S., Moessner, K., Tafazolli, R., 2011. An energy-efficient
clustering solution for wireless sensor networks. IEEE Trans. Wirel. Commun.
10, 3973–3983. http://dx.doi.org/10.1109/TWC.2011.092011.110717.
Xia, H., Zhang, R-hua, Yu, J., Pan, Z-kuan, 2016. Energy-efficient routing algorithm
based on unequal clustering and connected graph in wireless sensor networks.
Int. J. Wirel. Inf. Networks 23, 1–10. http://dx.doi.org/10.1007/s10776-016-
0304-5.
Yang, T.Z.D., 2009. An energy-balancing unequal clustering protocol for wireless
sensor networks. Inf. Technol. J. 8, 57–63.
Younis, M., Youssef, M., Arisha, K., 2003. Energy-aware management for cluster-
based sensor networks. Comput. Networks 43, 649–668. http://dx.doi.org/
10.1016/S1389-1286(03)00305-0.
Younis, O., Fahmy, S., 2004. HEED: A hybrid, energy-efficient, distributed clustering
approach for Ad Hoc sensor networks. IEEE Trans. Mob. Comput. 3, 660–669.
Yu, J., Qi, Y., Wang, G., 2011a. An energy-driven unequal clustering protocol for
heterogeneous wireless sensor networks. J. Control Theory Appl. 9, 133–139.
http://dx.doi.org/10.1007/s11768-011-0232-y.
Yu, J., Qi, Y., Wang, G., Guo, Q., Gu, X., 2011b. An energy-aware distributed unequal
clustering protocol for wireless sensor networks. Int. J. Distrib. Sens. Networks
2011. http://dx.doi.org/10.1155/2011/202145.
Zhu, G., 2009. Meme triangular probability distribution shuffled frog-leaping
algorithm. Comput. Integr. Manuf. Syst. 15, 1979–1985.
Zhou, Y., Fang, Y., Zhang, Y., 2008. Securing wireless sensor networks: a survey. IEEE
Commun. Surv. Tutorials 10, 6–28. http://dx.doi.org/10.1109/
COMST.2008.4625802.