You are on page 1of 20

Wireless Networks (0123456789().,-volV)(0123456789().,-volV)

Optimization energy consumption with multiple mobile sinks using

fuzzy logic in wireless sensor networks
Kambiz Koosheshi1,2 • Saeed Ebadi2

 Springer Science+Business Media, LLC, part of Springer Nature 2018

Each sensor in WSNs receives data from the limited area under its coverage. The received data is processed by the sensor;
then, it is wirelessly transmitted to the sink. Sensors’ energy consumption and the energy hole problem are considered as
outstanding challenge of these networks. That is, since sensors use batteries with limited lifetimes and sensors which are
closer to the sink transmit more data than remoter sensors, hence, they run up their energies sooner than other sensor nodes.
Consequently, optimizing energy consumption is regarded as one of the most critical issues throughout a network’s
operational lifetime. In this paper, by dividing a respective area into several smaller areas and using a multiple mobile sink
(MS), we proposed an unequal clustering method via fuzzy logic which leads to the reduction of distances among sensors
with respect to the movement direction of sink. As a result, the sizes of clusters are reduced. Accordingly, such a reduction
in the sizes of clusters and the smart selection of the route by the MS eliminates energy hole issue. Regarding the metrics of
FND and HNA, it was found that the proposed method optimized energy consumption for 19%. Hence, it was able to fix
energy hole problem in WSNs.

Keywords Wireless sensor network  Clustering  Energy hole problem  Multiple mobile sink  Fuzzy logic

1 Introduction into smaller groups called clusters [4]. Indeed, each cluster
has a cluster head (CH) which is responsible for coordi-
Wireless sensor networks consist of hundreds or thousands nating and gathering data from other nodes. Sensors con-
of sensors within a region [1]. Each of the sensors is cap- sume energy for transmitting receiving and processing
able of measuring magnetic, acoustic, spatial or seismic data. Since sensors are used via batteries which are neither
data from surrounding environment [2]; by transmitting replaceable nor rechargeable, hence, network lifetime
data to the final user, sensors provide the final users with depends on the remaining energy of the nodes [5]. As a
the chance to take the required measures. In general, result, optimal energy consumption is regarded as one of
wireless sensors are light sensors in small dimensions the main network challenges. In this network, nodes which
which are desirable in terms of costs. They are placed in a are more distant from the sink transmit their data in the
wide region and are able to handle themselves [3]. For relay format from one node to the other node so that data is
desirably gathering data in WSNs, nodes can be divided delivered to the sink node. This procedure is carried out
because nodes are limited in terms of their transmission
power. Consequently, nodes closer to the sink run up their
& Saeed Ebadi energy sooner because they transmit large amounts of data,
this is called energy hole problem or also called as the hot
Kambiz Koosheshi spot problem [6]. So the energy hole problem is created
due to energy depletion of the node which is near the sink.
Department of Computer Engineering, East Azarbaijan In this manner, no more data can be delivered to the sink
Science and Research Branch, Islamic Azad University, after energy hole appears. Consequently, a considerate
Tabriz, Iran amount of energy is wasted and the network lifetime ends
Department of Computer Engineering, Tabriz Branch, prematurely [7].
Islamic Azad University, Tabriz, Iran

Wireless Networks

Wireless sensors consist of four main units: a sensing approach in our algorithm the CH election occurs by
unit, a processing unit, a communicating unit and a power considering three fuzzy descriptors namely distance to RN,
unit [8]. Each sensor nodes has an analog to digital con- residual energy of each node and the density of the node
vertor. That is, a sensor firstly senses data from the sur- with the aim of estimating the competition radius for ten-
rounding environment and produces an analog signal. Next, tative CHs.
the analog to digital convertor converts the produced signal The rest of the paper is organized as follows: in section
into the digital form and delivers it to the processing unit. two, the related works are briefly reviewed. In section
Then, the processed data is delivered to the communication three, the system model is mentioned. In section four, a
unit so that they can be transmitted to other nodes [8, 9]. detailed description of the proposed algorithm is given. In
Also, power unit provides the required energy source. section five, the efficiency of the proposed algorithm is
In clustering WSNs, the nodes available within a cluster investigated and discussed via simulation. Finally, in sec-
directly transmit their data to the related CH [4]. In case the tion six, the conclusion of the study and directions for
CH is adequately close to the sink, it directly transmits the further research are reported.
received data to it. Otherwise, it transmits the data to
another CH which is closer to the sink; hence, its respective
data is delivered to the sink node in the relay format. It can 2 Related works
be argued that CHs consume more energy than the other
nodes of the network. As a result, an energy-aware model WSNs are composed of sets of sensors and a sink which are
is used for balancing CHs’ energy consumption and peri- in a region. Sensors sense data from their surrounding
odically replacing them. However, sensors closer to the environment and transmit them either directly or indirectly
sink lose their energy sooner than the other sensor nodes to the sink [8]. Nevertheless, transmitting data to long
because they are exposed to heavier traffic. For sorting out distances is not optimal in terms of energy consumption.
this problem, thanks to fuzzy logic, we created more Hence, one of the methods for enhancing network lifetime
clusters near the movement direction of the sink so that the is to reduce data transmission distance and use a MS in the
number of CHs near the sink is enhanced and the network network. In this method, by moving within the network, the
traffic is divided among more of them. In the proposed MS gathers data from the sensors [12–18]. However, it
algorithm, at first, tentative CHs are selected from network should be pointed out that using MS requires an optimal
nodes by a probable model. Then, the final CHs are protocol for smartly selecting sink trajectory in the net-
selected from among tentative CHs based on a competition work, see [19–28]. Also, one another optimization method
with regard to the remaining amount of nodes’ energy. is cluster network nodes. In this method, cluster members
In this paper, a method is proposed which can optimize transmit their data to the CH and the CH transmits the
sensor nodes’ energy consumption by means of a multiple gathered data to the sink. A few protocols in relation to MS
MS where fuzzy logic is used in WSNs. In this method, and clustering WSNs are briefly reviewed below.
rather than using a fixed sink in a predetermined place, a LEACH is regarded as the first hierarchical routing
MS is used for gathering sensors’ data in such a way that it protocol in WSNs. It is regarded as the basis for many other
smartly moves within the network and gathers data from hierarchical protocols [29]. In this method, network is
sensors. Also, the rest of the nodes are clustered based on divided into several small sections which are called clus-
fuzzy logic so as to prevent energy hole problem. The ters. Each cluster has its own related CH. In this protocol,
concept of Fuzzy Logic centers around the idea of partial CH is selected regardless of the remaining energy by
set membership, instead of crisp or discreet set membership means of a probable model. In general, LEACH protocol
[10]. Initially, it was introduced as an alternative approach uses a probabilistic method for selecting CH and nodes
to processing data that has behavior defined by a ‘‘fuzzy’’ decide by themselves about the CH to which they belong.
set [11], which contains elements whose degree of mem- COCA is considered to be an unequal clustering algo-
bership vary in the set. Fuzzy Logic is different from rithm in WSNs [3]. In this method, the node with the
conventional control methods by the fact that it incorpo- highest remaining energy among the neighbors is selected
rates simple if–then structure, rather than any complex as the CH. Also, for preventing the occurrence of energy
mathematical model [10]. Consequently, network lifetime hole problem, the more the distance from the sink, the
is enhanced. However, the motivation behind this work is bigger the clusters and vice versa. EAUCF is regarded as a
to overcome the energy hole problem especially in large- distributed competitive algorithm in which the CH is
scale networks by dividing WSN field into z autonomous selected from among tentative CHs based on competition
zones and using a MS in each one. In this work also, we use with regard to the remaining energy by means of fuzzy
fuzzy logic approach in order to handle uncertainties in logic [5]. Regarding the EAUCF protocol, fuzzy logic has
cluster head radius estimation. In the purposed fuzzy two inputs and an output. The inputs are remaining energy

Wireless Networks

and distance to base station and the only output is com- RNs. RNs buffer data in themselves and when the MS
petition radius which uses the Center of area. passes near them, they transmit data to the sink.
MOFCA is another distributed clustering algorithm in In the proposed algorithm, Instead of using a fixed sink
WSNs [30]. This algorithm uses three parameters, i.e. or predetermined MS trajectory the MS smartly moves
residual energy, distance to sink and node density for within the network in each round; at the beginning of
selecting the final CH through fuzzy logic. Similar to round, after the movement path of the MS is determined
EAUCF, tentative CHs will be randomly selected from and RNs are selected by MS, clustering is performed using
among the available nodes in the network. Then, using the distributed unequal protocol by the fuzzy logic. In this
fuzzy if–then mapping rules, the selected tentative CHs work, we determine the optimal trajectory of MS by
will measure their competition radius. It can be argued that dividing the area of interest into 16 equal parts and cal-
MOFCA has been exclusively designed for nonuniform culating average remaining energy of each one. We show
networks. Toloueiashtian et al. [31] proposed a novel that using combination of MS and fuzzy approach prolongs
approach for cluster formation in WSNs that uses fuzzy network lifetime and prevent energy hole problem partic-
logic to enhance network lifetime. The chance of each node ularly in large-scale networks. The proposed method
to be CH is calculated by using a fuzzy inference system. decreases energy consumption in WSNs further than tra-
Following three parameters as the input variables of the ditional algorithms which are using predetermined MS
fuzzy logic: remaining energy, distance to base station and trajectory or equal clustering methods which could easily
connections of CH. That is, these variables are taken into cause energy hole problem.
consideration for measuring this chance.
Multicasting, is one of the most common data trans-
mission patterns in WSNs. Hence, to make it efficient, 3 System model
multicast tree has been proposed, but most of the proposed
algorithms did not consider the length of tree which causes The proposed protocol consists of a number of homoge-
delay in the network. Toward Source Tree (TST) is a novel nous nodes which are randomly distributed in an environ-
distributed algorithm to construct an approximate mini- ment. The respective area is divided into several
mum-length multicast tree for WSNs [32]. ConMap is an autonomous zones (z). As shown in Fig. 1, each zone has
algorithm which successfully optimized the multicast an identical number of sensors and a particular MS. In this
energy in delay-constrained mobile WSNs [33]. Unlike protocol, the MS moves in the network based on the pro-
other protocols, Fu et al. proposed an algorithm which posed algorithm and gathers data from network nodes.
optimizes both the transmitting and receiving energy. Before discussing the proposed protocol in details,
Another method for enhancing network lifetime is to use system assumptions which have been considered in the
a MS. MS will cover and rectify the shortcomings of the valuation are mentioned as follows:
fixed sink and will lead to the enhancement of network
• All nodes are identical and deployed randomly.
lifetime, due to this, MAAM is a protocol for balancing
• MS consists of a mobile robot with an unlimited energy
energy consumption of entire network and cover the energy
hole problem in which mobile sink node is used to operate
• After being placed in the area, sensors are not fixed.
the routing process considering the shortest path between
However, this mobility is not tantamount to a lot of
multiple unequal clusters with reduced energy [34]. In this
movements. Rather, it refers to little movement which
method, the nodes are clustered by using K-medoids
is caused by natural sources such as erosion or external
algorithm. One of the algorithms in which MS was used in
its designing is referred to as optimizing leach clustering
• Since the movement of nodes is carried out by means of
algorithm with mobile sink and rendezvous nodes [8]. In
external factors, energy is not consumed.
this protocol, MS will go through a predetermined path
• At the distribution time, all sensor nodes have the same
within the network. Like LEACH, this algorithm consists
amount of energy levels and this battery-power is
of several rounds where there are three types of nodes in
initially 0.25 J same as [29, 39].
each round, rendezvous nodes (RN), CHs and normal
• Each sensor is able to adjust the strength of its
transmission signal according to its distance from the
It is clear that the MS cannot meet all the nodes of a
network and gather their information. Hence, it led to the
• The distance between two sensors is measured in terms
formation of the idea of rendezvous point (RP). RP is
of the reception signal strength (RSSI). Hence, sensors
regarded as a point near the movement route of the MS
need not be informed about each other’s placement
[35–38]. Nodes which are near this point are considered as

Wireless Networks

Fig. 1 The large-scale network

field with six autonomous zones

• MS is aware of the locations of all the nodes. node locally decides about tentative CH, competition
• Each node is aware of its location. radius and the final CH. As a result, a specific node such as
sink is not required for carrying out CH selection process.
The proposed algorithm is divided into several rounds
The proposed algorithm considers the following three
where each round begins with the set-up phase and con-
parameters as the input variables of the fuzzy logic: dis-
tinues with the steady-state phase. The set-up phase
tance to closest RN, remaining energy and calculated
includes three stages. The first stage is involved with the
density. That is, these variables are taken into consideration
coordination of tasks where the tentative CHs and RNs are
for measuring competition radius of each tentative CH.
selected. Since RNs are selected with respect to the
Indeed, a probabilistic model is used in the proposed
movement path of the MS, in the next stage, the ultimate
algorithm for selecting the tentative CH and a rotational
CHs are selected from among the chosen tentative CHs and
system is implemented based on profanity. After sensors
clustering is carried out based on fuzzy logic. The next
are placed in the respective location, the MS forms the first
stage is related to scheduling in which the CH announces
round via a broadcast message. At the outset of each round,
its members about the appropriate scheduling for data
all the nodes are normal node and since the sink is aware of
transmission through a broadcast message; accordingly,
the placement location of all the nodes, it measures the
each node should transmit its data to the CH at the specified
distances of all nodes from its own movement path which
is given in Eq. (1); then, it introduces the nodes whose
As mentioned above, the second phase is the steady-
distances are less than Rx (the optimal coverage radius of
state phase in which each node begins to transmit its data to
RN) as RNs and the rest of the network remains in the
the CH and, finally, to the MS. This phase consist of one
normal node state. The sink transmits the placement loca-
stage which is data transmission.
tion and IDs of RNs to the entire network via the broadcast
message. Hence, each normal node measures its distance
from the entire RNs; finally, it selects the closest RN for
4 The proposed algorithm
In designing the proposed algorithm, two important factors jax0 þ by0 þ cj
jABj ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð1Þ
were taken into consideration. Firstly, the algorithm should a 2 þ b2
optimize network efficiency and energy consumption.
After the placement of the nodes in the first round, the MS
Secondly, it should be light enough so that it can be
will pass through a predetermined path since all the nodes
implemented on real hardware. In the algorithm proposed
have identical initial energy. Different stages of imple-
in this paper, a distributed method based on the fuzzy logic
menting the proposed protocol are given in Algorithm 1.
was used for unequal clustering according to which each

Wireless Networks

Wireless Networks

When RNs are selected and the distance of each node a smaller competition radius so as to reduce CH coverage.
from the nearest RN is determined, each normal node will In general, in the proposed protocol, as the distance of the
select a random number between 0 and 1. In case this tentative CH becomes closer to the movement path of the
number is smaller than the threshold (Th) which is regar- MS the competition radius will be smaller. In other words,
ded as a predetermined value, hence, that node will change if there are higher number of final CHs near the MS tra-
its state from normal node to the tentative CH. Now, using jectory, there will be more nodes near which will function
three parameters of distance to the closest RN, remaining as relays; consequently, they will transmit the remaining
energy and calculated density, the tentative CHs measure information of the network so that the energy hole problem
and form their competition radius. can be sorted out in this way.
Since the proposed protocol uses the above-mentioned The density of the tentative CH (di) is measured through
three parameters for measuring competition radius, hence, Eq. (2). Each sensor is aware of the number of alive nodes
the competition radius of each tentative CH changes (nodes which have energy) within its coverage radius but
throughout the execution of the proposed protocol. Such a they need to be aware of the total number of alive nodes in
change in the competition radius is quite logical because if the entire network for measuring the amount of density
the competition radius does not change following the (denominator of Eq. 2); as this value is variable, depending
modifications in the input parameters, nodes’ energies upon the energy drainage of nodes at the end of each round,
drain will quickly. As a result, with respect to this the MS broadcasts this value in the entire network as a
notable issue, the competition radius will always change as message.
a function of the changes in the input parameters. Com- Number of Alive Nodes in Radius
petition radius is established and specified using a set of di ¼ ð2Þ
Number of All Alive Nodes in Netwok
predetermined fuzzy rules. This process is depicted in line
3.2 of Algorithm 1. The fuzzy rules are given in Table 1. In The only fuzzy variable of the output is the competition
order to evaluate the rules, the Mamdani Controller is used radius of the tentative CH. The fuzzy set related to this
as a fuzzy inference technique and the center of area output variable is depicted in Fig. 5.
(COA) method is employed for defuzzification of the As given in Table 1, if a particular tentative CH has the
competition radius. highest distance from RN (fuzzy input variable Distance to
For measuring competition radius, three fuzzy input Closest RN has the value of far), its battery is full (fuzzy
variables were used. The first variable is distance to the input variable Remaining Energy has the value of high) and
closest RN. The fuzzy set related to this fuzzy variable is the density of its tentative CH is high (fuzzy input variable
illustrated in Fig. 2. The linguistic variables related to this Calculated Density has the value of dense), hence, it will
input are close, medium and far. For close and far vari- have the highest value of competition radius (fuzzy output
ables, the trapezoidal membership function was used and variable Competition Radius has the value of 12XL). On
for the medium variable, the triangular membership func- the other hand, in case the battery amount of the respective
tion was used. node is running out (fuzzy input variable Remaining
The second fuzzy variable is the remaining energy of the Energy has the value of far the value of low) and the dis-
tentative CH. The fuzzy set related to this fuzzy variable is tance of that node from RN is close (fuzzy input variable
depicted in Fig. 3. Linguistic variables related to this input Distance to closest RN has the value of close) and the
are low, medium and high. For low and high variables, the measured density is high (fuzzy input variable Calculated
trapezoidal membership function and for the medium Density has the value of dense), consequently, the tentative
variable, the triangular membership function was used. CH will have the minimum competition radius (fuzzy
The third fuzzy variable is the density of the tentative output variable Competition Radius has the value of
CH. The fuzzy set describing this input variable is shown in 12XS). These are given in Table 1. The remaining states
Fig. 4. The linguistic variables related to this input are are included between the above-mentioned two states.
sparse, normal and dense. As shown in the figure, sparse Maximum competition radius is a fixed parameter which
and dense have trapezoidal membership functions and is transmitted by the MS throughout the entire network.
medium has triangular membership function. In contrast Hence, all the nodes are aware of this value in advance.
with the two fuzzy input variables which are aimed at Thus, each node will measure its competition radius based
optimizing energy consumption, this fuzzy variable adds on this value. After tentative CHs determine their own
robustness against the changes in node locations. In case competition radius, the competition on selecting the final
the tentative CH has high density and its distance from RN CH will start. Each tentative CH will transmit a message
is medium or far, its competition radius will be higher and announcing that it is a candidate and wants to compete with
will cover more nodes. On the other hand, when its dis- other tentative CHs for becoming final CH (CandidateCH
tance from RN is close, high density will be tantamount to message). This message is only transmitted to tentative

Wireless Networks

Table 1 Fuzzy rules in

Distance to RN Remaining energy Calculated density Competition radius
proposed protocol
Close Low Dense 12XS
Close Low Normal 11XS
Close Low Sparse 10XS
Close Medium Dense 9XS
Close Medium Normal 8XS
Close Medium Sparse 7XS
Close High Dense 6XS
Close High Normal 5XS
Close High Sparse 4XS
Medium Low Dense 3XS
Medium Low Normal 2XS
Medium Low Sparse XS (extra S)
Medium Medium Sparse Small (S)
Medium Medium Normal Medium (M)
Medium Medium Dense Large (L)
Medium High Sparse XL (extra L)
Medium High Normal 2XL
Medium High Dense 3XL
Far Low Sparse 4XL
Far Low Normal 5XL
Far Low Dense 6XL
Far Medium Sparse 7XL
Far Medium Normal 8XL
Far Medium Dense 9XL
Far High Sparse 10XL
Far High Normal 11XL
Far High Dense 12XL

Fig. 2 Fuzzy set defining the fuzzy input variable distance to rendezvous node

Wireless Networks

Fig. 3 Fuzzy set defining the fuzzy input variable remaining energy

Fig. 4 Fuzzy set defining the fuzzy input variable calculated density

CHs which are located within the competition radius of that tentative CH has higher remaining energy than all the other
respective node and includes id, its competition radius tentative CHs from which it has received the candidate
(Compi), the remaining energy (Ei) and density (di). message, or in case the remaining energies are identical but
In this competition, the remaining energy is regarded as one node has more density, that node will change its state
the key parameter in selecting the final CH. Now, if a to the CH state. This competition guarantees that there will
tentative CH receives the CandidateCH Message from not be another CH in the competition radius of the
another tentative CH which is inside its competition radius respective CH; accordingly, energy consumption will be
and the remaining energy of the sender is more than the balanced.
receiver’s remaining energy, hence, the receiving node will When the competition ends and the final CHs are
transmit a message and will quit from the competition selected, each CH will introduce itself as the final CH to
(Quit Election Message).However, in case the remaining the other nodes via broadcasting a message. According to
energy of both nodes is identical, then, the node with the receiving signal strength, the other nodes will select the
higher density will be selected as the final CH. Thus, if a closest CH and will send a message to it and will introduce

Wireless Networks

Fig. 5 Fuzzy set defining the fuzzy output variable competition radius

themselves as the member of that cluster. After the com-

pletion of clustering, each CH will create a scheduling
based on TDMA scheduling protocol and will broadcast it
as a message to their members so that they are informed of
it. Such a scheduling provides them with the opportunity to
transmit their data at a specific time span. Also, it will
make the nodes keep their radio on only at their own
dedicated time span and they will turn off their radios at
other times so as to reduce energy consumption.
For preventing radio disorder and interruption among
neighboring nodes, each cluster uses a different CDMA
code. That is, after the CH is selected, it will randomly
select a model from the list of its codes and will inform
them about it so that they transmit their data according to
the announced code. Consequently, the radio communica-
tion of the neighboring cluster will not impact one another
and the communication will not be destroyed.
At the end of the first round, when the MS reached the
predetermined location, the movement path of the MS
should be determined in the next round. With regard to the
Fig. 6 Mobile sink divides the network into 16 equal parts
network dimensions, the MS divides the network into 16
equal parts as shown in Fig. 6.
Sum of Each Alive Nodes Energy in PartðaÞ
Since the MS is aware of nodes remaining energy, after Eavgi ¼
network is divided into sections (each square indicating Number of All Alive Nodes in PartðaÞ
one section), the average remaining energy of alive nodes ð3Þ
in each section will be measured through Eq. 3 (Eavgi).
As shown in Fig. 6, now, the MS must determine the
Then averages will be placed in an array and sort in
number of paths which should be taken in the new round by
ascending order by the MS. Eavgi will be obtained by
Eq. (4) and by dividing the number of sections which have
dividing the energy of all alive nodes in that section to the
alive nodes on a fixed value (optimal value: 7).
total alive nodes of that section.

Wireless Networks

Number of Alive Cell the distance of CH from RN is less than this value, data
Number of MS Trajectory ¼ Ceil will be directly delivered to RN within one hop by means
of the CH.
If data transmission distance is less than d-0, energy
Indeed, number 7 is a fixed value and there are 16 sections consumption will be the second exponent of d. Also, if the
in the network in all scenarios; hence, until all the sections transmission distance is more than d0, energy consumption
have alive nodes, the MS will go through all the three paths will be the fourth exponent of d which justifies the uti-
in each round. Then, when the number of the sections lization of d0. In case the distance is more than this value,
having alive nodes is 14 or less than that, the number of data will be transmitted to RN and is saved there so that the
MS trajectory will be reduced to 2 paths in each round. MS reaches near the RN; hence, data will be at the sink’s
This process will continue until the number of sections disposal. Detecting and determining when the sink node is
having alive nodes is 7 or less than that; in that case, it is at an appropriate distance from RNs and from the CHs
obvious that the MS will go through only one path. Such a which aim to directly transmit data to the sink is as follows.
reduction in the number of path is taken for granted At first, the MS broadcasts an initial message (before
because if the number of sink paths does not decrease in starting to move and at the beginning of each round) telling
accordance with the reduction of the number of alive network nodes its current location as well as its movement
nodes, network efficiency will be, accordingly, diminished. speed D; this value indicates its movement speed in the
Hence, the network will become inefficient. After the sink current round. As a result, according to the MS’s move-
specifies the number of paths, it will go to those sections of ment speed, route and its distance, the CHs will be able to
the network which have the minimum remaining energy; measure when the sink will be at the closest location with
indeed, based on the results obtained from the simulations them. Figure 7 depicts an instance of clustering and
of the proposed protocol, RNs will suffer less stress than movement path of the MS by the proposed protocol. In this
the other nodes. As a result, the movement of the sink example, 100 sensors have been randomly distributed in a
towards the regions with the least remaining energy will 200*200 m region and the MS is indicated in the form of a
result in the enhancement of network lifetime and the black circle; also, CHs and RNs have been indicated in this
proper distribution of energy consumption among nodes. figure. As it is obvious in this figure, after normal nodes
The MS determines its movement trajectory; then, it select their CHs, they are connected to the closest CH and
measures its distances from all the nodes. In case the dis- the movement path of the MS is depicted via the available
tance is less than Rx, it will be specified as RN; otherwise, arrows in this figure. Since all 16 sections have alive nodes,
it will be introduced as normal node. Next, the algorithm the sink will go through three paths.
will be executed from the beginning for the next round.

6 Energy model
5 Routing
In the proposed protocol, we use the same radio model as
After CHs are formed and specified, it is the time for [29]. Also, the same energy model as described in [40]. The
transmitting data towards the MS. In this process, at first, initial energy at each sensor is 0.25 J. The energy con-
CHs begin to collect the sensed data according to the sumptions of transmitting and receiving one byte are ETx ¼
scheduling which was given to the cluster members at the 50 nJ and ERx ¼ 50 nJ, respectively. The consumed energy
time of cluster establishment. Next, the collected data is for transmitting L bits of data with the distance of d meters
delivered directly or through multi-hops to the respective is measured through Eq. (5).
RN and finally to the MS. In the proposed routing, in case ETx ðl; dÞ ¼ ETxelec ðlÞ þ ETxamp ðl; dÞ ! ð5Þ
the distance of the transmitting CH from the MS is more 
than DIST_MAX and TD_MAX, hence, the respective CH Eelec  l þ Eamp  l  d2 ; d\do
ETx ðl; dÞ ¼
transmits its gathered data through another CH in its Eelec  l þ Eamp  l  d4 ; d  do
neighborhood which is closer to the RN. This process
The energy consumed at the data reception time is mea-
continues until the collected data reaches the destination. If
sured through Eq. (6).
a CH is located in a place at the transmission time which is
closer than the DIST_MAX (DIST_MAX = d0), hence, the ERx ðlÞ ¼ ERxelec ðlÞ ð6Þ
data will be directly at the MS’s disposal. The value of d0 is ERx ðlÞ ¼ Eelec  l
measured through Eq. (7) which will be discussed later in
the paper. On the other hand, for optimizing routing effi- Also, d0 value is obtained through Eq. (7).
ciency, a value called TD_MAX was introduced; in case

Wireless Networks

Fig. 7 A WSN clustered by

using our proposed protocol

do ¼ ð7Þ
7 Simulation and evaluation of the results
ETX parameter indicates energy consumption for trans-
mitting one bit of data, Efs stands for energy consumed by The proposed protocol was simulated and the obtained
the power amplifier in the free space model. And Eamp results were investigated. The proposed algorithm was
refers to the energy consumed by the power amplifier in the compared with three other protocols, i.e. LEACH, MOFCA
multipath model. In LEACH algorithm, due to distance, and Optimizing LEACH clustering algorithm with mobile
data transmission from CH to the sink consumes lots of sink and rendezvous nodes (OPT LEACH) in the form of
energy. In the proposed protocol, energy consumption is four different scenarios. As we know, with increasing
reduced which is attributed to the placement location of the network size, sensors will lose their energy more quickly.
MS. That is, based on the placement location of the sink, if Due to this, four scenarios with different dimensions have
the distance of the CH is less than d0 , it will transmit the been used to evaluate energy consumption, which are
collected data to the nearest RN so that it is buffered there similar to real cases. All of these scenarios were simulated
and would be transmitted to the MS at the appropriate time. and their results were investigated. It was found that the
As given in Eq. (8) in case data transmission distance is proposed protocol had a better performance than the other
less than d0 , energy consumption will be the second protocols.
exponent of d. On the other hand, if the transmission dis- Handy et al., used the metrics first node dies (FND), half
tance is more than d0 , energy consumption will the fourth of the nodes alive (HNA) and last node dies (LND) in [41]
exponent of d. to estimate the lifetime of the WSNs.

lEelec þ lefs d 2 ; d\do FND indicates the number of round in which the first
ETx ðl; dÞ ¼ ð8Þ node dies. This metric is effective and efficient in networks
lEelec þ lemp d 4 ; d  do
with little number of nodes. However, it should be noted
As a result, the energy required for transmitting and that the death of a single node in networks with higher
receiving data for each node is obtained through Eq. (9). number of nodes is not that much important. As a result,
Etotal ¼ ETx ðl; dÞ þ ERx ðlÞ ð9Þ Handy et al., introduced another metric, namely HNA
which denotes the number of round in which half of the

Wireless Networks

nodes die. Furthermore, they provide another metric, Table 2 Configuration for scenario 1
referred to as LND, which denotes an estimated value for Parameter Value
the overall lifetime of the network. However, it can be
argued that LND is an ineffective metric because after the AOI (network boundaries) 200 m 9 200 m
death of the half of network nodes, the network will no Number of nodes 100
longer be effective and efficient. Consequently, FND, HNA Data packet size 4000 bits
metric and the total remaining energy (TRE) were used in Eamp 0.0013 pJ/bit/m4
the present study for investigating the proposed algorithm Efs 10 pJ/bit/m2
and the above-mentioned algorithms for comparison. TRE Aggregation ratio 10%
is measured through Eq. (10). Rx 16%
Remain  Engtotal ðrÞ ¼ Er ðNÞ ð10Þ

Er ðNÞ indicates the energy amount of N nodes in the rth Table 3 Simulation results for scenario 1
Algorithm FND HNA TRE (J)
In each round of the proposed protocol, the movement
path of the MS was firstly determined. Then, the tentative LEACH 267 349 10.605
CHs and RNs were selected. Next, using fuzzy logic, the MOFCA 309 364 11.228
competition radius and the final CHs were determined. OPT LEACH 366 442 13.945
Eventually, each node selects its own CH. Accordingly, Proposed algorithm 408 482 14.387
network is established. After the establishment of the net-
work, each normal node transmits 4000 data bits to the
previously selected CH. Also, using a specific aggregation
ratio, each CH collects the related data from its members 7.1 Scenario one
and transmits them to the nearest RN or directly to the MS.
In the conducted simulations, the considered aggregation In the first scenario, the network size was assumed to be
rate was 10% which was similar to that of CHEF algorithm. 200*200 m. The initial location of the MS was prespecified
The length of the aggregated data is measured through and the movement path of the MS in the first round was
Eq. (11). also predetermined. The details related to this scenario are
Lagg ¼ Lrec þ Lrec  Ragg  N ð11Þ illustrated in Table 2.
Maximum competition radius for this network size was
In this equation, Lagg parameter stands for the length of equal to 50 m. TD_MAX was equal to 60 m and DIS-
data received from cluster members; Ragg refers to the T_MAX was d0 (about 87 m). Table 3 gives the results
aggregation rate and N denotes the number of available with respect to different metrics for the first scenario. As
nodes in the cluster. For instance, in case a CH has 15 shown in Table 3, the proposed algorithm has better per-
members and each one transmits 100 data bits to the CH, formance than LEACH, MOFCA and OPT LEACH in
the aggregation ratio is considered to be 10%. Hence, the terms of all three metric. Also, it should be noted that
length of the aggregated data will be equal to LEACH has the poorest performance when compared with
(100 ? (100*0.1*15)) which is equal to 250 bits. the other methods.
All the stages in the simulation were carried out by Figure 8 depicts the number of dead nodes in relation to
MATLAB R2015a. The value of initial energy in all the the number of rounds. As shown in this figure, the proposed
scenarios was considered to be 0.25 J. Each scenario was method loses its energy later than all the other methods.
carried out more than 100 times. The average obtained The OPT LEACH method has a desirable approach; nev-
results are reported below. In the first and second scenarios, ertheless, it should be noted that the proposed method had a
the threshold of the tentative CH in each round of the two better performance in three metrics of FND, HNA and
fuzzy methods (the proposed protocol and MOFCA) was LND.
0.2 for 100 nodes. Also, in the two methods in which Figure 9 depicts the amount of remaining energy in
LEACH was used for selecting CH, P value (the proba- relation to the number of rounds. It was observed that
bility of becoming CH) was assumed to be 0.1. Further- LEACH had poorer performance than the other methods
more, in the third and fourth scenarios, P value was since it does not consider the amount of nodes’ remaining
considered to be 0.2. energy in clustering; rather, it uses a probabilistic model
which is not adequate for network efficiency with such a

Wireless Networks

Fig. 8 Distribution of dead nodes in relation to the number of rounds (scenario 1)

Fig. 9 Total remaining energy in relation to the number of rounds (scenario 1)

size. Also, although MOFCA uses fuzzy logic, it encoun- noted that the protocol proposed in this paper combined
ters the problem of the premature loss of nodes which are both fuzzy logic, MS and the variability of the movement
nearer to the sink. Indeed, this problem is attributed to the path of the MS in each round; consequently, capitalizing on
fixedness of the sink which has been solved in the proposed these features, the proposed protocol was able to have a
method. Nevertheless, MOFCA had a better performance more desirable approach and performance than the other
than LEACH. protocols.
Regarding OPT LEACH, obviously, capitalizing on MS The last column in Table 3 indicates the total remaining
led to a better performance of OPT LEACH in comparison energy (TRE) in round 200. Since the initial energy of each
with LEACH and MOFCA. However, due to the fixedness node was 0.25 J, the total energy in round one was 25 J. As
of the movement path of the sink, the pressure on some it can be expected, the increase in the number of rounds
specific nodes was more than other nodes. It should be leads to the reduction of the energy of nodes.

Wireless Networks

Table 4 Configuration for scenario 2 DIST_MAX was regarded as d0. The obtained results are
Parameter Value
given in Table 5. The results indicated that as the network
becomes smaller, LEACH protocol gets better approach
AOI (network boundaries) 150 m 9 150 m and performance than MOFCA. As given in Table 5,
Number of nodes 100 regarding FND metric, the proposed protocol, as in the
Data packet size 4000 bits previous scenario, had a better performance than the other
Eamp 0.0013 pJ/bit/m4 protocols. However, after losing the first node, its energy
Efs 10 pJ/bit/m2 quickly diminishes. As HNA metric shows, when half of
Aggregation ratio 10% the nodes has been died, it is OPT LEACH which had a
Rx 16% better performance than the proposed protocol.
Figure 10 illustrates the number of dead nodes in rela-
tion to the number of rounds. As shown in this figure, it is
observed that the proposed protocol loses its initial energy
Table 5 Simulation results for scenario 2 later than the other methods. Nevertheless, its energy
drains quickly; consequently, the efficiency of the proposed
Algorithm FND HNA TRE (J)
protocol diminishes sooner.
LEACH 295 384 12.244 Figure 11 depicts the amount of remaining energy in
MOFCA 252 337 10.006 relation to the number of nodes. According to this figure, it
OPT LEACH 420 483 14.85 can be observed that the proposed protocol and OPT
Proposed algorithm 441 464 13.979 LEACH lose their energies in a parallel way. However,
eventually, it was OPT LEACH which had a better per-
formance. In this scenario, although MOFCA uses fuzzy
logic, it had a weaker performance than the other methods.
It is obvious that using MS led to the improvement of
7.2 Scenario two
network efficiency. The last column of Table 5 indicates
the total remaining energy at the end of round 200.
In the second scenario, network size was 150*150 m. The
initial location of the MS and the movement path of the MS
7.3 Scenario three
in the first round of the proposed protocol were predeter-
mined. The details related to this scenario are given in
In this scenario, network size was equal to 300*300 m. The
Table 4.
initial location of MS and the movement path of the MS in
Maximum competition radius for this network dimen-
the first round of the proposed protocol were
sion was considered to be 50 m. TD_MAX was 50 m and

Fig. 10 Distribution of dead nodes in relation to the number of rounds (scenario 2)

Wireless Networks

Fig. 11 Total remaining energy in relation to the number of rounds (scenario 2)

available algorithms of WSNs in big dimensions encounter

Table 6 Configuration for scenario 3
issues and problems in terms of energy consumption.
Parameter Value However, the proposed protocol had high performance
AOI (network boundaries) 300 m 9 300 m
with regard to all three respective metrics. As it can be
expected, LEACH also had the most weaknesses and
Number of nodes 100
problems in a network with big dimensions.
Data packet size 4000 bits
Figure 12 illustrates the number of dead nodes in rela-
Eamp 0.0013 pJ/bit/m4
tion to the number of rounds. According to this figure, it
Efs 10 pJ/bit/m2
was found that the proposed protocol, with regard to the
Aggregation ratio 10%
three metrics of FND, HNA and LND, loses its nodes later
Rx 16%
than the other methods.
Furthermore, Fig. 13 depicts the amount of remaining
energy in relation to the number of rounds. As illustrated in
Figs. 12 and 13, it can be argued that LEACH performance
was weaker than the other protocols which are attributed to
Table 7 Simulation results for scenario 3
the high distance of nodes in relation to the sink. In a
Algorithm FND HNA TRE (J) similar vein, MOFCA suffers from the energy hole prob-
LEACH 129 235 6.9703 lem and quickly loses its energy. Regarding OPT LEACH,
MOFCA 216 288 8.5059
due to the fixedness of the MS trajectory, there is a pressure
OPT LEACH 318 376 12.591
on certain nodes. In contrast, the proposed protocol had a
significantly better performance which is due to the variety
Proposed algorithm 362 414 12.925
in selecting the trajectory of the MS. The last column of
Table 7 indicates the total remaining energy (TRE) at
round 200.

predetermined. The details related to this scenario are 7.4 Scenario four
given in Table 6.
Maximum competition radius with regard to the network In this scenario, network size was assumed to be
size was 70 m. TD_MAX and DIST_MAX were consid- 400*400 m. Like the previous scenarios, the initial location
ered to be d0. Table 7 gives the results for different metrics and the movement path of the MS at the first round were
which were obtained from the simulations. As shown in prespecified. The details of scenario are given in Table 8.
this table, it is obvious that the proposed algorithm had a In this scenario, due to the enhancement of length and
better performance than the other methods. In general, the width of network, the maximum competition radius was

Wireless Networks

Fig. 12 Distribution of dead nodes in relation to the number of rounds (scenario 3)

Fig. 13 Total remaining energy in relation to the number of rounds (scenario 3)

Table 8 Configuration for scenario 4 Table 9 Simulation results for scenario 4

Parameter Value Algorithm FND HNA TRE (J)
AOI (network boundaries) 400 m 9 400 m LEACH 55 166 0
Number of nodes 100 MOFCA 109 205 4.3092
Data packet size 4000 bits OPT LEACH 63 268 5.9669
Eamp 0.0013 pJ/bit/m4 Proposed algorithm 134 277 7.018
Efs 10 pJ/bit/m2
Aggregation ratio 10%
Rx 16% 90 m; TD_MAX and DIST_MAX was d0. Table 9 gives
the results for the different metrics of this scenario.

Wireless Networks

Fig. 14 Distribution of dead nodes in relation to the number of rounds (scenario 4)

Fig. 15 Total remaining energy in relation to the number of rounds (scenario 4)

As given in this table, the proposed protocol had a better Consequently, it leads to the reduction of network lifetime.
performance than LEACH, MOFCA and OPT LEACH As it can be expected, MOFCA had a better performance
with respect to all three metrics. than but weaker than the methods which had MS (due to
Figure 14 illustrates the number of dead nodes in rela- the occurrence of the energy hole problem). OPT LEACH
tion to the number of rounds. As it is obvious in the figure, had performance similar to that of the proposed protocol
the proposed method had a relatively higher performance but a little weaker than that. Within a very short time, OPT
than the other methods. Also, Fig. 15 illustrates the amount LEACH loses its first node.
of remaining energy in relation to the number of rounds. The proposed protocol duo to use of fuzzy logic for
Similar to the other previous scenarios with big unequal clustering instead of using fixed-size clusters like
dimensions, LEACH has the weakest performance among that used in LEACH, it has solved the energy hole problem.
all the available methods since the energy hole problem One of the other weaknesses of the LEACH protocol is that
emerges earlier due to the large distance to the sink. it uses fixed base station, but in the proposed protocol, MS

Wireless Networks

moves and gathers data from network nodes. Unlike the should be argued that the proposed protocol can be opti-
LEACH method, the MOFCA protocol, has used unequal mized and improved with regard to different aspects in
clustering in his method but using a fixed base station future studies. As a case in point, the optimization of the
causes to energy hole appears faster than our protocol. OPT movement path of the MS for smartly referring to the nodes
LEACH also suffers from energy hole problems despite the with the highest or lowest remaining energy is regarded as
use of MS, because it uses a predefined trajectory for MS a potential issue which can be addressed in future works.
and its clustering is similar to the LEACH method, but in Moreover, using multiple MSs simultaneously in a
our proposed protocol, in each round the movement path of specific region in such a way that sinks can define paths,
the MS was firstly determined, also, to prevent the energy cover nodes and gather data in collaboration with one
hole problem our protocol proposing a novel unequal another is considered to be another direction for further
clustering method by using fuzzy logic. As a result, the research. It should be maintained that the proposed method
proposed protocol is better than other methods. is appropriate for networks with relatively greater dimen-
As shown in scenarios one, three and four which are on sions. Thanks to utilizing the density parameter in clus-
a large-scale, due to the use of the MS with variable tra- tering via the fuzzy logic, the movement of the nodes
jectory based on the proposed method and fuzzy approach through the natural factors was considered in the proposed
the results of simulating indicated that it had significant method. Thus, the intentional movement and replacement
improvements in comparison with the other methods. In the of the nodes can be added to the network using the pro-
first scenario, regarding HNA and FND metrics, 9 and 11% posed method so that, if needed, nodes can change their
improvements were obtained in comparison with OPT locations.
LEACH which uses the predetermined path. In the second
scenario, regarding HNA metric, the proposed method 4%
weaker performance than OPT LEACH with respect to
HNA but it had 5% better performance in terms of FND. In References
the third scenario, HNA and FND metrics in the proposed
1. Chao, C., & Hsiao, T. (2014). Design of structure-free and
protocol were improved for 10 and 13%, respectively. energy-balanced data aggregation in wireless sensor networks.
Regarding the fourth scenario, network lifetime in the Journal of Network and Computer Applications, 37, 229–239.
proposed protocol was enhanced which is attributed to 3 2. Khedr, A. M., & Osamy, W. (2011). Effective target tarcking in a
and 112% improvements of HNA and FND, respectively. self-organizing wireless sensor network. Journal of Parallel and
Distributed Computing, 71, 1318–1326.
The findings show that the proposed method works better 3. Li, H., Liu, Y., Chen, W., Jia, W., Li, B., & Xiong, J. (2013).
than other protocol in most of the scenarios, especially COCA: Constructing optimal clustering architecture to maximize
large-scale networks. sensor network lifetime. Computer Communications, 36,
4. Liu, Z., Zheng, Q., Xue, L., & Guan, X. (2012). A distributed
energy-efficient clustering algorithm with improved coverage in
8 Conclusion wireless sensor networks. Future Generation Computer Systems,
28, 780–790.
The issue under focus in this paper was to improve and 5. Bagci, H., & Yazici, A. (2013). An energy aware fuzzy approach
to unequal clustering in wireless sensor networks. Applied Soft
optimize the performance of WSNs through proposing a Computing, 13, 1741–1749.
novel unequal clustering method by using fuzzy logic and 6. Li, C., Ye, M., Chen, G., & Wu, J. (2005). An energy-efficient
multiple sinks. Considering the problems and shortcomings unequal clustering mechanism for wireless sensor networks. In
of the earlier methods and related works, we developed a IEEE international conference on mobile ad hoc and sensor
systems conference.
novel method for sorting out the energy hole problem in 7. Bhagyalakshmi, L., Suman, S. K., & Murugan, K. (2012). Corona
WSNs by capitalizing on fuzzy logic and MS. We proved based clustering with mixed routing and data aggregation to
that, in order to avoid the energy hole problem, the number avoid energy hole problem in wireless sensor network. In IEEE-
of clusters in areas should strictly increase as the distance fourth international conference on advanced computing, ICoAC
MIT, Anna University, Chennai.
towards the MS trajectory decreases. The proposed algo- 8. Mottaghi, S., & Zahabi, M. R. (2015). Optimizing LEACH
rithm considers distance to closest RN, remaining energy clustering algorithm with mobile sink and rendezvous nodes.
and calculated density as the input variables of the fuzzy International Journal of Electronics and Communications (AEÜ),
logic. That is, these variables are taken into consideration 69, 507–514.
9. Baranidharan, B., & Santhi, B. (2016). DUCF: Distributed load
for measuring competition radius of each tentative CH. balancing unequal clustering in wireless sensor networks using
Also we developed and proposed a method for the smart fuzzy approach. Applied Soft Computing, 40, 495–506.
selection of MS route. Extensive simulations have shown 10. Misra, S., Roy, S., Obaidat, M. S., & Mohanta, D. (2009).
that our method extends the network lifetime as compared A Fuzzy logic-based energy efficient packet loss preventive
to the other protocols. As directions for further research, it

Wireless Networks

routing protocol. In IEEE performance evaluation of computer & 29. Heinzelman, W. R., Chandrakasan, A., & Balakrishnan, H.
telecommunication systems. (2000). LEACH: Energy-efficient communication protocol for
11. Rajsekaran, S., & Pai, G. A. V. (2003). Neural networks, fuzzy wireless microsensor networks. In Proceedings of the 33rd
logic and genetic algorithm, synthesis and application. Delhi: Hawaii international conference on system sciences.
Prentice Hall of India. 30. Sert, S. A., Bagci, H., & Yazici, A. (2015). MOFCA: Multi-
12. Liang, W., Luo, J., & Xu, X. (2010). Prolonging network lifetime objective fuzzy clustering algorithm for wireless sensor networks.
via a controlled mobile sink in wireless sensor networks. In Applied Soft Computing, 30, 151–165.
Global telecommunications conference (pp. 1–6). 31. Toloueiashtian, M., & Motameni, H. (2018). A new clustering
13. Akkaya, K., Younis, M., & Bangad, M. (2005). Sink reposi- approach in wireless sensor networks using fuzzy system. The
tioning for enhanced performance in wireless sensor networks. Journal of Supercomputing, 74(2), 717–737.
Computer Networks, 49(512), 534. 32. Gong, H., Fu, L., Fu, X., Zhao, L., Wang, K., & Wang, X. (2016).
14. Jain, S., Shah, R. C., Brunette, W., Borriello, G., & Roy, S. Distributed multicast tree construction in wireless sensor net-
(2006). Exploiting mobility for energy efficient data collection in works. IEEE Transactions on Information Theory, 63(1),
sensor networks. Mobile Networks and Applications, 11, 280–296.
327–339. 33. Fu, X., Xu, Z., Peng, Q., You, J., Fu, L., Wang, X., & Lu, S.
15. Chakrabarti, A., Sabharwal, A., & Aazhang, B. (2006). Com- (2017). ConMap: A novel framework for optimizing multicast
munication power optimization in a sensor network with a path- energy in delay-constrained mobile wireless networks. In ACM
constrained mobile observer. ACM Transactions on Sensor Net- MobiHoc.
works, 2, 297–324. 34. Nayak, S. P., Rai, S. C., & Pradhan S. (2017). A multi-clustering
16. Gu, Y., Bozdag, D., Brewer, R. W., & Ekici, E. (2006). Data approach to achieve energy efficiency using mobile sink in WSN.
harvesting with mobile elements in wireless sensor networks. In Springer nature Singapore, computational intelligence in data
Computer Networks, 50, 3449–3465. mining, advances in intelligent systems and computing p. 556.
17. Jea, D., Somasundara, A., & Srivastava, M. (2005). Multiple 35. Liu, B., Bras, P., Dousse, O., Nain, P., & Towsley, D. (2005).
controlled mobile elements (data mules) for data collection in Mobility improves coverage of sensor networks. In Proceedings
sensor networks. In IEEE/ACM conference on distributed com- of the 6th ACM international symposium on mobile ad hoc net-
puting in sensor systems (DCOSS). working and computing (pp. 300–308).
18. Akyildiz, I. F., Su, W., Sankarasubramaniam, Y., & Cayirci, E. 36. Liu, W., Lu, K., Wang, J., Xing, G., & Huang, L. (2012). Per-
(2002). Wireless sensor networks: A survey. Computer Networks, formance analysis of wireless sensor networks with mobile sinks.
38, 393–422. IEEE Transactions on Vehicular Technology, 61, 2777–2788.
19. Tashtarian, F., Yaghmaee, M. H., Sohraby, K., & Effati, S. 37. Liu, W., Lu, K., Wang, J., Huang, L., & Wu, D. O. (2012). On the
(2015). ODT: Optimal deadline-based trajectory for mobile sinks throughput capacity of wireless sensor networks with mobile
in WSN: A decision tree and dynamic programming approach. relays. IEEE Transactions on Vehicular Technology, 61,
Computer Networks, 77, 128–143. 1801–1809.
20. Shah, R. C., Jain, S., & Brunette, W. (2003). Data MULEs: 38. Xing, G., Wang, T., Xie, Z., & Jia, W. (2008). Rendezvous
Modeling a three-tier architecture for sparse sensor networks. In planning in mobility-assisted wireless sensor networks. IEEE
IEEE workshop on sensor network protocols and applications Transactions on Mobile Computing, 7, 1430–1443.
(SNPA) (pp. 30–41). 39. Duan, Z., Guo, F., Deng, M., & Yu. M. (2009). Shortest path
21. Tong, L., Zhao, Q., & Adireddy. S. (2003). Sensor networks with routing protocol for multi-layer mobile wireless sensor networks.
mobile agents. In Proceedings of the IEEE MILCOM (pp. In International conference on networks security, wireless com-
688–693). munications and trusted computing.
22. Shi, Y., & Hou, Y. T. (2009). Optimal base station placement in 40. Heinzelman, W. B. (2000). Application specific protocol archi-
wireless sensor networks. ACM Transactions on Sensor Net- tectures for wireless networks. Ph.D. thesis, Massachusetts
works, 5, 32. Institute of Technology.
23. Yun, Y., & Xia, Y. (2010). Maximizing the lifetime of wireless 41. Handy, M. J., Haase, M., & Timmermann, D. (2002). Low energy
sensor networks with mobile sink in delay-tolerant applications. adaptive clustering hierarchy with deterministic cluster-head
IEEE Transactions on Mobile Computing, 9, 1308–1318. selection. In Mobile and wireless communications network (pp.
24. Xing, G., Li, M., Wang, T., Jia, W., & Huang, J. (2012). Efficient 9–11).
rendezvous algorithms for mobility-enabled wireless sensor net-
works. IEEE Transactions on Mobile Computing, 11, 47–60.
25. Wang, Y.-C., Peng, Y.-C., & Tseng, Y.-C. (2010). Energy-bal- Kambiz Koosheshi received his
anced dispatch of mobile sensors in a hybrid wireless sensor M.S. in Software Engineering
network. IEEE Transactions on Parallel and Distributed Systems, from the Tabriz Branch of Isla-
21, 1836–1850. mic Azad University, Iran in
26. Gu, Y., Ji, Y., Li, J., & Zhao, B. (2013). ESWC: Efficient 2016 and his B.S. in Software
scheduling for the mobile sink in wireless sensor networks with Engineering from the Central
delay constraint. IEEE Transactions on Parallel and Distributed Tehran Branch of Islamic Azad
Systems, 24, 1310–1320. University, Iran in 2011. He has
27. Zhao, M., Ma, M., & Yang, Y. (2010). Efficient data gathering published several research
with mobile collectors and space-division multiple access tech- papers at both National and
nique in wireless sensor networks. IEEE Transaction on Com- International Conferences. His
puters, 60, 400–417. specializations include Wireless
28. Basagni, S., Carosi, A., Melachrinoudis, E., Petrioli, C., & Wang, Communication and Networks.
M. (2008). Controlled sink mobility for prolonging wireless His current research interests
sensor networks lifetime. Springer Science Wireless Network, 14, are Wireless Sensor and Ad hoc
831–858. Networks, IoT and 5G Wireless Systems and Networks. He is also a
member of the IEEE.

Wireless Networks

Saeed Ebadi is a Ph.D. student University, Tabriz Branch. His research interests include Cloud
in the Computer Architecture computing, grid computing and wireless sensor networks.
Engineering at Computer Engi-
neering Department of the Sci-
ence and Research Branch,
Islamic Azad University, Teh-
ran, Iran, since 2012. He
received M.S. (2010) from
Islamic Azad University, Tabriz
Branch, Iran, B.S. (2008) from
University of Applied Science
and Technology, Tabriz, Iran.
He is currently faculty member
in the Department of Computer
Engineering at Islamic Azad