sensors

Article
Energy Efficient Routing Algorithm with Mobile Sink
Support for Wireless Sensor Networks
Jin Wang 1,2,3 , Yu Gao 2 , Wei Liu 2 , Arun Kumar Sangaiah 4 and Hye-Jin Kim 5, *
1 Hunan Provincial Key Laboratory of Intelligent Processing of Big Data on Transportation,
School of Computer & Communication Engineering, Changsha University of Science & Technology,
Changsha 410000, China; [email protected]
2 College of Information Engineering, Yangzhou University, Yangzhou 225000, China;
[email protected] (Y.G.); [email protected] (W.L.)
3 School of Information Science and Engineering, Fujian University of Technology, Fuzhou 350000, China
4 School of Computing Science and Engineering, Vellore Institute of Technology (VIT), Tamil
Nadu 632014, India; [email protected]
5 Business Administration Research Institute, Sungshin W. University, Seoul 100744, Korea
* Correspondence: [email protected]; Tel.: +82-02-2287-0097




Received: 4 February 2019; Accepted: 26 March 2019; Published: 27 March 2019


Abstract: Recently, wireless sensor network (WSN) has drawn wide attention. It can be viewed as a
network with lots of sensors that are autonomously organized and cooperate with each other to collect,
process, and transmit data around targets to some remote administrative center. As such, sensors may
be deployed in harsh environments where it is impossible for battery replacement. Therefore, energy
efficient routing is crucial for applications that introduce WSNs. In this paper, we present an energy
efficient routing schema combined with clustering and sink mobility technology. We first divide
the whole sensor field into sectors and each sector elects a Cluster Head (CH) by calculating its
members’ weight. Member nodes calculate energy consumption of different routing paths to choose
the optimal scenario. Then CHs are connected into a chain using the greedy algorithm for intercluster
communication. Simulation results prove the presented schema outperforms some similar work
such as Cluster-Chain Mobile Agent Routing (CCMAR) and Energy-efficient Cluster-based Dynamic
Routing Algorithm (ECDRA). Additionally, we explore the influence of different network parameters
on the performance of the network and further enhance its performance.

Keywords: Wireless sensor network (WSN); clustering; mobile sink; moving trajectory scheduling;
energy efficiency

1. Introduction
Recent advances in wireless communication, and computer and micro-electronic technology have
enabled the rapid development of tiny, low-cost, and multi-functional sensors. These sensors are
commonly deployed in the target areas via a random way to monitor the physical features of the
surroundings such as temperature, humidity, and pressure. The monitored data is usually forwarded to
the data collector (sink) using a cooperative method (generally multi-hop) and the collector may relay
the data to a remote server for further data analysis. Besides, sensors can self-organize themselves
based on their local collaboration to form wireless sensor networks (WSNs) [1–3]. The favorable
characteristics of WSNs, such as rapid deployment, high fault tolerance, self-organization, real-time
data transport, etc., make them suitable to be deployed in unfriendly, or even harsh environments,
especially in military or disaster surveillance. Additionally, WSNs are also extensively applied in
industrial product line monitoring, agricultural and wildlife observation, healthcare, smart homes,
etc. [4,5].

Sensors 2019, 19, 1494; doi:10.3390/s19071494 www.mdpi.com/journal/sensors

Sensors 2019, 19, 1494 2 of 19

Sensors are generally powered by energy-constrained batteries and it is unrealistic for

battery replacement because of the enormous quantity of sensors and expensive cost. The best
solution for the insufficient energy in WSNs is to equip the sensors with higher-capacity batteries.
However, the performance of the battery has reached its choke point and is hard to be improved [6,7].
Therefore, the energy shortage problem in WSNs could only be addressed by adopting energy efficient
routing protocols or algorithms [8,9]. Additionally, WSNs are also troubled with the unbalanced energy
of sensors. Each sensor owns its monitoring range and once a node dies, fade zones will appear and
the performance of the network will also rapidly decrease. The phenomenon of unbalanced energy in
WSNs is similar to the energy holes, and they are both caused by the “hot spots” problem. The “hot
spots” problem usually happens in WSNs with a static sink and fixed network topology. Nodes around
the sink are usually burdened with busier traffic because they need to transfer the data packages from
the outer layer to the sink and exhaust their energy faster. Since the sink is static and the network
topology remains unchanged, the energy becomes more and more heterogeneous as time goes on.
Network lifetime is an important evaluation criterion to estimate the performance of the network and
it is commonly defined as the time when the first node dies.
A great deal of work has focused on the problem of energy efficiency and energy balance, and great
results have been achieved [10–13]. Clustering technology greatly decreases the energy consumption
of WSNs by dividing the sensors into clusters according to some special rules. In each cluster, one or
more cluster heads (CHs) are elected to be the relay nodes for its members. Clustering helps the
network simplify the topology structure and avoids the direct communication between sensors and
the sink. Additionally, data fusion can be adopted in CHs to filter the redundant data to decrease the
burden of CHs. One of the classic routing protocols that adopts clustering is the low-energy adaptive
clustering algorithm (LEACH); however, the approach for CHs selection is unreasonable and much
further work is being conducted based on LEACH.
Sink mobility technology is an efficient method to deal with the unbalanced energy in WSNs.
In mobile sink-supported WSNs, the sink is usually carried by the intelligent vehicles or robots
and it can move freely around the sensing field. The following advantages can be achieved by
introducing sink mobility technology. First, the “hot spots” problem can be greatly alleviated by the
movement of the sink. Areas around the sink are usually traffic hubs and when the sink moves,
the traffic hub also transfers. Sensors take turns to be the “hot spots” to evenly distribute the
energy consumption. Second, the total energy consumption can be greatly reduced by abridging
the transmission distance between the communication pairs, assuming that the sink mobility pattern
is well designed. Third, the latency of the network can be reduced and the throughput of the network
can be increased by using a mobile sink. Finally, network connectivity can be ensured, even under the
sparse or disconnected sensor networks. Although the introduction of sink mobility involves many
merits, it also faces some challenges [14]. The location of the mobile sink should be broadcast frequently
or predicted by sensors, which may increase the burden of the network. Additionally, the moving
pattern of the sink should be well designed to cooperate with the local nodes for data transmission.
For addressing the above problems, in this paper, we present an energy-efficient routing schema
combined with clustering and mobile sink technology. We first divide the whole sensor area into several
sectors with the same size. In each cluster, a CH is selected according to the weight, which is calculated
using the residual energy and the distance between the source node and the CH. Source nodes
communicate with the CH using single or multi-hop communication in accordance with the optimal
energy consumption. Additionally, CHs are connected by a reasonable chain, which is constructed by
the greedy algorithm for intercluster communication. The CH that is closest to the sink is chosen as the
leader to communicate with the sink. Then, the mobile sink moves along a predefined trajectory for
data gathering. Extensive simulations have been performed to compare our presented schema with
some similar work. Additionally, we further study the influence of some parameters on the network
and enhance the performance of the network.
Sensors 2019, 19, 1494 3 of 19

The rest of the paper is organized as follows. Section 2 discusses some classic routing protocols and
some latest research achievement adopting mobile sink. Section 3 presents the system model, which
contains the network and energy models. Section 4 describes our algorithm in detail. Large quantities
of simulations are conducted for comparing our presented schema with some similar work and the
results are discussed in Section 5. Section 6 analyzes the influence of different network parameters.
Section 7 discusses some phenomena during the simulation and proposed some open research issues.
Section 8 concludes this paper.

2. Related Work
During recent years, much attention has been paid to energy-efficient routing protocols.
Energy-efficient cluster-based dynamic routing algorithm (ECDRA) [13] is a mobile sink-based routing
schema. In ECDRA, the mobile sink is deployed at the outer side of the circular sensor field and rotates
in a circular manner. The topology of the network changes dynamically according to the position of
the mobile sink.
LEACH [15] is one of the most well-known and representative hierarchical routing protocols
that was first proposed. In LEACH, all sensors are divided into two types, cluster heads (CHs) and
ordinary nodes (ONs). An ON will deliver its monitored data to its corresponding CH, and the CH will
fuse and forward the monitored data to the base station (BS). LEACH is much superior to traditional
routing protocols in terms of extending the network lifetime. However, due to the random election of
CHs, CHs are often unevenly distributed, and CHs communicate with the BS directly, causing large
energy dissipation.
Power-efficient gathering in sensor information systems (PEGASIS) [16] is an enhanced version of
LEACH, which is a chain-based hierarchical protocol. In PEGASIS, each node only needs to transmit the
data package to its nearest neighbor, which is closer to the BS than the source node. CHs are connected
into a chain by the greedy algorithm for intercluster communication. Then, each chain leader, which is
closest to the BS, takes the responsibility to forward the data packages to the BS. The chain construction
makes an economical use of energy by avoiding long-distance communication. Meanwhile, because of
using multi-hop propagation, serious network delay could be caused. Therefore, this protocol is not
suitable for delay-sensitive applications.
Hybrid, energy-efficient, distributed clustering approach (HEED) [17] is greatly improved
alternative regarding aspects of CHs selection compared to LEACH. It uses the residual energy
as a main parameter and nodes with higher energy have a bigger possibility to be the CHs. In this
way, the time when the first node dies could be greatly postponed, and more uniform CHs could be
generated. Intracluster communication cost is the secondary parameter to determine which cluster
a node should take part in. Intracluster communication cost could be estimated using the average
minimum reachability power (AMRP) as shown in Equation (1):

∑iM
=1 MinPwri
AMRP = (1)
M
where MinPwr denotes the minimal transmission power demanded by a node to communicate with
its corresponding CH. Using AMRP further balances the energy consumption of the whole network.
In the above routing protocols, the position of the sink is usually fixed, which results in the
“hot spots” problem. With the popularity of the portable mobile devices, such as Radio Frequency
Identification (RFID), smart phones, and Personal Digital Assistants (PDAs), the further application
of the robots, Unmanned Ground Vehicles (UGVs), and Unmanned Aerial Vehicles (UAVs), routing
protocols using mobile sinks have become a hot topic in recent several years. Below are some routing
protocols using single mobile sink.
Zhao et al. [18] proposed an algorithm called Load Balanced Clustering and Dual Data Uploading
(LBC-DDU). In LBC-DDU, the whole network is divided into three layers: sensor layer, cluster head
layer, and SenCar layer. At the beginning of each round, the SenCar could calculate the optimized
Sensors 2019, 19, 1494 4 of 19

path in advance and then walk along the path to gather information utilizing single-hop transmissions.
After visiting each selected point, the SenCar will return to the base station and make preparations for
the next round. The SenCar is equipped with two antennas such that it can exchange information with
two CH simultaneously utilizing the technology of Multi-User Multiple-Input and Multiple-Output
(MU-MIMO), which makes contributions toward delay reduction and efficiency improvement.
Zhu et al. [19] proposed a tree-cluster-based data-gathering algorithm (TCBDGA). In TCBDGA,
the weight of each node is calculated using several main factors including residual energy, the number
of neighbors and the distance to the BS. Each node chooses its neighbor with maximum weight as
its parent node. In this way, a tree-construction is set up and then each tree is decomposed into
several sub-trees according to its depth and its traffic load. Simulation results show that it has better
performance in terms of energy consumption.
Xie and Pan [20] introduced the mobile sink in WSNs with obstacles, where a spanning
graphs-based scheduling mechanism was proposed. On the basis of this scheduling mechanism,
the authors propose a heuristic path planning algorithm to find a shortest path to avoid the obstacles.
A mobile sink is set to walk along the path and collect data from CHs via direct transmission.
The mobile sink returns to the origin after data collection. Simulation results demonstrate that
their proposed algorithm can extend network lifetime effectively. The scheduling mechanism they
proposed makes a good contribution to reduce the complexity traversing in WSNs with obstacles.
Velmani and Kaarthick [21] proposed a velocity energy-efficient and link-aware cluster-tree
(VELCT) scheme. This scheme mainly contains two phases, set-up phase and steady-state phase.
In the set-up phase, CHs are selected based on the threshold value and clusters are formed with better
performance via intra-cluster communication. After that, a few nodes are selected to serve as data
collection nodes (DCNs) to construct a data collection tree (DCT). In the steady-state phase, CHs collect
monitored data from its members and then forward the aggregated data to DCN. Experimental results
show that their proposed scheme has better performance in network throughput, energy consumption
and network latency.
Selvakumar and Swamynathan [22] presented an efficient data aggregation algorithm called
cluster-chain mobile agent routing (CCMAR). In CCMAR, the cluster head selection value (CHSV) is
utilized to choose the optimal CHs. Cluster members form chains to transmit their data to the sink.
The residual energy level, signal strength and path loss are considered to schedule an optimal routing
path for the mobile agent (MA).
Furthermore, some literature tries to optimize the performance of the network in aspects of energy
consumption and lifetime by combining the particle swarm optimization algorithm [23–25], immune
algorithm [26], bio-inspired ant colony optimization algorithm [27], clustering algorithms, etc.
Some researchers introduce multiple mobile sinks in WSNs and further improve the performance
of the network. Tashtarian et al. [28] presented an event-driven algorithm using multiple mobile sinks.
In the algorithm, each node has two states: monitoring and transmitting states. When an event is
captured, the related node turns its state from monitoring to transmitting and a group of active nodes
are generated to send packages to the mobile sink. In large scale WSNs, due to the limit of the mobility
of the mobile sink, the network is divided into several parts and each part introduces a mobile sink.
Wang et al. [29] presented a distance-aware routing algorithm using multiple mobile sinks to decrease
the energy consumption of the network. Wang et al. [30] introduced multiple mobile sinks to move
along the predetermined paths to gather raw data.
With the aim of prolonging the lifetime of the whole network, Banerjee et al. [31] introduced
multiple mobile CHs with rich energy. The mobile CHs work in a collaborative manner to collect
information from different sections of the sensor field. The base station takes responsibility for
communicating with the mobile CHs. Three different strategies were adopted to reduce multi-hop
communication and enhance the lifetime of the network.
Sensors 2019, 19, 1494 5 of 19

3. System Model

3.1. Basic Assumptions

Sensors 2019, 19, 1494 5 of 19
In this paper, we consider that the sensing field is a circular region. Each node owns a unique ID
that differs
In this from others’
paper, nodes. that
we consider We make some basic
the sensing assumptions
field is as follows:
a circular region. Each node owns a unique ID
•that differs from others’
All sensors nodes. We
are randomly make some
deployed by basic assumptions
vehicles, such as as
a follows:
plane, and remain stationary
• All sensors are
after deployment.randomly deployed by vehicles, such as a plane, and remain stationary after
• deployment.
All sensors have the knowledge of the location of the other nodes via the information exchange.
• All sensors have the knowledge of the location of the other nodes via the information exchange.
• All sensors have the same initial energy and their batteries cannot be changed. Once a sensor
• All sensors have the same initial energy and their batteries cannot be changed. Once a sensor
exhausts its energy, it will be useless.
exhausts its energy, it will be useless.
• The transmission power of sensors could be adjusted based on the communication distance.
• The transmission power of sensors could be adjusted based on the communication distance.
• The moving trajectory of the mobile sink is well-scheduled and it owns unlimited energy and
• The moving trajectory of the mobile sink is well-scheduled and it owns unlimited energy and
communication range.
communication range.
3.2. Network Model
3.2. Network Model
In this paper, we deploy N sensors in a circular field with radius R. The N sensors are denoted
In this paper, we deploy N sensors in a circular field with radius R. The N sensors are denoted
as: {N1, N2, . . . , Nn}. We arrange the mobile sink within the sensor field and it can move freely in the
as: {N1, N2……., Nn}. We arrange the mobile sink within the sensor field and it can move freely in
sensor field. The whole sensor area is evenly divided into several sectors and each sensor belongs to a
the sensor field. The whole sensor area is evenly divided into several sectors and each sensor belongs
sector based on its geographical position, as is shown in Figure 1.
to a sector based on its geographical position, as is shown in Figure 1.

Figure 1. Network model.

3.3. Energy Model

3.3. Energy Model
Here,
Here, we
we adopt
adopt the
the energy
energy model
model that
that is
is applied
applied in
in literature
literature [32]. As is
[32]. As is shown
shown inin Figure
Figure 2,
2, energy
energy
consumption mainly contains two parts, transmission consumption and reception
consumption mainly contains two parts, transmission consumption and reception consumption. The consumption.
The transmission
transmission partpart contains
contains the the signal
signal generation
generation andand its enhancement.
its enhancement. Once Once
thethe signal
signal is created,
is created, it
it will be enhanced by the amplifier. The amplifier uses two different powers to
will be enhanced by the amplifier. The amplifier uses two different powers to strengthen the strengthen the signal
signal
according
according to to the
the transmission
transmission distance. Therefore, the
distance. Therefore, the energy
energy model
model for
for transmission
transmission is is also
also divided
divided
into
into a free space model, which is for short-haul communication, and multi-path fading model, which
a free space model, which is for short-haul communication, and multi-path fading model, which
is
is for
for long
long distance
distance communication.
communication.
consumption mainly contains two parts, transmission consumption and reception consumption. The
transmission part contains the signal generation and its enhancement. Once the signal is created, it
will be enhanced by the amplifier. The amplifier uses two different powers to strengthen the signal
according to the transmission distance. Therefore, the energy model for transmission is also divided
Sensors
into 19, 1494
2019,space
a free 6 of 19
model, which is for short-haul communication, and multi-path fading model, which
is for long distance communication.

Energymodel.
Figure2.2.Energy
Figure model.

Transmission consumption represents the energy the transmission circuit and the amplifier use,
and the reception consumption represents the energy the reception circuit uses. Each sensor will
consume ETx energy to transmit an l-bit data package over the communication distance of d, as is
shown in Equation (2): (
l · Eelec + l ·ε f s ·d2 i f d < d0
ETx (l, d) = (2)
l · Eelec + l ·ε mp ·d4 i f d ≥ d0

where Eelec denotes the energy consumption to run the transmitter or receiver circuit. ε f s and
ε mp denote the amplification coefficient for the free space model and the multi-path fading model,
respectively. d0 is a threshold value and it can be calculated using:
s
ε fs
d0 = (3)
ε mp

Reception consumption represents the energy usage of the reception circuit. The energy
consumption to receive an l-bit data package can be calculated using Equation (4):

ERx (l ) = l · Eelec (4)

4. Our Proposed Routing Schema

4.1. Clustering Formation

We partition the whole sensor field into several sectors and the number of sectors is determined
by the CHs. Each node joins one cluster according to its geographical information. The angle of each
sector is shown as Equation (5):
2π
A= (5)
NCHs
where NCHs denotes the number of CHs. Then, we calculate the cluster that each node should take
part in using Equation (6):
  
arctan x
y
i f x ≥ 0 and y ≥ 0


A 



arctan yx +π
CHid = if x < 0 (6)

 A 

 arctan x +2π
y

i f x ≥ 0 and y < 0

A

where x denotes the position of the sensor in coordinate axis x and y denotes the position of the sensor
in coordinate axis y. We set the center of the sensor field as the origin of coordinates and the network
after clustering is shown as Figure 3.
Sensors 2019, 19, 1494 7 of 19
Sensors 2019, 19, 1494 7 of 19

Figure 3.
Figure 3. Clustering
Clustering formation.
formation.

4.2.
4.2. CH
CH Selection
Selection
We
Weexpect
expectto to select
select the
the nodes
nodes with
with as
as much
much energy
energy asas possible.
possible. Therefore,
Therefore, thethe residual
residual energy
energy of of
nodes
nodes is a significant metric in CH selection. Another important metric for CH selection is the distance
is a significant metric in CH selection. Another important metric for CH selection is the distance
between
between thethe CH
CH andand the
the sink.
sink. CHs
CHs take
take the
the responsibility
responsibility of of transmitting
transmitting their
their members’
members’ data
data toto the
the
sink and a great deal of energy is consumed. Therefore, it is necessary to select the nodes
sink and a great deal of energy is consumed. Therefore, it is necessary to select the nodes with as close with as close
aa distance
distanceto tothe
thesink
sinkasaspossible.
possible.WeWe use three
use different
three methods
different methods to calculate the weight
to calculate of sensors
the weight and
of sensors
the
andresult is discussed
the result in the
is discussed insimulation. We used
the simulation. Equations
We used (7)–(9)
Equations (7),to
(8),calculate
and (9) tothe weight the
calculate of sensors:
weight
of sensors:
Wi = Eresidual_i (7)
𝑊𝑖 = 𝐸𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙_𝑖 (7)
E
𝐸 residual_i
W𝑊i == 𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙_𝑖 (8)
(8)
𝑖 Dis i_to_sink
𝐷𝑖𝑠𝑖_𝑡𝑜_𝑠𝑖𝑛𝑘

E 22
𝐸residual_i
𝑟𝑒𝑠𝑖𝑑𝑢𝑎𝑙_𝑖
i 𝑖=
W𝑊 = Dis (9)
(9)
𝐷𝑖𝑠i_to_sink
𝑖_𝑡𝑜_𝑠𝑖𝑛𝑘

The
The node
nodeclosest
closesttotothe
thesink
sinkwill
willfirst broadcast
first broadcast its its
weight
weightto the other
to the members
other in the
members in cluster and
the cluster
claim itself to be a tentative CH. Only nodes with the higher weight will broadcast
and claim itself to be a tentative CH. Only nodes with the higher weight will broadcast their own their own weight
and
weightbecome a new CH
and become a newcandidate. Finally,Finally,
CH candidate. the node thewith
node the maximal
with weightweight
the maximal will be will
elected as the
be elected
final
as theCH in each
final CH incluster. The network
each cluster. after CH
The network selection
after is shown
CH selection as Figure
is shown 4.
as Figure 4.
Many
Many researches regarding LEACH, TTDD, and CCMAR conduct the selection of
researches regarding LEACH, TTDD, and CCMAR conduct the selection of CHs
CHs during
during
each
each round. However, frequent CH selection will greatly increase the control messagesthat
round. However, frequent CH selection will greatly increase the control messages such suchmuch
that
energy is wasted. In our presented schema, the CHs selection only happens
much energy is wasted. In our presented schema, the CHs selection only happens when any of the when any of the CHs do
not satisfy the condition of maximal weight. During each round, each
CHs do not satisfy the condition of maximal weight. During each round, each member node will member node will upload a
data
uploadpackage
a datathat contains
package thatits weightits
contains forweight
selection
for to its corresponding
selection CH. WhenCH.
to its corresponding CHsWhen
find that
CHstheir
find
member nodes have higher weight than themselves, a re-clustering message
that their member nodes have higher weight than themselves, a re-clustering message will be will be broadcast in those
clusters
broadcast byinthe CHs.clusters by the CHs.
those
Sensors 2019, 19, 1494 8 of 19
Sensors 2019, 19, 1494 8 of 19

Figure 4.
Figure 4. CHs
CHs selection.
selection.

4.3.
4.3. Intracluster
Intracluster Communication
Communication
Long distancecommunication
Long distance communicationusually usually consumes
consumes a great
a great deal deal of energy
of energy and shortens
and shortens the
the lifetime
lifetime of the network. Therefore, it is necessary to avoid long-distance
of the network. Therefore, it is necessary to avoid long-distance communication during the communication during
the intracluster
intracluster communication.
communication. In ourInpresented
our presented schema,
schema, membermember
nodesnodes can calculate
can calculate the costthe
of cost of
energy
energy consumption of different routing paths to select an optimal relay node or transmit
consumption of different routing paths to select an optimal relay node or transmit data directly. As data directly.
As Lambrou
Lambrou et al.
et al. [33]
[33] discusses,
discusses, theenergy
the energyconsumption
consumptionofoftransmission
transmissionisisabout
about 1000
1000 times
times that
that of
of
calculation. Therefore, it is wise for members to choose a better routing path via calculation.
calculation. Therefore, it is wise for members to choose a better routing path via calculation. The The energy
consumption in the routing
energy consumption path using
in the routing pathdirect
usingcommunication can be calculated
direct communication as:
can be calculated as:
2 )2
𝑙 ∙ 𝐸l𝑒𝑙𝑒𝑐
· Eelec++𝑙 l∙·ε𝜀f𝑓𝑠 ∙ 𝑑(𝑆𝑖 , 𝐶𝐻 Si ) 𝑆𝑖 i f d 𝑖𝑓
(Si ,𝑑(𝑆 < 𝑆𝑖d0) < 𝑑0
CH𝑖Si, )𝐶𝐻
(
s · d ( Si , CH
𝐸1 (𝑆𝑖1,(𝐶𝐻
E S i 𝑆𝑖 ) Si
, CH =) {
= 4 4 S , CH ) ≥ d )
(10)
(10)
· Eelec++𝑙 l∙·ε𝜀mp
𝑙 ∙ 𝐸l𝑒𝑙𝑒𝑐 𝑚𝑝·d∙(S 𝑑(𝑆 Si ) 𝑆𝑖i)f d (𝑖𝑓
𝑖 , 𝐶𝐻
i , CH i 𝑑(𝑆Si
𝑖 , 𝐶𝐻𝑆𝑖0 ≥ 𝑑0

Wheredd(𝑆
where (S ,𝑖 ,CH
𝐶𝐻Si
𝑆𝑖 ) denotes the
) denotes the distance
distance between
between node
node ii and
and its
itscorresponding
correspondingCH. CH.E𝐸11((𝑆
Si ,𝑖 ,CH
𝐶𝐻Si
𝑆𝑖 ) or
) or

i
𝐸 (𝑆 , 𝑆 ) represents the direct communication between the source and target nodes.
E11 S𝑖i , S𝑗j represents the direct communication between the source and target nodes. If the source node If the source
node
is is far from
far away awayits from
CH,its CH, anode
a relay relayj will
nodebej will be chosen
chosen to forward
to forward the datathe data package.
package. The total The total
energy
energy consumption for the whole routing can
consumption for the whole routing can be calculated using: be calculated using:

𝐸E2 (𝑆S𝑖 ,,𝑆S𝑗 ,, 𝐶𝐻

CH𝑆𝑖
)
= 𝐸𝑇𝑥 (𝑙, 𝑑(𝑆𝑖 , 𝑆𝑗


)) + 𝐸 (𝑙) + 𝑅 (𝑙, 𝑑(𝑆𝑗 , 𝐶𝐻


= E l, d S , S + E 𝑅𝑥(l ) + R 𝑇𝑥 l, d S , CH 𝑆𝑖 ))
2 i j Si Tx i j
2
Rx Tx
2
j Si (11)
(11)
==
3𝑙3l∙ 𝐸 + ε𝜀·d∙2𝑑S(i ,𝑆S𝑖j, 𝑆+
𝑗 ) ε+
·d2𝜀 ∙S𝑑j , CH
(𝑆𝑗Si, 𝐶𝐻𝑆𝑖 )


· E𝑒𝑙𝑒𝑐
elec +

compare E𝐸1((𝑆
Nodes compare 𝑖 , 𝐶𝐻𝑆𝑖 ) with 𝐸2 (𝑆𝑖 , 𝑆𝑗 , 𝐶𝐻𝑆𝑖 ) and choose the lower one as the routing path.


Nodes 1 Si , CHSi ) with E2 Si , S j , CHSi and choose the lower one as the routing path.
The intracluster
The intracluster communication
communication can canbe
bedetermined
determinedusing: using:

E(𝑆𝑖 ) = min(𝐸1 (𝑆𝑖 , 𝐶𝐻𝑆𝑖 ), 𝐸2 (𝑆𝑖 , 𝑆𝑗 , 𝐶𝐻

𝑆𝑖 )) (2)

E(Si ) = min E1 (Si , CHSi ), E2 Si , S j , CHSi (12)

4.4. Intercluster
4.4. Intercluster Communication
Communication
In this
In this section,
section, we
we use
use a greedy
greedy algorithm
algorithm to
to construct
construct a chain for intercluster
intercluster communication
communication to to
avoid long-distance
avoid long-distancecommunication.
communication.The Theprocess
processofof
thethe formation
formation of of
thethe chain
chain cancan be divided
be divided intointo
the
the following
following steps:
steps:
Step1: The sink broadcasts a Chain_Formation
Step1: Chain_Formation message in the whole network to require all all the
the
CHs to report their ID and position.
CHs to report their ID and position.
Step2: Once the sink receives the information
Step2: information regarding
regarding CHs,
CHs, itit will
will choose
choose the
the closest
closest CH
CH as
as the
the
leader. The
leader. The leader
leader takes
takes the
the responsibility
responsibilityto
totransmit
transmitthethedata
datato
tothe
thesink
sinkdirectly.
directly.
Step3: The greedy algorithm is executed in other CHs and each CH only transmits its data to a
relay CH that is closer to the sink compared with itself.
The result of chain generation is shown in Figure 5.
Sensors 2019, 19, 1494 9 of 19

Step3: The greedy algorithm is executed in other CHs and each CH only transmits its data to a
relay
SensorsCH that
2019, is closer to the sink compared with itself.
19, 1494 9 of 19
Sensors 2019, 19, 1494 9 of 19
The result of chain generation is shown in Figure 5.

Figure 5.
5. Chain of
of CHs generation.
generation.
Figure 5. Chain
Figure Chain of CHs
CHs generation.
4.5.
4.5. Migration
Migration Strategy
Strategy of
of the
the Mobile
Mobile Sink
Sink
4.5. Migration Strategy of the Mobile Sink
In
In our
our proposed
proposed schema,
schema, the the moving
moving trajectory
trajectory ofof the
the mobile
mobile sink
sink is is aa circle,
circle, as
as Figure
Figure 66
In our proposed schema, the moving trajectory of the mobile sink is a circle, as Figure 6
demonstrates.
demonstrates. The The mobile
mobilesink sinkmoves
moveswith
witha aconstant
constantangular
angularvelocity
velocityand and thethe current
current position
position of
demonstrates. The mobile sink moves with a constant angular velocity and the current position of
of the mobile sink can be calculated using its initial position and moving time.
the mobile sink can be calculated using its initial position and moving time. Hence, the mobile sink Hence, the mobile sink
the mobile sink can be calculated using its initial position and moving time. Hence, the mobile sink
only
only needs
needstotoannounce
announceitsitsinitial
initialposition
positionandandangular
angular velocity
velocitysuch thatthat
such thethe
broadcast
broadcast message
messageof the
of
only needs to announce its initial position and angular velocity such that the broadcast message of
network is greatly reduced. We suppose the initial position of the mobile sink is
the network is greatly reduced. We suppose the initial position of the mobile sink is 𝑃0 , and after aP 0 , and after a ∆t time
the network is greatly reduced. We suppose the initial position of the mobile sink is 𝑃0 , and after a
t time interval,
interval, the mobile thesink will sink
mobile havewill
moved
havetomoved to a newPposition
a new position ∆t , as is shown
𝑃 , asinisFigure
shown6.inThe prediction
Figure 6. The
t time interval, the mobile sink will have moved to a new position 𝑃∆𝑡 ∆𝑡 , as is shown in Figure 6. The
of the mobile
prediction ofsink
the demands
mobile sink thatdemands
the clock that
of sensors is synchronized
the clock of sensors isand the position and
synchronized information will
the position
prediction of the mobile sink demands that the clock of sensors is synchronized and the position
be calibrated every few rounds. When there is only one mobile sink in the network,
information will be calibrated every few rounds. When there is only one mobile sink in the network, the initial position
information will be calibrated every few rounds. When there is only one mobile sink in the network,
of
thethe mobile
initial sink Pof
position 0 can be set atsink
the mobile any𝑃position
can be ofsetthe
at circle of ther.circle
with radius
any position Oncewith thereradius
are multiple
r. Once
the initial position of the mobile sink 𝑃00 can be set at any position of the circle with radius r. Once
mobile
there aresinks, their initial
multiple mobilepositions are initial
sinks, their evenlypositions
distributedareat the edge
evenly of the circle
distributed at thewith radius
edge r. circle
of the
there are multiple mobile sinks, their initial positions are evenly distributed at the edge of the circle
with radius r.
with radius r.

Figure 6.
6. Trajectory of
of mobile sink.
sink.
Figure 6. Trajectory
Figure Trajectory ofmobile
mobile sink.
Sensors 2019, 19, 1494 10 of 19

5. Performance Evaluation

5.1. Simulation Environment

In order to test our presented schema, we adopted the MATLAB with the Chinese version of
R2016a simulation environment. Some relevant parameters are shown as Table 1.

Table 1. Simulation parameters.

Parameter Name Value

Network radius (R) [100, 200, 300, 400] m
Mobile sink radius (r) [0, 0.25R, 0.5R, 0.75R, R]
Mobile sink number (MN) [1, 2, 3, 4, 5]
Mobile sink speed (w) [π/20, π/10, π/5]
Cluster number (CN) [3, 4, 5, 6, 7]
Number of nodes (N) 100
Packet length (l) 500 bits
Initial energy (E0 ) 0.5 J
Energy consumption on circuit (Eelec ) 50 nJ/bit
Free-space model parameter (ε f s ) 10 pJ/bit/m2
Multi-path model parameter (ε mp ) 0.0013
q pJ/bit/m4
ε fs
Distance threshold (do ) ε mp m

5.2. Comparation of Different Algorithms

We compared our presented schema with CCMAR and ECDRA. In order to equally evaluate the
performance of different algorithms, we adopted the same network model to execute the algorithms.
As is commonly discussed in similar work, the network lifetime is commonly defined as the time
when the first dead node appears in the network. We compared the network lifetime of different
algorithms and the simulation result is shown as Figure 7. From Figure 7, we can clearly see that
the schema we present had a better performance in the aspect of network lifetime compared to
CCMAR and ECDRA. CCMAR adopts chain structure for intracluster transmission, which causes
the heavy burden of forwarding. Additionally, intercluster communication in CCMAR adopts direct
communication between CHs and the mobile sink, which greatly increases the energy consumption of
CHs. ECDRA also adopts a mobile sink to rotate in the sensor field; however, it only considered the
mobile sink moving in the outer side of the sensor field. Therefore, sensors in the boundary dissipate
more energy and that hastens their demise. Our proposed schema performed best because it calculated
the energy for different routing paths and chooses the optimal one.
Then, we compared the energy consumption between different algorithms and the result is
illustrated as Figure 8. We can clearly see from Figure 8 that our presented schema still had the best
performance. Too much data forwarding resulted in the unnecessary energy dissipation in CCMAR
and that led to a higher energy consumption compared to the other two algorithms. ECDRA consumed
more energy than our presented algorithm because the average communication distance in ECDRA
was much longer than that in our presented schema.
 Sensors 2019, 19, 1494 11 of 19
Sensors 2019, 19, 1494 11 of 19

Figure 7. Comparison of the network lifetime between different algorithms.

Then, we compared the energy consumption between different algorithms and the result is
illustrated as Figure 8. We can clearly see from Figure 8 that our presented schema still had the best
performance. Too much data forwarding resulted in the unnecessary energy dissipation in CCMAR
and that led to a higher energy consumption compared to the other two algorithms. ECDRA
consumed more energy than our presented algorithm because the average communication distance
in ECDRA was much Figure
Figure7.7.Comparison
longer than thatofof
Comparison inthe
thenetwork
ournetworklifetime
lifetime
presented between
betweendifferent
schema. differentalgorithms.
algorithms.

Then, we compared the energy consumption between different algorithms and the result is
illustrated as Figure 8. We can clearly see from Figure 8 that our presented schema still had the best
performance. Too much data forwarding resulted in the unnecessary energy dissipation in CCMAR
and that led to a higher energy consumption compared to the other two algorithms. ECDRA
consumed more energy than our presented algorithm because the average communication distance
in ECDRA was much longer than that in our presented schema.

Comparisonofoftotal
Figure8.8.Comparison
Figure totalenergy
energyconsumption
consumptionbetween
betweendifferent
differentalgorithms.
algorithms.

6. Network Parameters Adjustment and Performance Enhancement

6. Network Parameters Adjustment and Performance Enhancement
In this section, we discuss the influence of different parameters on the performance of the
In this section, we discuss the influence of different parameters on the performance of the
network. The radius of the moving trajectory of the mobile sink has a significant influence on the
network. The radius of the moving trajectory of the mobile sink has a significant influence on the
topology construction. Then, the method for weight calculation determines the usage ratio of the
topology construction. Then, the method for weight calculation determines the usage ratio of the
energy. The number of the clusters and the speed of the sink may also affect the performance.
energy. The number of the clusters and the speed of the sink may also affect the performance.
Figure 8. mobile
Additionally, multiple Comparison
sinksofare
total energy
also consumption
discussed in this between
section. different algorithms.
Additionally, multiple mobile sinks are also discussed in this section.
6. Network
6.1. Parameters
Study of the Adjustment
Radius of the and Performance Enhancement
Moving Trajectory
6.1. Study of the Radius of the Moving Trajectory
Inour
In thisproposed
section, we discuss
schema, thethe influence
mobile of different
sink rotates parameters
in a circle and theon the performance
radius of the
of the circle needs
In our proposed schema, the mobile sink rotates in a circle and the radius of the circle needs to
tonetwork. The radius
be optimized. of the
We first movingthe
changed trajectory ofthe
radius of themoving
mobile sink has a and
trajectory significant influence
kept other on the
parameters
be optimized. We first changed the radius of the moving trajectory and kept other parameters
topology construction.
unchanged. Then,tothe
We set the radius method
0, 0.25R, for0.75R,
0.5R, weight calculation
and R, and thedetermines usage ratio
lifetime of the network of the
is shown
asenergy.
FigureThe number
9. From of the
Figure clusters
9, we found and
that the
whenspeed of the sink moved
the mobile may also affect
along thethe performance.
radius of 0.25R,
Additionally,
the multiplethe
network achieved mobile sinks
longest are also
lifetime. discussed
However, in this
when thesection.
mobile sink ran at the outside of the
circle, the network had the poorest performance.
6.1. Study of the Radius of the Moving Trajectory
In our proposed schema, the mobile sink rotates in a circle and the radius of the circle needs to
be optimized. We first changed the radius of the moving trajectory and kept other parameters
Sensors 2019,
Sensors 19,19,
2019, 1494
1494 1212ofof1919

unchanged.
unchanged.We Wesetsetthe
theradius
radiustoto0,0,0.25R,
0.25R,0.5R,
0.5R,0.75R,
0.75R,and
andR,R,and
andthe
thelifetime
lifetimeofofthethenetwork
networkisisshown
shown
asasFigure
Figure9.9.From
FromFigure
Figure9,9,we
wefound
foundthat thatwhen
whenthe themobile
mobilesink
sinkmoved
movedalongalongthe theradius
radiusofof0.25R,
0.25R,the
the
Sensors 2019, 19, 1494 12 of 19
network
networkachieved
achievedthe thelongest
longestlifetime.
lifetime.However,
However,when whenthe
themobile
mobilesink
sinkranranatatthe
theoutside
outsideofofthe
thecircle,
circle,
the
thenetwork
networkhad hadthethepoorest
poorestperformance.
performance.

Figure
Figure9.9.9.Different
Figure Differentradius
Different radiusof
radius ofmobile
of mobilesink.
mobile sink.
sink.

We
WeWecompared
compared
compared the
the energy
theenergy
energyconsumption
consumption
consumption of of
different radiiradii
ofdifferent
different for single
radiifor mobile
forsingle
single sink. Simulation
mobile
mobile sink. results
sink.Simulation
Simulation
are shown
results
results are inshown
areshownFigure
inin10. We can
Figure 10.
Figure obviously
10.WeWecan see that see
canobviously
obviously the energy
seethat
thatthe consumption
theenergy rose much
energyconsumption
consumption roseslower
rose much
muchwhen
slowerthe
slower
radius
when of
the the
radiusmobile
of thesink was
mobile about
sink 0.25R.
was aboutFurthermore,
0.25R. when
Furthermore, the radius
when of
the the mobile
radius
when the radius of the mobile sink was about 0.25R. Furthermore, when the radius of the mobile sink of sink
the was
mobile about
sink
R,
was
wasthe energy
about
about consumption
R,R,the
theenergy of the network
energyconsumption
consumption ofoftherose rapidly.
network
the network The
rose
rose reasonThe
rapidly.
rapidly. for the
Thereason change
reason for could
forthe be that
thechange
change we
could
could
took
bebethatresidual
thatwe
wetook energy
tookresidualof each
residual energynode
energy into
ofofeach consideration
eachnode
nodeinto such that the
intoconsideration
consideration suchenergy
such that consumption
thatthe
theenergy of the whole
energyconsumption
consumption ofof
network
the
thewhole was
wholenetworknot optimal.
networkwas wasnot
notoptimal.
optimal.

Figure 10. Energy consumption of different radius of single mobile sink.

Figure
Figure10.
10.Energy
Energyconsumption
consumptionofofdifferent
differentradius
radiusofofsingle
singlemobile
mobilesink.
sink.
Additionally, we studied the optimal radius of the moving trajectory under the different network
sizesAdditionally,
and we set thewe
Additionally, westudied
studied
network sizethe
as optimal
the 100, 200, radius
optimal andof400
radius
300, ofthethe moving
moving
meters. trajectory
With trajectoryunder
the network sizethe
under Rthedifferent
different
increasing,
network
network sizes
sizesand
andwewe set the
set network
the network size as
size 100,
as 200,
100, 200,300,
300,and
and 400 meters.
400 meters. With
With
the whole energy consumption increased and the lifetime of the network decreased. As is shown the
thenetwork
network size
sizeR
inR
increasing,
increasing,
Figure the
11, 0.25Rwhole
the whole energy
energy
was still consumption
consumption
the optimal increased
radius increased and the
and the
for the mobile lifetime
sinklifetimeof
under ofthe network
thethe network
different decreased.
decreased.
network As is
As is
scales.
shown
shownininFigure
Figure11,
11,0.25R
0.25Rwas
wasstill
stillthe
theoptimal
optimalradius
radiusfor
forthe
themobile
mobilesink
sinkunder
underthe
thedifferent
differentnetwork
network
scales.
scales.
 Sensors 2019, 19, 1494 13 of 19
Sensors 2019, 19, 1494 13 of 19
Sensors 2019, 19, 1494 13 of 19

Figure 11.Different
Figure11. Differentradius
radiusof
ofmobile
mobilesink
sinkunder
underdifferent
differentnetwork
networksize.
size.
Figure 11. Different radius of mobile
6.2. Study of the Methods of Weight Calculation of CHs sink under different network size.
6.2. Study of the Methods of Weight Calculation of CHs
It is aofgood
6.2. Study choice of
the Methods toWeight
select sensors with
Calculation high residual energy as CHs. However, those nodes
of CHs
It is a good choice to select sensors with high residual energy as CHs. However, those nodes
with high residual energy tend to be very far away from the mobile sink. Consequently, if we only
withIt high
is a good choice
residual to select
energy sensors
tend to be verywith
farhigh
awayresidual
from theenergy
mobile assink.
CHs.Consequently,
However, those nodes
if we only
considered the residual energy of nodes, long distance communication will occur between CHs and
with high residual
considered energyenergy
the residual tend toofbe very long
nodes, far away fromcommunication
distance the mobile sink. Consequently,
will occur between if we
CHs only
and
the sink. Thus, it was necessary to achieve a balance between the residual energy and distance from
considered the residual
the sink. Thus, energy ofto
it was necessary nodes, long
achieve distancebetween
a balance communication will energy
the residual occur between CHs and
and distance from
CHs to mobile sink to prolong the lifetime of the system. We first set the radius of the mobile sink to be
the
CHssink. Thus, itsink
to mobile wastonecessary to achieve
prolong the lifetimeaofbalance between
the system. We the
firstresidual energyofand
set the radius thedistance fromto
mobile sink
0.25R, and then we calculated the weight of each node by using Equations (7)–(9). Simulation results
CHs to mobile
be 0.25R, andsink
thento we
prolong the lifetime
calculated of the system.
the weight of each We first
node byset the radius
using of the(7),
Equations mobile sink (9).
(8) and to
are provided in Figure 12, and it illustrates that it is more reasonable for us to use Equation (9) to
beSimulation
0.25R, and then are
results weprovided
calculated in the weight
Figure of each
12, and node by
it illustrates thatusing Equations
it is more (7), (8)
reasonable for and
us to(9).
use
calculate the weight of nodes.
Simulation
Equation (9) results are provided
to calculate in Figure
the weight 12, and it illustrates that it is more reasonable for us to use
of nodes.
Equation (9) to calculate the weight of nodes.

Figure 12.Different
Figure12. Differentmethods
methodsofofweight
weightcalculation.
calculation.
Figure 12. Different methods of weight calculation.
Sensors 2019, 19, 1494 14 of 19
Sensors 2019, 19, 1494 14 of 19

AAgreat
greatdeal
deal ofof research
research defines
defines the
the network
network lifetime
lifetime as
as the
the time
time when
when the
the first
first node
node dies.
dies.
However, using Equation (8) may make the node’s death premature because
However, using Equation (8) may make the node’s death premature because the distance accounts the distance accounts
for
foraalarge
largeproportion
proportionininthe
theweight
weightcalculation.
calculation.In Inorder
ordertotohighlight
highlightthetheimportance
importanceof ofthe
theresidual
residual
energy,
energy, we adopted Equation (9) as the method of weight calculation under the circumstances the
we adopted Equation (9) as the method of weight calculation under the circumstances that that
initial energy
the initial of each
energy ofnode
each was
nodeless than
was less1 J.than
Simulation results demonstrate
1 J. Simulation that it performed
results demonstrate better in
that it performed
terms
betterofinthe timeofwhen
terms the first
the time whennode dies.node dies.
the first

6.3.
6.3.Study
Studyofofthe
theNumber
NumberofofClusters
Clusters
The
Thenumber
numberofof clusters was
clusters another
was important
another importantparameter of the
parameter ofnetwork. Too few
the network. TooCHs
fewmay
CHscause
may
long-distance communication between cluster members and CHs. However,
cause long-distance communication between cluster members and CHs. However, excessive excessive CHs caused
CHs
oversized chains tochains
caused oversized be generated during intracluster
to be generated communication.
during intracluster Plenty ofPlenty
communication. energyofwill be wasted
energy will be
towasted
forward the data packages and the network latency will greatly increase. We set different
to forward the data packages and the network latency will greatly increase. We set different cluster
numbers in our schema
cluster numbers in ourwhen
schemathewhen
radiustheof the mobile
radius sink
of the was 0.25R,
mobile and the
sink was simulation
0.25R, and the results are
simulation
shown in Figure 13. When the cluster number was 5, the network had a better performance,
results are shown in Figure 13. When the cluster number was 5, the network had a better performance, especially
with regardwith
especially to the time when
regard to the the first
time node
when thedied.
first node died.

Figure 13.Different
Figure13. Differentnumber
numberofofclusters.
clusters.
6.4. Study of the Speed of the Mobile Sink
6.4. Study of the Speed of the Mobile Sink
We also studied the speed of the mobile sink. As the sink is carried by vehicles, such as the
We also studied the speed of the mobile sink. As the sink is carried by vehicles, such as the
intelligent car, its speed is controllable. Therefore, we set the angular velocity of the mobile sink as
intelligent car, its speed is controllable. Therefore, we set the angular velocity of the mobile sink as
π/20, π/10, and π/5, and the radius of the mobile sink was 0.25R. The simulation results are shown
π/20, π/10, and π/5, and the radius of the mobile sink was 0.25R. The simulation results are shown in
in Figure 14. It can be clearly seen that the lifetime of the network was almost unchanged with the
Figure 14. It can be clearly seen that the lifetime of the network was almost unchanged with the
increasing speed of the mobile sink. Therefore, we draw a conclusion that the speed of the mobile sink
increasing speed of the mobile sink. Therefore, we draw a conclusion that the speed of the mobile
has little influence on the lifetime of the network, and it should not be too fast or too slow.
sink has little influence on the lifetime of the network, and it should not be too fast or too slow.
 Sensors 2019, 19, 1494 15 of 19
Sensors 2019, 19, 1494 15 of 19
Sensors 2019, 19, 1494 15 of 19

Figure 14.Different
Different speedof
of themobile
mobile sink.
Figure 14. Differentspeed
Figure14. speed ofthe
the mobilesink.
sink.
6.5.
6.5.Study
Studyofof
ofMultiple
MultipleMobile
MobileSinks
Sinks
6.5. Study Multiple Mobile Sinks
We
Wealso
alsostudied
studied thethe
influence
influenceof multiple mobile
of multiple
multiple sinks sinks
mobile on theonlifetime of the network.
the lifetime
lifetime of the The moving
the network.
network. The
We also studied the influence of mobile sinks on the of The
radius
moving of each
radius sink
of was
each set
sink to 0.25R.
was set Simulation
to 0.25R. results
Simulation illustrate
results that with
illustrate the
thatincreasing
with the numbers
increasing
moving radius of each sink was set to 0.25R. Simulation results illustrate that with the increasing
of mobile sinks,
numbers the lifetime
of mobile
mobile sinks, the
theoflifetime
the network theimproved,
of the as is shown
network improved,
improved, asin
is Figure
shown15. We also15.
in Figure
Figure noticed
We alsoa
also
numbers of sinks, lifetime of network as is shown in 15. We
trend that with
noticed aa trend the
trend that number
that with of
with the mobile
the number sinks
number of increasing,
of mobile
mobile sinks the improvement
sinks increasing,
increasing, the of the lifetime
the improvement
improvement of becomes
of the fewer
the lifetime
lifetime
noticed
and fewer.fewer
becomes Sinceandthe fewer.
mobileSince
sinks the
are mobile
more expensive
sinks are than
more ordinal
expensivesensors,
than three mobile
ordinal sinksthree
sensors, are
becomes fewer and fewer. Since the mobile sinks are more expensive than ordinal sensors, three
recommended
mobile sinks
sinks areto be
are the most cost-effective.
recommended to be
be the
the most
most cost-effective.
cost-effective.
mobile recommended to

Figure 15.Different
Figure15. Differentmobile
mobilesink
sinknumber.
number.
Figure 15. Different mobile sink number.
We
Wecompared
comparedthe energy
the energy consumption
consumption of multiple
of multiplemobile
multiple sinks
mobile when
sinks the radius
when of theofmobile
the radius
radius sink
the mobile
mobile
was We compared
0.25R. It can the
clearly energy
be seenconsumption
from Figure of
16 that when mobile
the sinks
number when
of the the
mobile of the
sinks increased,
sink was
sink was 0.25R.
0.25R. ItIt can
can clearly
clearly be
be seen
seen from
from Figure
Figure 1616 that
that when
when the
the number
number of of the
the mobile
mobile sinks
sinks
the upward
increased, thetrend of
the upward the energy
upward trend
trend of consumption
of the
the energy of the
energy consumptionwhole
consumption of network
of the
the wholedecreased.
whole network When
network decreased.the number
decreased. When
When theof
the
increased,
the mobile sinks exceeded three, the performance of the whole network does not change much.
number of
number of the
the mobile
mobile sinks
sinks exceeded
exceeded three,
three, the
the performance
performance ofof the
the whole
whole network
network does
does not
not change
change
much.
much.
Sensors
Sensors 2019, 19,
Sensors 2019,
2019, 19,1494
19, 1494
1494 16 of
16
16 of 19
19
19

Figure 16. Energy consumption of different numbers of mobile sinks.

Figure 16. Energy consumption of different numbers of mobile sinks.
7. Discussions and Open Research Issues
7. Discussions and Open Research Issues
7.1. Uneven Energy Distribution Between Clusters
7.1. Uneven Energy
The results Distribution
show that ourBetween Clusters
proposed algorithm had good performance in terms of its lifetime;
however, we found
The results a common
show that ourphenomenon that the energy
proposed algorithm distribution
had good between
performance clusters
in terms of was uneven.
its lifetime;
As is shown
however, weinfound
the Figure 17, some
a common clusters hadthat
phenomenon no node alive, whereas
the energy the nodes
distribution betweenin some clusters
clusters was
uneven.
were As is The
all alive. shown in the Figure
following reasons17, some
may clusters
cause had no node alive, whereas the nodes in some
the problem:
clusters were all alive. The following reasons may cause the problem:
• Sensors were deployed unevenly and each cluster could have different number of nodes.
• Sensors were deployed unevenly and each cluster could have different number of nodes.
• The average distance between nodes and the regional center in one cluster could be longer than
• The average distance between nodes and the regional center in one cluster could be longer than
that in other clusters.
that in other clusters.

Figure 17. Uneven energy distribution between clusters.

Figure 17. Uneven energy distribution between clusters.
Sensors 2019, 19, 1494 17 of 19

Node deployment had a great influence on the performance of the network. The study of
node deployment strategies has been a hot research issue in recent years since it can improve
the performance of WSNs in the aspects of energy consumption, network lifetime, and coverage
rate. Homogeneous deployment of sensor nodes, if possible, can eliminate the phenomenon
mentioned above.

7.2. Open Research Issues

It has been proven that the optimal sink moving trajectory problem is an NP-hard issue [34,35]
that is, by definition, hard to solve. More importantly, the sink mobility issue is tightly related with
routing algorithms used therein. Thus, jointly optimizing mobile sink strategies and routing algorithms
is one of the primary research challenges. Not only does a proper moving trajectory, cluster head
selection, sink number, speed, moving pattern, and sojourn positions need to be carefully designed,
but also the corresponding routing algorithms need to be selected for different applications.
Security is always an important and challenging research issue for WSNs. Restricted to the
limitation of resources and deployment environment, it is difficult to implement strong security
algorithms [36]. Besides, due to the potential threatens like data modification and malicious attacks,
communications between source and destination nodes may have serious influence on the network
security and increase the energy consumption during data transmission. Using sink mobility
technology in WSNs, various types of data reports may get lost or modified during data transmission
due to the change of the location of the sink and selective forwarding. Therefore, security as an
essential component for data transmission in dynamic sensor networks environment needs further
research attention. It is likely that the security for store and forward [37] will help in this regard.
In many location-based applications, data obfuscation technologies can be used to protect location
privacy to avoid the damage caused by having the position leaked. By reducing location accuracy,
data obfuscation technology can achieve the fuzzy effect of the locations of multiple data information.
However, the safety of this technology remains to be further researched. For example, a general
three-tier security framework has been proposed [38] where two separate key pools are respectively
designed for a mobile sink node and pair-wise key establishment between sensor nodes. This scheme
can effectively protect sensor networks from stationary access node replication attacks.

8. Conclusions
Clustering and sink mobility technology as effective methods to improve the performance of
WSNs have drawn much research attention recently. In this paper, we first divided the whole sensor
field into several sectors with the same size, and each sector selected a CH according to its members’
weight. Cluster members calculated the routing path with the optimal energy consumption to transmit
data to their corresponding CHs via single or multi hops communication. Then, the CHs adopted
a greedy algorithm to form a chain for intercluster communication. The latest routing schemas,
such as CCMAR and ECDRA, were compared with our presented algorithm regarding aspects of
energy consumption and network lifetime by simulation. The simulation results proved our presented
algorithm outperformed the other two algorithms. Additionally, we explored the influence of different
parameters on the performance of the network and further improved its performance.

Author Contributions: J.W. and H.-J.K. conceived and designed the experiments; Y.G. performed the experiments;
W.L. and A.K.S. analyzed the data and helped to revise this paper, and J.W. wrote this paper.
Funding: This work was supported by the National Natural Science Foundation of China (61772454, 61811530332,
61811540410, U1836208). Professor Hye-Jin Kim is the corresponding author.
Conflicts of Interest: The authors declare no conflict of interest.
Sensors 2019, 19, 1494 18 of 19

References
1. Akyildiz, I.F.; Su, W.; Sankarasubramaniam, Y.; Cayirci, E. Wireless sensor networks: A survey. Comput. Netw.
2002, 38, 393–422. [CrossRef]
2. Zeng, D.; Dai, Y.; Li, F.; Sherratt, R.S.; Wang, J. Adversarial learning for distant supervised relation extraction.
Comput. Mater. Contin. 2018, 55, 121–136.
3. Pan, J.S.; Kong, L.; Sung, T.W.; Tsai, P.W.; Snasel, W. α-Fraction First Strategy for Hirarchical Wireless Sensor
Neteorks. J. Internet Technol. 2018, 19, 1717–1726.
4. Chanak, P.; Indrajit Banerjee, I.; Wang, J.; Sherratt, R.S. Obstacle Avoidance Routing Scheme through Optimal
Sink Movement for Home Monitoring and Mobile Robotic Consumer Devices. IEEE Trans. Consum. Electron.
2014, 60, 23–29. [CrossRef]
5. Wang, J.; Zhang, Z.; Li, B.; Lee, S.; Sherratt, R.S. An Enhanced Fall Detection System for Elderly Person
Monitoring using Consumer Home Networks. IEEE Trans. Consum. Electron. 2014, 60, 23–29. [CrossRef]
6. Al-Karaki, J.N.; Kamal, A.E. Routing techniques in wireless sensor networks: A survey. IEEE Wirel. Commun.
2004, 11, 6–28. [CrossRef]
7. Yin, C.; Xi, J.; Sun, R.; Wang, J. Location privacy protection based on differential privacy strategy for big data
in industrial internet of things. IEEE Trans. Ind. Inform. 2018, 14, 3628–3636. [CrossRef]
8. Wang, J.; Gao, Y.; Yin, X.; Li, F.; Kim, H. An Enhanced PEGASIS Algorithm with Mobile Sink Support for
Wireless Sensor Networks. Wirel. Commun. Mob. Comput. 2018, 2018, 9472075. [CrossRef]
9. Yao, J.; Zhang, K.; Yang, Y.; Wang, J. Emergency vehicle route oriented signal coordinated control model
with two-level programming. Soft Comput. 2018, 22, 4283–4294. [CrossRef]
10. Ren, Y.; Liu, Y.; Ji, S.; Sangaiah, A.K.; Wang, J. Incentive Mechanism of Data Storage Based on Blockchain for
Wireless Sensor Networks. Mob. Inf. Syst. 2018. [CrossRef]
11. Li, X.; Yang, J.L.; Nayak, A.; Stojmenovic, I. Localized geographic routing to a mobile sink with guaranteed
delivery in sensor networks. IEEE J. Sel. Area Commun. 2012, 30, 1719–1729. [CrossRef]
12. Wang, J.; Yang, X.; Li, B.; Lee, S.Y.; Jeon, S. A mobile sink based uneven clustering algorithm for wireless
sensor networks. J. Internet Technol. 2013, 14, 895–902.
13. Wang, J.; Cao, J.; Ji, S.; Park, J.H. Energy-efficient cluster-based dynamic routes adjustment approach for
wireless sensor networks with mobile sinks. J. Supercomput. 2017, 73, 3277–3290. [CrossRef]
14. Khan, A.W.; Abdullah, A.H.; Anisi, M.H.; Bangash, J.I. A Comprehensive Study of Data Collection Schemes
Using Mobile Sinks in Wireless Sensor Networks. Sensors 2014, 14, 2510–2548. [CrossRef] [PubMed]
15. Heinzelman, W.R.; Chandrakasan, A.; Balakrishnan, H. Energy-efficient communication protocol for wireless
microsensor networks. Ad Hoc Sens. Wirel. Netw. 2000, 2. [CrossRef]
16. Lindsey, S.; Raghavendra, C.S. PEGASIS: Power efficient gathering in sensor information systems. Proc. IEEE
Aerosp. Conf. 2003, 3, 1125–1130.
17. Younis, O.; Fahmy, S. HEED: A Hybrid, Energy-efficient, Distributed Clustering Approach for Ad hoc Sensor
Networks. IEEE Trans. Mob. Comput. 2004, 3, 366–379. [CrossRef]
18. Zhao, M.; Yang, Y.; Wang, C. Mobile Data Gathering with Load Balanced Clustering and Dual Data Uploading
in Wireless Sensor Networks. IEEE Trans. Mob. Comput. 2015, 14, 770–785. [CrossRef]
19. Zhu, C.; Wu, S.; Han, G.; Shu, L.; Wu, H. A Tree-Cluster-Based Data-Gathering Algorithm for Industrial
WSNs With a Mobile Sink. IEEE Access. 2015, 3, 381–396. [CrossRef]
20. Xie, G.; Pan, F. Cluster-Based Routing for the Mobile Sink in Wireless Sensor Networks with Obstacles.
IEEE Access 2016, 4, 2019–2028. [CrossRef]
21. Velmani, R.; Kaarthick, B. An efficient cluster-tree based data collection scheme for large mobile wireless
sensor networks. IEEE Sens J. 2015, 15, 2377–2390. [CrossRef]
22. Sasirekha, S.; Swamynathan, S. Cluster-chain mobile agent routing algorithm for efficient data aggregation
in wireless sensor network. J. Commun. Netw. 2017, 19, 392–401.
23. Wang, J.; Cao, Y.; Li, B.; Kim, H.J.; Lee, S. Particle Swarm Optimization based Clustering Algorithm with
Mobile Sink for WSNs. Future Gener. Comput. Syst. 2017, 76, 452–457. [CrossRef]
24. Wang, J.; Ju, C.; Gao, Y.; Sangaiah, A.K.; Kim, G. A PSO based Energy Efficient Coverage Control Algorithm
for Wireless Sensor Networks. Comput. Mater. Contin. 2018, 56, 433–446.
Sensors 2019, 19, 1494 19 of 19

25. Wang, J.; Gao, Y.; Liu, W.; Sangaiah, A.K.; Kim, H.J. An Improved Routing Schema with Special Clustering
using PSO Algorithm for Heterogeneous Wireless Sensor Network. Sensors 2019, 19, 671. [CrossRef]
[PubMed]
26. Abo-Zahhad, M.; Ahmed, S.M.; Sabor, N.; Sasaki, S. Mobile Sink-Based Adaptive Immune Energy
Efficient Clustering Protocol for Improving the Lifetime and Stability Period of Wireless Sensor Networks.
IEEE Sens. J. 2015, 15, 4576–4586. [CrossRef]
27. Wang, J.; Cao, J.; Sherratt, R.S.; Park, J.H. An improved ant colony optimization-based approach with mobile
sink for wireless sensor networks. J. Supercomput. 2018, 74, 6633–6645. [CrossRef]
28. Tashtarian, F.; Hossein, Y.M.M.; Sohraby, K.; Effati, S. On maximizing the lifetime of wireless sensor networks
in event-driven applications with mobile sinks. IEEE Trans. Veh. Technol. 2015, 64, 3177–3189. [CrossRef]
29. Gao, Y.; Wang, J.; Wu, W.; Sangaiah, A.K.; Lim, S. A Hybrid Method for Mobile Agent Moving Trajectory
Scheduling using ACO and PSO in WSNs. Sensors 2019, 19, 575. [CrossRef] [PubMed]
30. Wang, J.; Zuo, L.; Shen, J.; Li, B.; Lee, S. Multiple mobile sink-based routing algorithm for data dissemination
in wireless sensor networks. Concurr. Comput.-Pract. Exp. 2015, 27, 2656–2667. [CrossRef]
31. Banerjee, T.; Agrawal, D.P. Increasing lifetime of wireless sensor networks using controllable mobile cluster
heads. Wirel. Commun. Mob. Comput. 2010, 10, 313–336.
32. Wang, J.; Gao, Y.; Liu, W.; Wu, W.; Lim, S. An Asynchronous Clustering and Mobile Data Gathering Schema
based on Timer Mechanism in Wireless Sensor Networks. Comput. Mater. Contin. 2019, 58, 711–725.
33. Lambrou, T.P.; Panayiotou, C.G. A survey on routing technique supporting mobility in sensor networks.
In Proceedings of the 5th International Conference on Mobile Ad-Hoc and Sensor Networks, Fujian, China,
14–16 December 2009; pp. 78–85.
34. Vukasinovic, I.; Babovic, Z.; Rakocevic, G. A survey on the use of mobile agents in wireless sensor networks.
In Proceedings of the IEEE Conference on Industrial Technology, Athens, Greece, 19–21 March 2012;
pp. 271–277.
35. Luo, H.; Ye, F.; Cheng, J.; Lu, S.; Zhang, L. TTDD: Two-tier data dissemination in large-scale wireless sensor
networks. Wirel. Netw. 2005, 11, 161–175. [CrossRef]
36. Zhou, Y.; Fang, Y.; Zhang, Y. Securing wireless sensor networks: A survey. IEEE Commun. Surv. Tutor. 2008,
10, 6–28. [CrossRef]
37. Díaz-Sánchez, D.; Sherratt, R.S.; Almenarez, F.; Arias, P.; Marín, A. Secure store and forward proxy for
dynamic IoT applications over M2M networks. IEEE Trans. Consum. Electron. 2017, 62, 389–397. [CrossRef]
38. Rasheed, A.; Mahapatra, R.N. The three-tier security scheme in wireless sensor networks with mobile sinks.
IEEE Trans. Parallel Distrib. Syst. 2012, 23, 958–965. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).