sensors
Article
An Energy-Efficient Clustering Routing Protocol
Based on a High-QoS Node Deployment with an
Inter-Cluster Routing Mechanism in WSNs
Kaida Xu 1 , Zhidong Zhao 1,2,3, *, Yi Luo 1 , Guohua Hui 4 and Liqin Hu 5
1
2
3
4
5
*
School of Communication Engineering, Hangzhou Dianzi University, Hangzhou 310018, China;
xukaidahdu@gmail.com (K.X.); luoyi@hdu.edu.cn (Y.L.)
Hangdian Smart City Research Center of Zhejiang Province, Hangzhou Dianzi University,
Hangzhou 310018, China
College of Electronics and Information, Hangzhou Dianzi University, Hangzhou 310018, China
Key Laboratory of Forestry Intelligent Monitoring and Information Technology of Zhejiang Province,
School of Information Engineering, Zhejiang A & F University, Linan 311300, China;
guohua_hui@zafu.edu.cn
Department of Construction Engineering, Zhejiang College of Construction, Hangzhou 311231, China;
huliqin98@gmail.com
Correspondence: zhaozd@hdu.edu.cn; Tel.: +86-135-8804-5453
Received: 29 April 2019; Accepted: 16 June 2019; Published: 19 June 2019
Abstract: Currently, wireless sensor network (WSN) protocols are mainly used to achieve low power
consumption of the network, but there are few studies on the quality of services (QoS) of these
networks. Coverage can be used as a measure of the WSN’s QoS, which can further reflect the quality
of data information. Additionally, the coverage requirements of regional monitoring target points
are different in real applications. On this basis, this paper proposes an energy-efficient clustering
routing protocol based on a high-QoS node deployment with an inter-cluster routing mechanism
(EECRP-HQSND-ICRM) in WSNs. First, this paper proposes formula definitions for information
integrity, validity, and redundancy from the coverage rate and introduces a node deployment strategy
based on twofold coverage. Then, in order to satisfy the uniformity of the distribution of cluster
heads (CHs), the monitoring area is divided into four small areas centered on the base station (BS),
and the CHs are selected in the respective cells. Finally, combined with the practical application of
the WSN, this paper optimizes the Dijkstra algorithm, including: (1) nonessential paths neglecting
considerations, and (2) a simultaneous introduction of end-to-end weights and path weights, achieving
the selection of optimal information transmission paths between the CHs. The simulation results
show that, compared with the general node deployment strategies, the deployment strategy of
the proposed protocol has higher information integrity and validity, as well as lower redundancy.
Meanwhile, compared with some classic protocols, this protocol can greatly reduce and balance
network energy consumption and extend the network lifetime.
Keywords: wireless sensor network; network coverage; quality of services; inter-cluster routing
mechanism; Dijkstra algorithm
1. Introduction
The wireless sensor network (WSN), which has exerted a great influence in the 21st century,
combines several advanced technologies, such as wireless communication and sensor technology,
to effectively achieve an interaction between human society and the physical world, resulting in a
profound impact on promoting social progress. WSNs currently have a wide range of applications in
Sensors 2019, 19, 2752; doi:10.3390/s19122752
www.mdpi.com/journal/sensors
Sensors 2019, 19, 2752
2 of 23
military defense, environmental monitoring, and smart homes. As the basic component of a WSN,
the sensor node is generally powered by a dry battery. Once the battery is exhausted, the node
is dead and cannot participate in subsequent data operations. If these nodes are placed in harsh
environments, it becomes impractical to replace the batteries. Therefore, how to effectively reduce the
energy consumption of sensor nodes for long-term operation has become a research hotspot [1].
It was under this context that the WSN clustering routing protocol came into being. The WSN
protocol combines a large number of sensor nodes into clusters to form a hierarchical management
system from cluster members (CMs) to cluster heads (CHs) to base stations (BS), which can effectively
reduce the energy consumption of the network, thus extending the lifetime of the network and
achieving more rounds of data iteration.
In recent years, researchers in various countries have proposed a variety of WSN clustering
routing protocols. The most classic protocol is LEACH (Low-Energy Adaptive Cluster Hierarchical) [2],
which achieves the goal of energy balance by periodically changing the CHs in the area. However,
because it adopts a random number mechanism, the suitability of the CH is full of uncertainties.
Hosen and Cho [3] proposed an energy center-based WSN routing mechanism that mainly grades
all nodes in the area according to the residual energy of the node and the average distance from the
member nodes then selects higher-level nodes to become the CHs. A hierarchical routing protocol
based on a k-d tree algorithm was proposed in [4], which uses a spatial partitioned data structure to
organize the nodes into clusters. A multi-aware query-driven (MAQD) routing protocol based on a
neuro-fuzzy inference system was introduced in [5]. An energy-aware cluster-based CRSN routing
protocol (EACRP) was introduced in [6], which considers several factors such as the energy and the
dynamic spectrum.
A multi-access WSN routing protocol based on the transmission power control applied to
underground coal mines was introduced in [7], which uses the transmission power control algorithm
to find the optimal transmission radius and transmission power, and the non-uniform clustering
idea is used to optimize the CH selection mechanism. An energy chain-based WSN routing protocol
(E-CBCCP) for underwater environments was introduced in [8].
Bozorgi et al. [9] combined static clustering and dynamic clustering, then took the nodes’ residual
energy, the energy required to receive the information, and the number of neighbor nodes into
consideration to select the CHs. The simulation results showed that, compared with other methods,
this method could improve the stability and efficiency of the network. The CH selection mechanism
introduced in [10] was divided into three stages: only the advanced node was selected as the CH in
Stage 1; all the nodes performed CH selection with the same probability in Stage 2; and the planar
topology replaced the clustering topology in Stage 3. The cluster structure optimization problem
was divided into multiple sub-optimization problems and solved by multi-objective evolutionary
algorithms (MOEAs) in [11]. Combining a new network structure model with the original energy
consumption model, a new method to determine the optimal number of clusters was proposed in [12],
and through the AGglomerative NESting (AGNES) algorithm, including: (1) introduction of distance
variance, (2) the dual-cluster heads (D-CHs) division of the energy balance strategy, and (3) the node
dormancy mechanism, it can achieve a reduction in the network energy consumption decay rate and
prolong the network lifetime.
In the process of cluster construction, we can also consider the use of principal component
analysis [13], fuzzy logic [1], K-means [14], Markov models [15], the introduction of mobile nodes [16,17],
etc., to construct a reliable and secure sensor network [18–20].
Meanwhile, it is necessary to consider the optimization problem of the information transmission
path between the CHs in the clustering process, which must involve the concept of multi-hops [21].
Common path optimization algorithms include the Dijkstra algorithm [22–24], the ant colony
optimization (ACO) algorithm [25], etc. Through the proper path optimization, the phenomenon of
premature death of the CHs and information congestion [26] could be alleviated effectively.
Sensors 2019, 19, 2752
3 of 23
In addition, a large amount of data is generated during the WSN’s environmental monitoring [27].
How to obtain the most complete and effective information possible while minimizing the amount of
redundant information is related to the WSN’s quality of service (QoS) [28]. How to build a compliant
WSN through a reasonable node deployment [17], [29,30] has also become one of the factors to consider
in the protocol. In [31], a clustering technique is proposed that uses the fuzzy c-means clustering
method for the formation of clusters and a cluster head is selected based on four parameters: sensor’s
location within cluster, location with respect to the fusion center (FC), its signal-to-noise (SNR), and its
residual energy.
In real life, a WSN often faces two problems: (1) There are often some monitoring targets in the
monitoring area that have key roles. Compared with other points, these targets often have higher
coverage requirements. Therefore, for these points we need to arrange more sensor nodes around
them to meet their coverage requirements. (2) With the continuous operation of the WSN, the CH
far from the BS tends to die prematurely due to the long transmission distance. Therefore, how to
build an energy-efficient and reliable information transmission path between the CHs is also worthy
of our attention. It is worth noting that the above two problems are still in [12]. The random node
deployment strategy in [12] does not guarantee the obtained information’s QoS and thus must not
meet the requirement of the QoS for the actual application. In addition, it does not consider the
inter-cluster routing mechanism, which will cause some CHs to deplete energy prematurely because of
the long transmission distances from the BS. From the perspective of the two aspects above, this paper
proposes an energy-efficient clustering routing protocol based on a high-QoS node deployment with an
inter-cluster routing mechanism (EECRP-HQSND-ICRM). The basic premise is as follows: For all the
monitoring target points with a coverage requirement of two, this paper proposes a series of definition
formulas for information integrity, validity, and redundancy and a node deployment strategy based
on twofold coverage. Subsequently, in order to satisfy the uniformity of the CH distribution, the BS
is divided into four small cells, and the CHs are selected in the respective cells. Finally, combined
with the practical application of the WSN, this paper optimizes the Dijkstra algorithm, including:
(1) nonessential path neglecting considerations, and (2) a simultaneous introduction of end-to-end
weights and path weights, achieving the selection of optimal information transmission paths between
the CHs.
The main contributions of this paper include the following:
•
•
•
•
A proposal for calculating the information integrity, validity, and redundancy based on coverage.
The determination of a node deployment strategy based on twofold coverage.
The construction of the CH selection mechanism through partition based on the uniformity of the
CH distribution, the residual energy of the nodes, and the distances from the nodes to the BS.
The optimization of Dijkstra to achieve Dijkstra’s computational complexity reduction.
The remainder of this paper is organized as follows: Section 2 introduces a node deployment
strategy based on twofold coverage. Section 3 describes the details of the protocol in the paper. The
simulation study is conducted in Section 4. Finally, Section 5 summarizes the research and prospects
for the future.
2. The Node Deployment Strategy Based on 2-Fold Coverage
In this section, we propose a series of calculation formulas for the coverage-based QoS metrics,
including information integrity, validity, and redundancy, according to the phenomenon of the different
coverage requirements of the monitoring target points in the real area. At the same time, we also
determine the node deployment strategy based on 2-fold coverage.
2.1. Coverage
As a key factor of the WSN’s QoS, coverage is defined as the proportion of target points that
all sensor nodes can monitor to the number of all target points in the area under the 0-1 perception
Sensors 2019, 19, 2752
4 of 23
Sensors
Sensors2019
2019
44 ofof 24
24
model in this paper, where the 0-1 perception model is a typical idealized model. The target within
perceived
radius
of
sensor
node
can
be
perceived
by
node,
but
the
target
outside
the
perceived
radius
ofthe
the
sensor
node
cancan
be100%
100%
perceived
bythe
thethe
node,
butbut
thethe
target
outside
the
the
perceived
radius
of the
sensor
node
be 100%
perceived
by
node,
target
outside
perceived
radius
cannot
be
perceived
by
the
sensor
node.
Figure
1
shows
a
typical
0-1
perception
perceived
radius
cannot
be
perceived
by
the
sensor
node.
Figure
1
shows
a
typical
0-1
perception
the perceived radius cannot be perceived by the sensor node. Figure 1 shows a typical 0-1 perception
model
diagram.
modeldiagram.
diagram.
model
0%
0%
100%
100%
Figure
Figure1.1.AAtypical
typical0-1
0-1perception
perceptionmodel
modeldiagram.
diagram.
The
coverage
isisas
as
follows:
Thecalculation
calculationformula
formulafor
forthe
thenetwork
networkcoverage
coverageis
asfollows:
follows:
P
(Ts)∆s
TT
ss ΔΔss

 (( ))
Coverage === P
.
Coverage
Coverage
∆sΔ s
Δ
 s
s∈S 
s∈S
s∈
sarea
∈SS
area
area
..
(1)
(1)
(1)
area
s∈
s ∈SSarea
area
In the above formula, s represents the monitoring target point, ∆S represents the area occupied
ΔΔSS representsthe
In
the
ssrepresents
the
target
Inmonitoring
theabove
aboveformula,
formula,
represents
themonitoring
monitoring
targetpoint,
point,
thearea
area
occupied
by the
target point,
Sarea represents
the monitoring
area, andrepresents
T(s) represents
the occupied
coverage
by
the
monitoring
target
point,
S
area
represents
the
monitoring
area,
and
T(s)
represents
the
coverage
by the monitoring
target point,
Sarea
represents
the monitoring
area, and T(s)
represents
the coverage
condition
of the monitoring
target
point
in the area.
If it can be monitored
by the
surrounding
sensor
condition
of
the
monitoring
target
point
in
the
area.
If
it
can
be
monitored
by
the
surrounding
sensor
condition
of
the
monitoring
target
point
in
the
area.
If
it
can
be
monitored
by
the
surrounding
sensor
nodes, then T(s) = 1; otherwise, T(s) = 0.
nodes,
nodes,then
thenT(s)
T(s)==1;1;otherwise,
otherwise,T(s)
T(s)==0.0.
2.2. The Difference in Coverage Requirements
2.2.
2.2.The
TheDifference
Differencein
inCoverage
CoverageRequirements
Requirements
In real-world applications, some key monitoring targets in the area often require more sensor
In
real-world
monitoring
in
area
often
more
sensor
nodes
monitor, applications,
that
is, their some
coverage
are higher
Figure
2 shows
a
Into
real-world
applications,
somekey
keyrequirements
monitoringtargets
targets
inthe
thethan
areaothers.
oftenrequire
require
more
sensor
nodes
that
is,
coverage
requirements
others.
Figure
shows
3-coverage
example.
If the
coverage
requirement
of the are
monitoring
target
is 2, then
the22information
nodesto
tomonitor,
monitor,
that
is,their
their
coverage
requirements
arehigher
higherthan
than
others.
Figure
showsaa33coverage
example.
IfIfthe
the
coverage
requirement
of
the
monitoring
target
then
the
information
will
will
be redundant.
number
coverage
example.If
thecoverage
coveragerequirement
requirementof
ofthe
themonitoring
monitoringtarget
targetisisis3,2,2,then
thenthe
thecoverage
information
will
be
redundant.
IfIfthe
coverage
requirement
of
the
monitoring
target
then
coverage
of
target point
meets
its requirement.
requirement
thethe
monitoring
target
is of
4,
bethe
redundant.
the
coverage
requirementIf
ofthe
thecoverage
monitoring
targetisis3,3,of
then
the
coveragenumber
number
of
the
point
meets
its
thetarget
target
point
meets
itsrequirement.
requirement.
Ifthe
thecoverage
coveragerequirement
requirementof
ofthe
themonitoring
monitoringtarget
targetisis4,4,then
then
then
a partial
loss
of valid
information If
occurs.
aapartial
partialloss
lossof
ofvalid
validinformation
informationoccurs.
occurs.
Monitoring
Monitoring
Target
Target
Monitoring
Monitoring
Range
Range
Sensor
Sensor
Nodes
Nodes
Figure
2.2.Threefold
Figure2.
Threefoldcoverage.
coverage.
Figure
Threefold
coverage.
2.3.
2.3.Coverage-Based
Coverage-BasedQoS
QoSMetrics
Metrics
In
Inthe
theprocess
processof
ofdefining
definingthe
thecoverage-based
coverage-basedQoS
QoSmetrics,
metrics,this
thispaper
paperisisbased
basedon
onthe
thefollowing
following
three
assumptions:
three assumptions:

•

•

During
each
iteration
round,
the
data
each
sensor
node
perceives
from
During
iteration
round,
the amount
of dataof
each
sensor
node
perceives
the environment
Duringeach
each
iteration
round,
the amount
amount
of
data
each
sensor
nodefrom
perceives
from the
the
environment
is
the same. isisthe
environment
thesame.
same.
the
0-1
perception
model.
Thesensory
sensorymodel
modelofofthe
thesensor
sensornode
nodeadopts
adoptsthe
the0-1
0-1perception
perceptionmodel.
model.
The
The
area
contains
a
large
number
of
monitoring
target
points,
and
The area contains a large number of monitoring target points, andthe
theoccupied
occupiedareas
areasof
ofthese
these
points
pointsare
areequal
equalin
insize.
size.
Sensors 2019, 19, 2752
•
5 of 23
The area contains a large number of monitoring target points, and the occupied areas of these
points are equal in size.
(1) Information Integrity: Information integrity is the percentage of the active ingredient in the
obtained information required for the entire monitoring area. The effective component, EIG, of the
information refers to information that is not greater than the target’s coverage requirement in the
obtained information, and the calculation formula is as follows:
X
EIG =
min(M(s), J (s))∆s,
(2)
s∈Sarea
In the above formula, M(s) and J(s) represent the coverage requirement and the actual coverage
number of the monitoring target point in the area, respectively.
The information required for the entire monitoring area, RIF, refers to the information that satisfies
all the coverage requirements, and its calculation formula is as follows:
X
RIF =
M(s)∆s,
(3)
s∈Sarea
The calculation formula for information integrity is as follows:
P
Integrity =
min(M(s), J (s))∆s
s∈Sarea
P
,
M(s)∆s
(4)
s∈Sarea
If the coverage requirement of all target points is 2, then the above formula can be modified to
min(2, J (s))∆s
P
Integrity =
s∈Sarea
P
2∆s
,
(5)
s∈Sarea
(2) Information Validity: Information validity is the percentage of the active ingredients in the
obtained information. The calculation formula is as follows:
P
min(M(s), J (s))∆s
Validity =
s∈Sarea
P
,
J (s)∆s
(6)
s∈Sarea
If the coverage requirement of all target points is 2, then the above formula can be modified to
min(2, J (s))∆s
P
Validity =
s∈Sarea
P
J (s)∆s
,
(7)
s∈Sarea
(3) Information Redundancy: Redundancy is the percentage of the redundant ingredients in the
obtained information. The calculation formula is as follows:
P
[ J (s) − min(M(s), J (s))]∆s
Redundancy =
s∈Sarea
P
s∈Sarea
J (s)∆s
= 1 − Validity,
(8)
Sensors 2019, 19, 2752
6 of 23
Sensors
2019Deployment
2.4. Node
6 of 24
Sensors 2019
6 of 24
From the
the three
three coverage-based
coverage-based QoS
QoS metrics
metrics proposed
proposed above,
above, aa node
node deployment
deployment strategy
strategy needs
needs
From
to
improve
information
integrity
and
validity,
as
well
as
reduce
information
redundancy
as
much
From the
three coverage-based
metrics
above,
a node deployment
strategy
needs
to improve
information
integrity andQoS
validity,
as proposed
well as reduce
information
redundancy
as much
as
as
possible. information integrity and validity, as well as reduce information redundancy as much as
to improve
possible.
According to
a 1-fold
coverage-based
node
deployment,
if the
possible.
According
to existing
existingresearch,
research,when
whenconducting
conducting
a 1-fold
coverage-based
node
deployment,
if
full
coverage
condition
is
satisfied
and
the
number
of
nodes
in
the
area
is
minimized,
it
is
necessary
According
to condition
existing research,
when
conducting
a 1-fold
coverage-based
deployment,
if
the full
coverage
is satisfied
and
the number
of nodes
in the areanode
is minimized,
it is
to
full
ofcondition
the use
coverage
disk
of each
node;
that node;
is,of
thenodes
overlapping
area is
between
nodes
themake
full coverage
satisfied
and
theofnumber
theoverlapping
area
minimized,
it is
necessary
to use
make
full
ofisthe
coverage
disk
each
that
is,inthe
areathe
between
should
be as
as
possible.
This
study This
found
that
the effect
constructing
a hexagonal
necessary
to small
makebe
full
use ofasthe
coverage
disk
of each
node;
is, theofoverlapping
area
between
the
nodes
should
as
small
possible.
study
found
thatofthat
the
effect
constructing
ahoneycomb
hexagonal
grid
is
optimal.
Based
on
this,
we
propose
a
2-fold
coverage-based
node
deployment
strategy
in a
the nodes should
beoptimal.
as small as
possible.
Thiswe
study
founda that
thecoverage-based
effect of constructing
a hexagonal
honeycomb
grid is
Based
on this,
propose
2-fold
node deployment
cellular
network
for
the
case Based
where
the
coverage
requirement
of
all monitoring
nodes
in the
honeycomb
is optimal.
on
this,
we propose
a 2-fold
coverage-based
node
deployment
strategy
in a grid
cellular
network
for the
case
where
the
coverage
requirement
of alltarget
monitoring
target
region
is
2,
as
shown
in
Figure
3.
strategy
cellularisnetwork
for the
case where
the coverage requirement of all monitoring target
nodes
in in
thea region
2, as shown
in Figure
3.
nodes in the region is 2, as shown in Figure 3.
Figure 3.
schematic
diagram
of a of
2-fold
coverage-based
node deployment
strategystrategy
in a cellular
3. A A
schematic
diagram
a 2-fold
coverage-based
node deployment
in a
Figure 3.
A schematic diagram of a 2-fold coverage-based node deployment strategy in a cellular
network.
cellular
network.
network.
node deployment
deployment strategy
strategy in
in this
this paper
paper is
is based
based on
on aacellular
cellularnetwork.
network.
As shown in Figure 4, the node
steps in
areFigure
as follows:
As shown
4, the node deployment strategy in this paper is based on a cellular network.
The main
The main steps are as follows:
(1) A
A cellular
cellular network
network is
is built
built in
in the
the monitoring
monitoring area.
area.
(1)
A
cellular
network
is
built
in
the
monitoring
area.
(2) The
The sensor
sensor nodes
nodes are
are arranged
arranged at
at the
the upper
upper left,
left, upper
upper right,
right, and lower middle vertices of the
(2) grid.
The
at the upper left,
upper
right, and
lower middle vertices of the
At
point,
the
1-fold
node
deployment
is
grid.sensor
Atthis
thisnodes
point,are
thearranged
1-foldcoverage-based
coverage-based
node
deployment
is completed.
completed.
grid.
At
this
point,
the
1-fold
coverage-based
node
deployment
is
completed.
(3)
(3) The
Thesensor
sensor nodes
nodes are
are arranged
arranged at the
the vertices
vertices of the center of the grid. At
At this
this point,
point, the
the 2-covered
2-covered
(3) node
The
sensor
nodes are arranged at the vertices of the center of the grid. At this point, the 2-covered
is
node
is deployed.
deployed.
node is deployed.
Figure 4. A schematic diagram of the node deployment for a single cellular grid.
Figure 4. A schematic diagram of the node deployment for a single cellular grid.
3. The Clustering Protocol
3.
The Clustering
Clustering Protocol
Protocol
3. The
The main steps of the clustering protocol in this paper are as follows:
The
The main
main steps
steps of
of the
the clustering
clustering protocol
protocol in
in this
this paper
paper are
are as
as follows:
follows:
(1) The sensor nodes in the area are deployed according to the 2-fold coverage-based node
(1) deployment
The sensor strategy
nodes inintroduced
the area are
deployed
in Section
2. according to the 2-fold coverage-based node
deployment strategy introduced in Section 2.
Sensors 2019, 19, 2752
(1)
7 of 23
The sensor nodes in the area are deployed according to the 2-fold coverage-based node deployment
7 of 24
introduced in Section 2.
Centering on the BS, the monitoring area is divided into four small cells, and a certain number of
Centering on the BS, the monitoring area is divided into four small cells, and a certain number
CHs in each cell are selected to meet the uniformity of the CHs distribution.
of CHs in each cell are selected to meet the uniformity of the CHs distribution.
All CHs send broadcasts to the surrounding ordinary nodes, and the latter sequentially add the
All CHs send broadcasts to the surrounding ordinary nodes, and the latter sequentially add the
clusters to which the strongest signals belong according to the strength of the signals received,
clusters to which the strongest signals belong according to the strength of the signals received,
notifying the corresponding CHs to complete the cluster construction.
notifying the corresponding CHs to complete the cluster construction.
Combined with practical applications, this paper optimizes the Dijkstra algorithm to achieve the
Combined with practical applications, this paper optimizes the Dijkstra algorithm to achieve the
selection of optimal information transmission paths among the CHs.
selection of optimal information transmission paths among the CHs.
Data
Dataare
aretransmitted
transmittedand
andenergy
energyisisupdated.
updated.
Sensors
2019
strategy
(2)
(2)
(3)
(3)
(4)
(4)
(5)
(5)
To
the energy
energy consumption
consumptionof
ofthe
thenodes
nodesin
Tominimize
minimizethe
thetotal
totalenergy
energyconsumption
consumption and
and balance
balance the
inthe
thenetwork,
network,we
we
reselect
the
CHs
in
the
area
after
each
round
of
data
transmission.
Additionally,
reselect the CHs in the area after each round of data transmission. Additionally, Step
Step
is called
the node
deployment
phase
the protocol,
3 are collectively
the
1 is 1called
the node
deployment
phase
of theofprotocol,
StepsSteps
2 and23and
are collectively
calledcalled
the cluster
cluster
of protocol,
the protocol,
Step
4 iscalled
calledthe
the inter-cluster
inter-cluster routing
setup setup
phasephase
of the
andand
Step
4 is
routing mechanism
mechanism(ICRM)
(ICRM)
establishment
phase
of
the
protocol.
Steps
2,
3,
and
4
are
collectively
called
the
preparation
phase
establishment phase of the protocol. Steps 2, 3, and 4 are collectively called the preparation phaseofof
the
theprotocol.
protocol.Finally,
Finally,Step
Step5 5isiscalled
calledthe
thestabilization
stabilizationphase
phaseofofthe
theprotocol.
protocol.
Before
Beforeentering
enteringthe
thestabilization
stabilizationphase,
phase,the
theCH
CHofofeach
eachcluster
clusterneeds
needstotocreate
createa acorresponding
corresponding
time
division
multiple
access
(TDMA)
schedule
and
send
a
control
message
named
Schedule_Msg
time division multiple access (TDMA) schedule and send a control message named Schedule_Msgtoto
itsitsmember
membernodes
nodesininthe
theform
formofof(Node
(NodeNO.1,
NO.1,Time
TimeSlot
Slot1;1;Node
NodeNO.2,
NO.2,Time
TimeSlot
Slot2;......).
2;......).The
Thetime
timeslot
slot
allocation
of
the
clustering
routing
protocol
in
this
paper
is
provided
in
Figure
5.
A
flowchart
of
the
allocation of the clustering routing protocol in this paper is provided in Figure 5. A flowchart of the
clustering
6.6.
clusteringrouting
routingprotocol
protocolininthis
thispaper
paperisispresented
presentedininFigure
Figure
Node Deployment Preparation
Phase
Phase
Slot for CH1
Slot for Member
Node1 of CH1
Slot for Member
Slot for CH2
Nodec1 of CH1
Clusters Setup
Time
Inter-Cluster Routing Mechanism Setup
Figure5.5.The
Thetime
timeslot
slotallocation
allocationofofthe
theclustering
clusteringprotocol.
protocol.CH—cluster
CH—clusterhead.
head.
Figure
3.1. CH Selection
In the traditional protocols, while selecting the CHs, the residual energy and position of the node
are often neglected, resulting in an uneven final distribution of the selected CHs, among which the
CHs with lower residual energy tend to die prematurely.
In Figure 7, the green dot is the CH, and the red dot is the BS. It can be determined that in the
process of CH selection, the LEACH protocol will generate a serious phenomenon of an uneven CH
distribution, resulting in a relatively large distance between the CH and its member node, which
inevitably has an unfavorable effect on the extension of the
Sensors 2019
8 of 24
Start
Sensors 2019, 19, 2752
Sensors 2019
8 of 23
8 of 24
Node deployment based
on 2-fold coverage
Start
Division of
monitoring area
Node deployment based
on 2-fold coverage
CHs
CH Selection
selection
of
inDivision
each area
monitoring area
Cluster Setup Phase
i=1
Cluster Setup Phase
CHs
CH Selection
selection
in each area
Yes
Node i is a CH?
i=1
Data Transmission
Phase
Inter-Cluster Routing
Data Transmission
Mechanism Setup
Phase
Phase
No
Yes
Inter-Cluster
i<n? Routing
Mechanism Setup
Phase
No
Broadcast CH message
to the nodes nearby
Yes
Receive the message
broadcast
CHs
Node i isby
a CH?
Receive clustering request
from member
nodes
Broadcast
CH message
to the nodes nearby
Join the clusterNowith the
strongest
Receive
the signal
message
broadcast by CHs
Create a TDMA schedule and
Receive
clustering
send
it to the
memberrequest
nodes
from member nodes
Receive the Schedule_Msg
Join the cluster
the
message
sent bywith
the CH
strongest signal
No
Yes
i<n?
i++
i++
Create a TDMA schedule and
Receive the Schedule_Msg
send it to the member nodes
message sent by the CH
Figure 6. Flowchart of the clustering protocol, including node deployment, CHs selection, cluster
setup, and data transmission.
3.1. CH Selection
In
the traditional
protocols,
while selecting
CHs, the
residual
energyCHs
and selection,
position
the node
Figure
6.
of the
protocol,
including
node
deployment,
cluster
Figure
6. Flowchart
the clustering
clustering
protocol,the
including
node
deployment,
CHs
selection, of
cluster
are often
neglected,
resulting
in
an
uneven
final
distribution
of
the
selected
CHs,
among
which
the
setup,
setup, and
and data
data transmission.
transmission.
CHs with lower residual energy tend to die prematurely.
network
lifetime.
3.1. CH
Selection
In the traditional protocols, while selecting the CHs, the residual energy and position of the node
are often neglected, resulting in an uneven final distribution of the selected CHs, among which the
CHs with lower residual energy tend to die prematurely.
Figure 7. The three rounds of uneven CH distributions generated by the LEACH (Low-Energy Adaptive
Cluster Hierarchical) protocol.
distribution, resulting in a relatively large distance between the CH and its member node, which
inevitably has an unfavorable effect on the extension of the network lifetime.
Based on this phenomenon, this paper uses the BS as the center and divides the area into four
small cells. Then, based on the idea of CH selection in [31], combining the residual energy of a node
Sensors
19, 2752 from the node to the BS, we can construct a CH selection factor and select the
9 of top
23
with2019,
the distance
25% of the CH selection factors as the CHs in each of the small cells. The formula for the CH selection
factor is as follows:
Based on this phenomenon, this paper uses the BS as the center and divides the area into four
small cells. Then, based on the idea of CH
i selection in [31], combining the
i residual energy of a node
,
SCH E (i )，dtoBS
= α ⋅ Nor(E (i )) + β ⋅ Nor dtoBS
(9)
with the distance from the node to the BS, we can construct a CH selection factor and select the top
25% of the CH selection factors as the CHs in each of the small cells. The formula for the CH selection
In the above formula, α and β represent weight factors, where α + β = 1 ; E (i ) represents
factor is as follows:
i
i
i
represents
the
distance
from
node i to the BS. Nor(9)
()
the residual energy of node
(
(
))
SCH Ei;(i),ddtoBS
=
α
·
Nor
E
i
+
β
·
Nor
d
toBS
toBS ,
represents the normalization. The larger the value of SCH of node i, the more likely it is to be
In the above formula, α and β represent weight factors, where α + β = 1;E(i) represents the
selectedenergy
as the CH.
residual
of node i; ditoBS represents the distance from node i to the BS. Nor() represents the
Through The
the larger
mechanism,
an of
appropriate
of nodes
residual
energy
normalization.
the value
SCH of nodeamount
i, the more
likely with
it is tolarger
be selected
as the
CH. and
shorter
distances
from
the
BS
are
selected
as
CHs
in
each
small
area,
which
can
better
ensure
that
Through the mechanism, an appropriate amount of nodes with larger residual energy and shorter
there is afrom
closethe
transmission
distance
between
each
CM area,
and the
CH of
joined
cluster.
It there
avoidsisthe
distances
BS are selected
as CHs
in each
small
which
canthe
better
ensure
that
a
loss
of
too
much
residual
energy
of
some
nodes
due
to
its
large
transmission
distance.
Figure
8
shows
close transmission distance between each CM and the CH of the joined cluster. It avoids the loss of
a CH
distribution
diagram
of thenodes
protocol
It can be determined
the8distribution
too
much
residual energy
of some
dueintothis
its paper.
large transmission
distance. that
Figure
shows a CHof
the CHs is diagram
relativelyofuniform
and can
meet
the low
power
consumption
network. of the CHs
distribution
the protocol
in this
paper.
It can
be determined
that of
thethe
distribution
is relatively uniform and can meet the low power consumption of the network.
(
)
(
)
Figure
Figure8.8.AACH
CHdistribution
distributiondiagram
diagramofofthe
theprotocol
protocolininthis
thispaper.
paper.
3.2. Energy Consumption Model
3.2. Energy Consumption Model
This paper adopts the energy consumption model proposed in [32]. In the process of information
This paper adopts the energy consumption model proposed in [32]. In the process of information
transmission, the model contains two additional models based on the transmission distance: the free
transmission, the model contains two additional models based on the transmission distance: the free
space model and the multipath fading channel model.
space model and the multipath fading channel model.
ET (l, d) represents the energy consumed by wirelessly transmitting an l-bit message. The expression
ET (l , d ) represents the energy consumed by wirelessly transmitting an l-bit message. The
is as follows:
(
expression is as follows:
l · Eelec + l · ε f s · d2 , d < d0
ET (l, d) =
,
(10)
l · Eelec + l · εmp · d4 , d ≥ d0
In the above formula, Eelec represents the energy consumed per bit by the transmitter or receiver
circuit, d represents the transmission distance between the transmitter and the receiver, ε f s and εmp
represent the energy factor per bit in the free space model and the multipath fading channel model,
respectively, l represents the size of the information, and d0 represents the transmission distance
threshold, which is calculated as follows:
s
d0 =
εfs
εmp
(11)
Sensors 2019, 19, 2752
10 of 23
ER (l) represents the energy consumed by receiving an l-bit message. The expression is as follows:
s
d0 =
εfs
εmp
ER (l) = l · Eelec ,
(12)
Next we can calculate the energy consumed by the CH and the energy consumed by the CM
in a cluster according to the energy consumption model. During the operation of the WSN, the CM
is responsible for transmitting the information sensed from the environment to the corresponding
CH. In general, the transmission distance dtoCH is less than d0 . Therefore, the corresponding energy
consumption EMem is calculated as follows:
EMem = l · Eelec + l · ε f s · d2toCH ,
(13)
On the other hand, the CH is also responsible for receiving and fusing the information of the CMs
of the cluster and finally transmitting the information after completing the fusion to the BS for decision
making. The energy ECH consumed by the CH is divided into three parts: the energy ECH_R consumed
by receiving, the energy ECH_FU consumed by fusing, and the energy ECH_T consumed by transmitting.
The formulas of the above three are shown as 14, 15, and 16.
ECH_R = k · l · Eelec ,
(14)
In the above formula, k represents the number of CMs in the cluster.
ECH_FU = (k + 1) · l · EDA ,
(15)
In the above formula, EDA represents the energy consumption of each bit of data processed by the
CH. Since the CH itself can also sense the surrounding environment information, plus the information
of the k · l size received from the CMs, the total information of the (k + 1) · l size needs to be processed.
(
ECH_T = l · Eelec +
l · ε f s · d2toBS ,
l · εmp · d4toBS ,
dtoBS < d0
,
dtoBS ≥ d0
(16)
In the above formula, dtoBS represents the distance between the CH and the BS. After information
fusion, the original information of the (k + 1) · l size is compressed into the size of l.
Therefore, the energy ECH consumed by the CH is as follows:
(
ECH = ECH_R + ECH_FU + ECH_T = k · l · Eelec + (k + 1) · l · EDA + l · Eelec +
l · ε f s · d2toBS ,
l · εmp · d4toBS ,
dtoBS < d0
,
dtoBS ≥ d0
(17)
3.3. The Dijkstra Algorithm
The Dijkstra algorithm is a routing algorithm proposed by the Dutch computer scientist Dijkstra
in 1959 to solve the shortest path optimization problem between the source point and the remaining
points in the directed graph. The algorithm takes a greedy strategy, and its main steps are as follows:
(1)
(2)
Construction of an initial weight matrix
Updating the weight
Suppose there are n vertices in the region, and the set of vertices is T = {1,2,...,n}. In Step 1, we have
n2 initial weights, and a two-dimensional matrix W is used to store the n2 initial weights, where W(i,j)
represents the weight of point i to point j. It is only if point i and point j can directly convey information
to one another that W(i,j) = W(j,i). Otherwise, only point i can transmit information directly to point j,
and point j cannot directly transmit information to point i, so W(i,j) is a finite positive number, and
W(j,i) has a value of positive infinity. For i = 1,2, . . . ,n, W(i,i) = 0.
Suppose there are n vertices in the region, and the set of vertices is T={1,2,...,n}. In Step 1, we have
n2 initial weights, and a two-dimensional matrix W is used to store the n2 initial weights, where W(i,j)
represents the weight of point i to point j. It is only if point i and point j can directly convey
information to one another that W(i,j)=W(j,i). Otherwise, only point i can transmit information
Sensors 2019, 19, 2752
11 of 23
directly to point j, and point j cannot directly transmit information to point i, so W(i,j) is a finite
positive number, and W(j,i) has a value of positive infinity. For i = 1,2,…,n, W(i,i) = 0.
InInStep
Step2,2,we
wefirst
firstdetermine
determinethe
thesource
sourcepoint
points, s,then
thenestablish
establishthe
thevertex
vertexsetsetP Pofofthe
theknown
known
shortest
paths
and
thethe
vertex
setset
Q of
unknown
shortest
paths.
NowNow
P={s},Pand
Q=T-P.
Here,
shortest
paths
and
vertex
Q the
of the
unknown
shortest
paths.
= {s},
and Q
= Twe
− P.
declare
an declare
array disantoarray
storedisthe
weight
the source
s to
thesremaining
points.points.
The
Here, we
to initial
store the
initialofweight
of thepoint
source
point
to the remaining
representation
of dis is
in Figure
9.
The representation
of shown
dis is shown
in Figure
9.
dis
1
2
n
W(s,1)
W(s,1)
W(s,n)
Figure
9. 9.
The
representation
ofof
dis.
Figure
The
representation
dis.
Then,
from
itself
totos from
Then,we
wecan
canfind
findthe
thevertex
vertexx xwith
withthe
thesmallest
smallestweight
weight
from
itself
s fromQ,Q,then
thenremove
removex x
from
it it
toto
P,P,and
the
fromQ,Q,add
add
andcompare
comparewhether
whetherthe
thesum
sumofofthe
theweights
weightsfrom
frompoint
points to
s topoint
pointx xtotoone
oneofof
the
other
points
i isi smaller
than
thethe
weight
of point
s to spoint
i directly.
If yes,Ifthen
other
points
is smaller
than
weight
of point
to point
i directly.
yes, W(s,i)=W(s,x)+W(x,i),
then W(s,i) = W(s,x) +
and
we replace
corresponding
value W(s,i)
in dis.
Then,
we can
find
vertex
y with
the ysmallest
W(x,i),
and wethe
replace
the corresponding
value
W(s,i)
in dis.
Then,
wethe
can
find the
vertex
with the
weight
from
itselffrom
to s from
y from
Q, yand
add
to P,
repeating
the previous
steps until
P
smallest
weight
itselfQ,
to remove
s from Q,
remove
from
Q,itand
add
it to P, repeating
the previous
steps
contains
all the vertexes
in the graph.
until P contains
all the vertexes
in the graph.
The
Thedetail
detailprocedure
procedureofofthe
theDijkstra
Dijkstraalgorithm
algorithmisisgiven
givenby
bythe
thepseudocode
pseudocodeininAlgorithm
Algorithm1.1.
Algorithm 1. Dijkstra Algorithm
Algorithm 1. Dijkstra Algorithm
Inputs: (1) n2 weights
Inputs:(2) (1)
n2 weights
vertex
set, T
(2)
vertex
set, T
Result: an array of weights
from source point to remaining points, dis
Result:
an
array
of
weights
from source point to remaining points, dis
1: for i = 1 → n
for i = 1 → n
2: 1:for
j=1→n
2: for j = 1 → n
3: W(i, j) = weight from Point i to Point j;
3:
W(i, j) = weight from Point i to Point j;
4: end for
4: end for
5: end
for
5: end for
6: determine
the
theregion,
region,s;s;
6: determine
thesource
sourcepoint
point in
in the
7: P7:=P{s},
Q
=
T
P;
= {s}, Q = T − P;
8: dis
= [W(s,
1),1),
W(s,
2),2),…,
8: dis
= [W(s,
W(s,
. . .W(s,
, W(s,n)];
n)];
9: while
(Q(Q
~=~=
Ø)Ø)
9: while
vertex
x withthe
thesmallest
smallest weight
vertex
s from
Q; Q;
10: 10:findfind
thethe
vertex
x with
weightfrom
fromthe
the
vertex
s from
+ {x},
- {x};
11: 11:P = PP =+ P{x},
Q =QQ=-Q{x};
12: 12:for for
i = 1i =→1 n→ n
13:
if i is in Q
14:
if W(s, i) > W(s, x) + W(x, i)
15:
W(s, i) = W(s, x) + W(x, i);
16:
update the corresponding value, W(s, i) in dis;
17:
end if
18:
end if
19: end for
20: end while
3.4. Inter-Cluster Routing Mechanism
In this section, we optimize the Dijkstra algorithm: when defining the weights, we ignore the
consideration of the nonessential paths, thus reducing the computational complexity and achieving
the selection of optimal information transmission paths among the CHs.
(1) The Definition of Weights: First, based on the energy consumption of the nodes in the network
during data transmission, we propose formulas for calculating the weight between the end-to-end and
the path weight.
Sensors 2019, 19, 2752
12 of 23
The calculation formula for the weight between the end-to-end is as shown in (18):

2


 l · Eelec + l · ε f s · d(i, j) ,
W (i, j) = ET (l, d(i, j)) = 

 l · Eelec + l · εmp · d(i, j)4 ,
d(i, j) < d0
d(i, j) ≥ d0
(18)
In the above formula, d(i,j) represents the distance between node i and node j, and the weight
between the two is directly represented by the required energy consumption for transmitting information
between the two.
The calculation formula for the path weight is as shown in (19):
n
1 X
W (Mi , Mi+1 ),
W (Path(M1 , Mn+1 )) = ·
n
(19)
i=1
In the above formula, we can obtain the weight of the path W (Path(M1 , Mn+1 )), in which the
information starting point is node M1 , which passes through M2 , M3 ... in order and finally reaches
node Mn + 1 .
Because in the WSN applications information ultimately needs to be transmitted to the BS, the
path is the information transmission path from a CH to the BS in the area. Thus, (19) can be modified to:
1
W (Path(M1 , BS)) =
n
 n−1

X


· 
W (Mi , Mi+1 ) + W (Mn , BS),
(20)
i=1
(2) The Neglect Consideration of Non-Essential Paths: As described in Section 3.4, the Dijkstra
algorithm mainly selects the optimal path by focusing on the starting point and expanding the outer
layer. However, when the optimal path traverses the set Q, it often generates some unnecessary weight
calculations. As an example of a WSN, as shown in Figure 10, it is obvious that the distance d(i,x) from
node i to node x is larger than the distance d(i,BS) from node i to the BS. If node x acts as the relay node
of node i, unnecessary energy consumption will be generated. When the path of node i to the BS is
selected by the Dijkstra algorithm, the BS serves as the source point s. When the BS selects the node
with the smallest weight from itself, once x is selected, the values of W(BS,i) and W(BS,x) + W(x,i) are
inevitably compared. As mentioned above, x must not act as the relay node between node i and the BS,
Sensors 2019
13 of 24
which inevitably causes redundant calculations.
x
d(x,BS)
BS
d(i,x)
d(i,BS)
i
Figure
Figure10.
10.Schematic
Schematicdiagram
diagramofofinformation
informationtransmission
transmissionbetween
betweennodes.
nodes.BS—base
BS—basestation.
station.
Basedon
onthis,
this,we
wecould
couldignore
ignorethis
thissituation
situationin
inthe
theprocess
processofofoptimizing
optimizingDijkstra,
Dijkstra,thus
thusreducing
reducing
Based
the
computational
complexity.
In
addition,
the
Dijkstra
algorithm
selects
the
node
x
closest
tothe
the
the computational complexity. In addition, the Dijkstra algorithm selects the node x closest to
source
point
s
from
the
set
Q
as
a
multi-hops
relay
node,
and
then
whether
other
nodes
in
the
set
source point s from the set Q as a multi-hops relay node, and then whether other nodes in the set QQ
haveaabetter
bettermulti-hops
multi-hopseffect
effectthrough
throughthe
thenode
nodexxisisdetermined,
determined,thereby
therebycontinuously
continuouslyupdating
updatingthe
the
have
weights and selecting the optimal information transmission paths. From another point of view, in the
process of optimizing Dijkstra, we order all the CHs according to distance from the BS to form an
ordered set of CHs, then we traverse the CHs, and each CH is respectively judged with the previous
CHs by the path weights until all the CHs in the area find the optimal information transmission paths
to the BS.
Sensors 2019, 19, 2752
13 of 23
weights and selecting the optimal information transmission paths. From another point of view, in the
process of optimizing Dijkstra, we order all the CHs according to distance from the BS to form an
ordered set of CHs, then we traverse the CHs, and each CH is respectively judged with the previous
CHs by the path weights until all the CHs in the area find the optimal information transmission paths
to the BS.
The detailed procedure of the Optimized Dijkstra algorithm in the WSN is given by the pseudocode
in Algorithm 2.
Algorithm 2. Optimized Dijkstra Algorithm in WSN
Inputs: (1) CHs set, T
(2) the coordinate of BS, (BS.x, BS.y)
Result: (1) the set of the next hop CH of each CH, NH
(2) path set from each CH to BS, PA
(3) an array of path weights from remaining nodes to BS, dis
1: for i = 1 → n
2: for j = 1 → n
3:
W(i,j) = weight from Node i to Node j according to (14);
4: end for
5: end for
6: dis = [W(1, BS), W(2, BS), . . . , W(n, BS)]T , NH(1) = BS, PA(1) = {BS};
7: sort all CHs in T in order of the distances from the BS to form an ordered set, CHs;
8: initialize an array of the hop number for each CH, HP = [1,1, . . . ,1]n ;
9: for i = 2 → length(CHs)
10: for j = 1 → i − 1
11:
if d(i, j) < d(i, BS)
12:
if W(Path(j, BS))*HP(j) + W(i, j) < W(i, BS)*(HP(j)+1)
13:
W(i, BS) = [W(Path(j, BS))*HP(j) + W(i, j)]/(HP(j)+1);
14:
HP(i) = HP(j) + 1, NH(i) = CHs(j), PA(i) = PA(j) + {CHs(i)};
15:
update the corresponding value, W(i,BS) in dis;
16:
end if
17:
end if
18: end for
19: end for
4. Simulation Results
In this section, we evaluate the proposed protocol by simulation using Spyder (Python 3.6) on a
desktop PC (Lenovo, made in Beijing, China) with Intel(R) Core (TM) i3-4170 CPU @ 3.70 GHz, 8GB
RAM. The BS of this paper is located at coordinates (50, 45), and the area is a square of 100 m × 90 m.
The specific simulation parameters are shown in Table 1. (Note that J in this paper stands for Joule,
which is a unit of energy.)
We mainly compare the node deployment strategy proposed in this paper with the random node
deployment strategy, the uniform node deployment strategy, and the 1-fold coverage-based node
deployment strategy in a cellular network in terms of coverage, information integrity, information
validity, and information redundancy. At the same time, this paper selects three classic protocols,
including LEACH, Distribute Energy-Efficient Clustering (DEEC), and Group-Based Sensor Network
(GSEN). Then, the protocol of this paper (EECRP-HQSND-ICRM) is compared with the three protocols
and the protocol without the inter-cluster routing mechanism (ICRM) of this paper (EECRP-HQSND) to
compare the network lifetime, network coverage, information integrity, network energy consumption,
the iteration round at which the WSN starts to decline, and the network throughput to reflect the
advantages of the protocol in this paper, and especially the feasibility of the ICRM.
Sensors 2019, 19, 2752
14 of 23
Table 1. Simulation parameters.
Parameter
Value
Eelec
EDA
εfs
εmp
The size of monitoring area
The coordinate of BS
Initial number of nodes, N
Size of message, l
Initial energy
Perceptual radius, R
50 nJ/bit
5 nJ/bit/message
10 pJ/bit/m2
0.0013 pJ/bit/m4
100 m*90 m
(50, 45)
91
4000 bits
0.5 J
10 m
4.1. Comparison of the QoS of the Node Deployment Strategies
In this section, we mainly compare the node deployment strategy proposed in this paper
with the random node deployment strategy, the uniform node deployment strategy, and the 1-fold
coverage-based node deployment in a cellular network in terms of coverage, information integrity,
information validity, and information redundancy to reflect the advantages of the node deployment
strategy proposed in this paper.
4.1.1. Node Distribution of the Node Deployment Strategies
Figures 11–14 show the node distribution diagrams of the random node deployment strategy, the
uniform node deployment strategy, the 1-fold coverage-based node deployment strategy in the cellular
network,
and the node deployment strategy of this paper. The blue point represents the sensor
Sensors 2019
15 ofnode,
24
the red point represents the BS, each black circle represents the sensing range of each sensor node, and
the redSensors
square
2019is the monitoring area. It can be determined that in order to satisfy the coverage
15 of of
24 the
WSN, some of the nodes in Figures 13 and 14 are located outside of the area.
Figure 11. The node distribution diagram of the random node deployment strategy.
Figure
11. The
node
distribution
diagramofofthe
therandom
random node
strategy.
Figure
11. The
node
distribution
diagram
nodedeployment
deployment
strategy.
Figure
12.12.The
the uniform
uniformnode
nodedeployment
deploymentstrategy.
strategy.
Figure
Thenode
nodedistribution
distribution diagram
diagram of the
Figure 12. The node distribution diagram of the uniform node deployment strategy.
Figure 12. The node distribution diagram of the uniform node deployment strategy.
Sensors 2019, 19,Figure
2752 12. The node distribution diagram of the uniform node deployment strategy.
15 of 23
Figure 13. The node distribution diagram of the 1-fold coverage-based node deployment strategy in
a cellular
network.
Figure
deployment
strategy
in in
a
Figure13.
13. The
Thenode
nodedistribution
distributiondiagram
diagramofofthe
the1-fold
1-foldcoverage-based
coverage-basednode
node
deployment
strategy
cellular
network.
a cellular
network.
Figure 14. The node distribution diagram of the node deployment strategy proposed in this paper.
4.1.2. QoS of the Node Deployment Strategies
According to the calculation formulas of the coverage-based QoS metrics introduced in Section 2.3,
we compare the coverage, the information integrity, the information validity, the information
redundancy, and the number of sensor nodes of the four node deployment strategies above. The specific
results are shown in Table 2.
Table 2. Comparison of the quality of service (QoS) of the four node deployment strategies.
Strategy
Coverage
Integrity
Validity
Redundancy
Amount
Random
Uniform
1-Coverage
Cellular Grid
Our Strategy
94.233%
100%
86.689%
100%
58.854%
63.134%
41.146%
36.866%
91
110
100%
59.833%
99.889%
0.111%
46
100%
100%
83.480%
16.520%
91
As can be seen from Table 2, in terms of coverage, the random node deployment strategy only
reaches 94.233%, while the other three deterministic node deployment strategies reach 100%. In terms
of information validity and redundancy, we find that the 1-fold coverage-based node deployment
strategy in a cellular network achieves the best effect; that is, this strategy can make the obtained
information have a high validity and a low redundancy, which is the main reason why it is widely
used in the case of 1-fold coverage requirements. However, the strategy’s information integrity under
the 2-coverage requirement is only 59.833%, which obviously cannot meet the basic requirements of
the project. In terms of information integrity, the uniform node deployment strategy and the node
deployment strategy of this paper both reach 100%. However, compared with the 63.134% information
validity, 36.866% information redundancy, and the 110 nodes required by the uniform node deployment
strategy, the node deployment strategy in this paper can achieve 83.480% information validity and
16.520% information redundancy and only requires 91 sensor nodes. Obviously, the node deployment
strategy in this paper is more advantageous than the uniform node deployment strategy.
Sensors 2019, 19, 2752
16 of 23
In summary, in the four node deployment strategies, the node deployment strategy proposed in
this paper has a superior QoS and can meet the basic requirements of the project.
4.2. Comparison of the Network Lifetime
As a visual evaluation standard reflecting the performance of clustering routing protocols, the
network lifetime refers to the live nodes in the network with the iteration rounds. The steady lifetime
refers to the number of rounds until the first node death. In a steady lifetime, the nodes in the network
are full of energy and can perform large-scale data communication. This lifetime is the most vigorous
moment of the WSN’s vitality. As shown in Figure 15, the steady lifetimes of the LEACH, DEEC, and
GSEN are 1061, 1135, and 1163 rounds, respectively, and the steady lifetime of the EECRP-HQSND is
1331, which further exceeds LEACH, DEEC, and GSEN by 25.45%, 17.27%, and 14.45%, respectively. In
addition, through a reasonable ICRM, the EECRP-HQSND-ICRM extends the steady lifetime by 33
rounds compared to the EECRP-HQSND. With the iteration rounds, there are few nodes in LEACH
Sensors 2019
17 of 24
and DEEC around the 1300th round. At this time, there is no node death in EECRP-HQSND or
EECRP-HQSND-ICRM. After around the 1356th round, the number of nodes in EECRP-HQSND is
HQSND is almost 0, but EECRP-HQSND-ICRM still maintains 91. So, we find that ICRM can reduce
almost 0, but EECRP-HQSND-ICRM still maintains 91. So, we find that ICRM can reduce the network
the network energy consumption well. In addition, in terms of the downward trend of the curves of
energy consumption well. In addition, in terms of the downward trend of the curves of the live nodes,
the live nodes, it is obvious that compared with LEACH, DEEC, and GSEN, after the death of the first
it is obvious that compared with LEACH, DEEC, and GSEN, after the death of the first node, the slopes
node, the slopes of EECRP-HQSND and EECRP-HQSND-ICRM change the fastest, and the slopes
of EECRP-HQSND and EECRP-HQSND-ICRM change the fastest, and the slopes can almost reach
can almost reach negative infinity. Thus we find that compared with the traditional CH selection
negative infinity. Thus we find that compared with the traditional CH selection mechanisms, CH
mechanisms, CH selection through partition based on the uniformity of the CH distribution in this
selection through partition based on the uniformity of the CH distribution in this paper can balance
paper can balance the network energy consumption well.
the network energy consumption well.
Figure 15. The number of live nodes varies with the iteration round. DEEC— Distribute Energy-Efficient
Figure 15. The number of live nodes varies with the iteration round. DEEC— Distribute EnergyClustering; EECRP-HQSND—energy-efficient clustering routing protocol based on a high-QoS node
Efficient Clustering; EECRP-HQSND—energy-efficient clustering routing protocol based on a highdeployment; EECRP-HQSND-ICRM—energy-efficient clustering routing protocol based on a high-QoS
QoS node deployment; EECRP-HQSND-ICRM—energy-efficient clustering routing protocol based
node deployment with an inter-cluster routing mechanism; GSEN— and Group-Based Sensor Network;
on a high-QoS node deployment with an inter-cluster routing mechanism; GSEN— and GroupLEACH—Low-Energy Adaptive Cluster Hierarchical.
Based Sensor Network; LEACH—Low-Energy Adaptive Cluster Hierarchical.
4.3. Comparison of the Network Coverage
4.3. Comparison of the Network Coverage
The network coverage reflects the proportion of monitored targets covered in the area out of
Thenumber,
networkreflecting
coveragethe
reflects
the proportion
of monitored
covered
the area coverage
out of the
the total
network’s
monitoring
capability. targets
A decline
in theinnetwork
total
number,
reflecting
the
network's
monitoring
capability.
A
decline
in
the
network
means that the monitoring capability of the network begins to decline. As shown in Figure 16,coverage
among
means
the monitoring
capability
of the network
beginscapability
to decline.under
As shown
in Figure
among
the
five that
protocols,
the decline
of the network
monitoring
LEACH
is the16,earliest,
the five protocols,
the GSEN,
decline while
of theEECRP-HQSND
network monitoring
capability
undera LEACH
is monitoring
the earliest,
followed
by DEEC and
begins
to experience
decline in
followed by DEEC and GSEN, while EECRP-HQSND begins to experience a decline in monitoring
capacity in the 1333rd round. Through an effective ICRM, the EECRP-HQSND-ICRM delays the
monitoring capability decline to the 1367th round, thus ensuring that the network can perform more
rounds of monitoring during this period to achieve higher monitoring requirements. Meantime, with
a reasonable CH selection mechanism through partition based on the uniformity of the CH
distribution, compared with LEACH, DEEC, and GSEN, the round when the network coverage of
Sensors 2019, 19, 2752
17 of 23
capacity in the 1333rd round. Through an effective ICRM, the EECRP-HQSND-ICRM delays the
monitoring capability decline to the 1367th round, thus ensuring that the network can perform more
rounds of monitoring during this period to achieve higher monitoring requirements. Meantime, with a
Sensors 2019 CH selection mechanism through partition based on the uniformity of the CH distribution,
18 of 24
reasonable
compared
with
LEACH,
DEEC,
and
GSEN,
the
round
when
the
network
coverage
of
EECRP-HQSND
Sensors 2019
18 of 24
and EECRP-HQSND-ICRM begins to decline can be greatly delayed.
Figure 16. The network coverage varies with the iteration round.
Figure
Figure16.
16.The
Thenetwork
networkcoverage
coveragevaries
varieswith
withthe
theiteration
iterationround.
round.
4.4. Comparison of the Information Integrity
4.4. Comparison of the Information Integrity
4.4. Comparison
of integrity
the Information
Information
reflectsIntegrity
the proportion of information required to meet the monitored area
Information
integrity
reflects
the
proportion
of information
required
to meet the
monitored
as
as a percentage of the total required
information.
The decline
in information
integrity,
to aarea
certain
Information
integrity
reflects
the proportion
of information
required
to meet to
thea monitored
area
a extent,
percentage
of
the
total
required
information.
The
decline
in
information
integrity,
certain
extent,
means that the obtained information has become less desirable. As shown in Figure 17, similar
as a percentage
of the total
required
information.
declineAs
in shown
information
integrity,
to a to
certain
means
that
the obtained
information
has
lessThe
desirable.
Figure
similar
the
to the network
lifetime and
coverage,
thebecome
information
integrity of
LEACH in
is the
first17,
to decline
among
extent,
means
that
the
obtained
information
has
become
less
desirable.
As
shown
in
Figure
17,
similar
network
andfollowed
coverage,bythe
of LEACH
the first to declineonly
among
the five
the fivelifetime
protocols,
theinformation
DEEC and integrity
the GSEN,
while theisEECRP-HQSND
experienced
to the network
lifetime
and
coverage,
theGSEN,
information
integrity
of LEACH only
is theexperienced
first to decline
among
protocols,
followed
by
the
DEEC
and
the
while
the
EECRP-HQSND
a
a decline in information integrity in the 1332nd round. Finally, the ICRM in this paper candecline
greatly
five protocols,
followed
the DEEC
and the GSEN,
while
only
experienced
inthe
information
integrity
inload,
the by
1332nd
round.
ICRM
in the
thisEECRP-HQSND
paper
can greatly
reduce
CHs’
reduce
the CHs’
energy
especially
for Finally,
the CHsthe
that
are remotely
located,
resulting
in thethe
EECRPa
decline
in
information
integrity
in
the
1332nd
round.
Finally,
the
ICRM
in
this
paper
can greatly
energy
load,
especially
for
the
CHs
that
are
remotely
located,
resulting
in
the
EECRP-HQSND-ICRM
HQSND-ICRM postponing the decline of information integrity to the 1365th round, thus ensuring
reduce the the
CHs’
energyofload,
especially
for the CHs
that
are remotely
located, resulting in the EECRPpostponing
decline
information
integrity
toduring
the
1365th
that the network
can obtain
more complete
data
this round,
period. thus ensuring that the network
HQSND-ICRM
postponing
the
decline
information integrity to the 1365th round, thus ensuring
can
obtain more complete
data
during
thisofperiod.
that the network can obtain more complete data during this period.
Figure
Figure17.
17.The
Theinformation
informationintegrity
integrityvaries
varieswith
withthe
theiteration
iterationround.
round.
Figure 17. The information integrity varies with the iteration round.
4.5. Comparison of the Round at Which the WSN Begins to Decline
4.5. Comparison
of thewe
Round
Which the
the WSN
Begins
to Decline
In this section,
will at
analyze
network's
decline
from three aspects: the number of rounds
until the death of the first node, the number of rounds leading to the decline in network coverage,
In this section, we will analyze the network's decline from three aspects: the number of rounds
Sensors 2019, 19, 2752
18 of 23
4.5. Comparison of the Round at Which the WSN Begins to Decline
Sensors
19 of 24
In 2019
this section, we will analyze the network’s decline from three aspects: the number of rounds
until the death of the first node, the number of rounds leading to the decline in network coverage, and
of number
Sectionsof4.2,
4.3, and
4.4, to
our
in Table
3 andCombined
Figure 18.with
It can
seen that
the
rounds
leading
theresults
declineare
in shown
information
integrity.
the be
contents
of
compared
with
the
traditional
protocols,
including
LEACH,
DEEC,
and
GSEN,
the
CH
selection
Sections 4.2–4.4, our results are shown in Table 3 and Figure 18. It can be seen that compared with
the
mechanism
of partition
based LEACH,
on the uniformity
ofGSEN,
the CHthe
distribution
proposed
in this
can
traditional
protocols,
including
DEEC, and
CH selection
mechanism
of paper
partition
balance
the
network
energy
consumption
well,
thus
greatly
delaying
the
number
rounds
until
the
based on the uniformity of the CH distribution proposed in this paper can balance the network energy
death
of
the
first
node,
the
number
of
rounds
leading
to
the
decline
in
network
coverage,
and
the
consumption well, thus greatly delaying the number rounds until the death of the first node, the
numberofofrounds
rounds
leading
decline
in information
integrity.
by constructing
the optimal
number
leading
to to
thethe
decline
in network
coverage,
and theThen,
number
of rounds leading
to the
path
from
each
CH
to
the
BS,
the
ICRM
proposed
in
this
paper
can
further
reduce
the
network
energy
decline in information integrity. Then, by constructing the optimal path from each CH to the BS, the
consumption,
optimizing
above
threethe
aspects.
ICRM
proposedthus
in this
paper canthe
further
reduce
network energy consumption, thus optimizing the
above three aspects.
Table 3. Comparison of the number of rounds leading to decline in the network for each protocol.
Table 3. Comparison of the number of rounds leading to decline in the network for each protocol.
Protocol
LEACH
Protocol
DEEC
LEACH
GSEN
DEEC
GSEN
EECRP-HQSND
EECRP-HQSND
EECRP-HQSND-ICRM
EECRP-HQSND-ICRM
First Node Death
1061Death
First Node
1135
1061
1163
1135
1163
1331
1331
1364
1364
Coverage Degradation Integrity Degradation
1093
1072
Coverage Degradation
Integrity Degradation
10931175
1072 1149
1227
1175
1149 1172
12271333
1172 1332
13331367
1332 1365
1367
1365
Figure
Figure18.
18.Comparison
Comparisonofofthe
thenumber
numberofofrounds
roundsleading
leadingtotodecline
declineininthe
thenetwork
networkfor
foreach
eachprotocol.
protocol.
4.6. Comparison of the Network Energy Consumption
4.6. Comparison of the Network Energy Consumption
In the analysis of network energy consumption, this paper mainly considers two aspects: the
In the analysis of network energy consumption, this paper mainly considers two aspects: the
network energy consumption size and the network energy consumption uniformity. The index of the
network energy consumption size and the network energy consumption uniformity. The index of the
network energy consumption uniformity is mainly the range and variance of the residual energy of the
network energy consumption uniformity is mainly the range and variance of the residual energy of
nodes in the network.
the nodes in the network.
4.6.1. Comparison of Network Energy Consumption
4.6.1. Comparison of Network Energy Consumption
In this paper, the BS is centered, and the area is divided into 4 cells. Then, we select the top 25%
In this
paper,
BS is cell
centered,
and the the
areaestablishment
is divided into
cells.
Then,
we select
top 25%
of nodes
as the
CHsthe
in each
and perform
of4an
ICRM
through
the the
optimized
of
nodes
as
the
CHs
in
each
cell
and
perform
the
establishment
of
an
ICRM
through
the
optimized
Dijkstra algorithm to reduce network energy consumption and balance the energy consumption of
Dijkstra algorithm to reduce network energy consumption and balance the energy consumption of
the network nodes. As shown in Figure 19, compared with LEACH and DEEC, the GSEN has an
advantage in reducing energy consumption. Compared to these three, the EECRP-HQSND
effectively reduces the energy consumption of the network, which proves the effectiveness of the CH
Sensors 2019, 19, 2752
19 of 23
the network nodes. As shown in Figure 19, compared with LEACH and DEEC, the GSEN has an
Sensors 2019
20 of 24
advantage
in reducing energy consumption. Compared to these three, the EECRP-HQSND effectively
Sensors 2019
20 of 24
reduces
the energy consumption of the network, which proves the effectiveness of the CH selection
selection mechanism in this paper. In addition, EECRP-HQSND-ICRM further reduces the network
mechanism in this paper. In addition, EECRP-HQSND-ICRM further reduces the network energy
energy
consumption
to EECRP-HQSND,
reflecting the feasibility
of the
ICRM.the network
selection
mechanism compared
in this paper.
In addition, EECRP-HQSND-ICRM
further
reduces
consumption compared to EECRP-HQSND, reflecting the feasibility of the ICRM.
energy consumption compared to EECRP-HQSND, reflecting the feasibility of the ICRM.
Figure
Figure19.
19.The
Thenetwork
networkresidual
residualenergy
energyvaries
varieswith
withthe
theiteration
iterationround.
round.
Figure
19. The
network
residual energy
varies with the iteration round.
4.6.2. Comparison of
Network
Energy
Consumption
Uniformity
4.6.2. Comparison of Network Energy Consumption Uniformity
It Comparison
can be seen of
from
Figures
20 and
21 that theUniformity
protocol can effectively balance the energy
4.6.2.
Network
Energy
Consumption
consumption of nodes in the network compared to LEACH, DEEC, and GSEN.
Figure 20. The range of the residual energy of the nodes in the network.
Figure
Figure20.
20.The
Therange
rangeofofthe
theresidual
residualenergy
energyofofthe
thenodes
nodesininthe
thenetwork.
network.
Sensors
2019
Sensors
2019,
19, 2752
2021ofof
23 24
Figure 21. The variance of the residual energy of the nodes in the network.
Figure 21. The variance of the residual energy of the nodes in the network.
In the network iteration, the maximum energy range of LEACH reaches approximately 0.13 J, that
It can be seen from Figures 20 and 21 that the protocol can effectively balance the energy
of DEEC and GSEN both reach approximately 0.10 J, that of EECRP-HQSND reaches approximately
consumption of nodes in the network compared to LEACH, DEEC, and GSEN.
0.028 J, and that of EECRP-HQSND-ICRM reaches a maximum of 0.01564 J.
In the network iteration, the maximum energy range of LEACH reaches approximately 0.13 J,
As for the energy variance, EECRP-HQSND EECRP-HQSND-ICRM achieve variances of
that of DEEC and GSEN both reach approximately 0.10 J, that of EECRP-HQSND reaches
0–0.000036. The maximum variance of LEACH reaches 0.0010, that of DEEC reaches 0.0003, and that of
approximately 0.028 J, and that of EECRP-HQSND-ICRM reaches a maximum of 0.01564 J.
GSEN reaches 0.0004.
As for the energy variance, EECRP-HQSND EECRP-HQSND-ICRM achieve variances of 0–
Based on the comprehensive energy range and variance, it can be determined that the protocol in
0.000036. The maximum variance of LEACH reaches 0.0010, that of DEEC reaches 0.0003, and that of
this paper can achieve a good balance of energy consumption among the nodes in the network.
GSEN reaches 0.0004.
Based onofthe
energy range and variance, it can be determined that the protocol
4.7. Comparison
thecomprehensive
Network Throughput
in this paper can achieve a good balance of energy consumption among the nodes in the network.
The network throughput is an important indicator that fundamentally reflects the performance of
a protocol.
The network
throughput
refers to the number of packets in the network that are ultimately
4.7. Comparison
of the Network
Throughput
sent to the BS. The CMs transmit the sensed information to the CH in the form of packets, and the
The network throughput is an important indicator that fundamentally reflects the performance
CH fuses this information with that sensed by itself and finally sends the information to the BS in the
of a protocol. The network throughput refers to the number of packets in the network that are
form of packets. As the nodes in the network die continuously, the number of CHs in the network
ultimately sent to the BS. The CMs transmit the sensed information to the CH in the form of packets,
decreases, and the success rate of transmitting packets to the BS also decreases. Therefore, the number
and the CH fuses this information with that sensed by itself and finally sends the information to the
of data packets transmitted in each round will gradually decrease. Until the last node dies, the network
BS in the form of packets. As the nodes in the network die continuously, the number of CHs in the
throughput reaches a maximum value, which is just the final network throughput.
network decreases, and the success rate of transmitting packets to the BS also decreases. Therefore,
As seen from Figure 22, the final network throughput of LEACH is 11005, that of DEEC is 13901,
the number of data packets transmitted in each round will gradually decrease. Until the last node
and that of GSEN is 16239. Compared with these three, the EECRP-HQSND increases the final
dies, the network throughput reaches a maximum value, which is just the final network throughput.
network throughput by 167.53%, 111.80%, and 81.30%, respectively. With an effective ICRM, the
As seen from Figure 22, the final network throughput of LEACH is 11005, that of DEEC is 13901,
EECRP-HQSND-ICRM increases the final network throughput by 856 compared to EECRP-HQSND.
and that of GSEN is 16239. Compared with these three, the EECRP-HQSND increases the final
network throughput by 167.53%, 111.80%, and 81.30%, respectively. With an effective ICRM, the
EECRP-HQSND-ICRM increases the final network throughput by 856 compared to EECRP-HQSND.
Sensors
Sensors2019,
201919, 2752
21
22 of
of2324
Figure
Figure22.
22.The
Thecomparison
comparisonof
ofthe
thefinal
finalthroughput
throughputininthe
thefive
fiveprotocols.
protocols.
5. Conclusions
5. Conclusions
Aimed at the different coverage requirements of regional monitoring target points in real-world
Aimed at
thepaper
different
coverage
requirements
of regional
monitoring
points
in real-world
applications,
this
proposes
a node
deployment
strategy
based on target
twofold
coverage
for all
applications,
this
paper
proposes
a
node
deployment
strategy
based
on
twofold
coverage
for all
coverage requirements of two in the region. Then, from the network coverage, this paper proposes
a
coverage
requirements
of
two
in
the
region.
Then,
from
the
network
coverage,
this
paper
proposes
series of definition formulas for information integrity, validity, and redundancy. In the CH selection,a
series
of definition
formulas
information
integrity,
validity, and
In the CH
selection,
this
paper
fully considers
the for
uniformity
of the
CH distribution
andredundancy.
proposes a strategy
of dividing
thismonitoring
paper fullyarea
considers
thesmall
uniformity
of the
distribution
a strategy
of dividing
the
into four
cells with
theCH
BS as
the center and
and proposes
then selecting
the CHs
in their
the
monitoring
area
into
four
small
cells
with
the
BS
as
the
center
and
then
selecting
the
CHs
in their
respective cells. Finally, combined with the actual requirements of the WSN, this paper optimizes
the
respective
cells.
Finally,
combined
with
the
actual
requirements
of
the
WSN,
this
paper
optimizes
the
Dijkstra algorithm, including (1) nonessential paths neglecting considerations and (2) a simultaneous
Dijkstra algorithm,
including
(1) nonessential
neglecting
(2) a simultaneous
introduction
of end-to-end
weights
and path paths
weights.
Then, considerations
the optimized and
Dijkstra
algorithm is
introduction
of
end-to-end
weights
and
path
weights.
Then,
the
optimized
Dijkstra
algorithm is
applied to the establishment of the ICRM.
applied
the establishment
of thethat,
ICRM.
The to
simulation
results show
compared with the random node distribution strategy, the
The
simulation
results
show
that,
with
the random node
uniform node distribution strategy, and compared
the onefold
coverage-based
node distribution
deploymentstrategy,
strategy the
in
uniform
node
distribution
strategy,
and
the
onefold
coverage-based
node
deployment
strategy
in a
a cellular network, the node deployment strategy in this paper achieves relatively high coverage,
cellular network,
the
node
deployment
in low
this information
paper achieves
relatively
coverage,
information
integrity,
and
validity,
as well asstrategy
a relatively
redundancy.
Athigh
the same
time,
information
integrity,
and
validity,
as
well
as
a
relatively
low
information
redundancy.
At
the same
compared with LEACH, DEEC, and GSEN, EECRP-HQSND-ICRM can reduce and balance network
time, compared
with
LEACH,
DEEC,
and GSEN,
EECRP-HQSND-ICRM
can reduce
and balance
energy
consumption,
thus
extending
the network
lifetime
and improving the network
throughput.
It is
network
thus extending
network
andofimproving
the network
noted
that energy
the CH consumption,
selection mechanism
of partitionthe
based
on thelifetime
uniformity
the CH distribution
in
throughput.
It is noted
that thefor
CH
selection
mechanism
partition
basedand
on uniform
the uniformity
of the
this
paper is mainly
responsible
selecting
CHs
with largeofresidual
energy
distribution,
CH distribution
this papernetwork
is mainly
responsible
for selecting
CHs with
residual
energy
and
thereby
effectivelyinbalancing
energy
consumption
and delaying
thelarge
number
of rounds
until
uniform
distribution,
thereby
effectively
balancing
network
energy
consumption
and
delaying
the
the network performance begins to decline. The ICRM is mainly responsible for finding the optimal
number
of
rounds
until
the
network
performance
begins
to
decline.
The
ICRM
is
mainly
responsible
transmission path from each CH to the BS in the network and preventing some CHs from dying early
for finding
thelong
optimal
transmission
path from
CHthereby
to the BS
in the
networknetwork
and preventing
because
of the
transmission
distances
from each
the BS,
further
reducing
energy
some
CHs
from
dying
early
because
of
the
long
transmission
distances
from
the
BS,
thereby further
consumption and realizing more rounds of network data iterations, which has great significance
for
reducing
network
energy
consumption
and
realizing
more
rounds
of
network
data
iterations,
which
engineering. However, there are some shortcomings in our protocol:
has First,
great significance
engineering.
some shortcomings
our protocol:
we used thefor
0-1
perception However,
model forthere
nodeare
deployment
and QoSinformula
derivation.
First,
we
used
the
0-1
perception
model
for
node
deployment
and
QoS
formula
In
In actual situations, the perception of nodes is very complicated, and it is much less idealderivation.
than the 0-1
actual situations,
perception
of nodes
very complicated,
and it is much
less
ideal
than
the 0-1
perception
model. the
Therefore,
we may
laterisconsider
a node deployment
scheme
and
QoS
formulas
perception
model.
Therefore,
we
may
later
consider
a
node
deployment
scheme
and
QoS
formulas
based on a more complex perceptual model.
based
on a more
complex
perceptualonly
model.
Finally,
the protocol
is applicable
to 2D scenarios. Typical 3D scenarios, such as underground
Finally,
the
protocol
is
applicable
only to 2D scenarios. Typical 3D scenarios, such as
coal mines, underground pipe corridors, and indoor homes require the WSN to be arranged in three
underground coal mines, underground pipe corridors, and indoor homes require the WSN to be
Sensors 2019, 19, 2752
22 of 23
dimensions. Therefore, in the future, we will consider proposing a clustering routing protocol suitable
for 3D scenarios based on this protocol.
Author Contributions: Z.Z. and K.X. conducted the research and performed the simulations. Z.Z. led the research.
Y.L., G.H. and L.H. supervised the work and suggested revisions. All authors participated in the writing of
the article.
Funding: This research was supported by National Natural Science Foundation of China (Grant No. U1709212),
the Zhejiang Province Public Welfare Technology Application Research Project (Grant No. GF18F030010), and the
Smart City Collaborative Innovation Center of Zhejiang Province. This work was also supported by the Zhejiang
Provincial Key Lab of Equipment Electronics.
Conflicts of Interest: The authors declare no conflict of interest.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
Jan, B.; Farman, H.; Javed, H.; Montrucchio, B.; Khan, M.; Ali, S. Energy Efficient Hierarchical Clustering
Approaches in Wireless Sensor Networks: A Survey. Wirel. Commun. Mob. Comput. 2017, 2017, 1–4.
[CrossRef]
Amodu, O.A.; Raja Mahmood, R.A. Impact of the energy-based and location-based LEACH secondary
cluster aggregation on WSN lifetime. Wirel. Netw. 2018, 24, 1379–1402. [CrossRef]
Hosen, A.; Cho, G. An Energy Centric Cluster-Based Routing Protocol for Wireless Sensor Networks. Sensors
2018, 18, 1520. [CrossRef] [PubMed]
Anzola, J.; Pascual, J.; Tarazona, G.; González, R. A Clustering WSN Routing Protocol Based on k-d Tree
Algorithm. Sensors 2018, 18, 2899. [CrossRef] [PubMed]
Saleh, A.I.; Abo-Al-Ez, K.M.; Abdullah, A.A. A Multi-Aware Query Driven (MAQD) routing protocol for
mobile wireless sensor networks based on neuro-fuzzy inference. J. Net. Comput. Appl. 2017, 88, 72–98.
[CrossRef]
Yadav, R.N.; Misra, R.; Saini, D. Energy aware cluster based routing protocol over distributed cognitive radio
sensor network. Comput. Commun. 2018, 129, 54–66. [CrossRef]
Xia, X.; Chen, Z.; Liu, H.; Wang, H.; Zeng, F. A Routing Protocol for Multisink Wireless Sensor Networks in
Underground Coalmine Tunnels. Sensors 2016, 16, 2032. [CrossRef]
Rani, S.; Ahmed, S.H.; Malhotra, J.; Talwar, R. Energy efficient chain based routing protocol for underwater
wireless sensor networks. J. Net. Comput. Appl. 2017, 92, 42–50. [CrossRef]
Bozorgi, S.M.; Shokouhi Rostami, A.; Hosseinabadi, A.A.R.; Balas, V.E. A new clustering protocol for energy
harvesting-wireless sensor networks. Comput.Electr. Eng. 2017, 64, 233–247. [CrossRef]
Sharma, S.; Bansal, R.K.; Bansal, S. Heterogeneity-aware Energy-efficient Clustering (HEC) Technique for
WSNs. KSII Trans. Int. Inf. Syst. 2017, 11, 1866–1888.
Xu, Y.; Ding, O.; Qu, R.; Li, K. Hybrid multi-objective evolutionary algorithms based on decomposition for
wireless sensor network coverage optimization. Appl. Soft Comput. 2018, 68, 268–282. [CrossRef]
Zhao, Z.; Xu, K.; Hui, G.; Hu, L. An Energy-Efficient Clustering Routing Protocol for Wireless Sensor
Networks Based on AGNES with Balanced Energy Consumption Optimization. Sensors 2018, 18, 3938.
[CrossRef] [PubMed]
Chidean, M.I.; Morgado, E.; Del Arco, E.; Ramiro-Bargueno, J.; Caamano, A.J. Scalable Data-Coupled
Clustering for Large Scale WSN. IEEE Trans. Wirel. Commun. 2015, 14, 4681–4694. [CrossRef]
Wang, Q.; Guo, S.; Hu, J.; Yang, Y. Spectral partitioning and fuzzy C-means based clustering algorithm for
big data wireless sensor networks. EURASIP J. Wirel. Commun. Net. 2018, 54, 1–11. [CrossRef]
Palma, J.; Duran-Faundez, C.; de P Carvalho, L.; Morais, C.F.; Oliveira, R.; Rubio, E.; Herman, K. Protocol
for Energy-Efficiency in Networked Control Systems Based on WSN. Sensors 2018, 18, 2590. [CrossRef]
[PubMed]
Gharaei, N.; Abu Bakar, K.; Mohd Hashim, S.; Hosseingholi Pourasl, A.; Siraj, M.; Darwish, T. An
Energy-Efficient Mobile Sink-Based Unequal Clustering Mechanism for WSNs. Sensors 2017, 17, 1858.
[CrossRef] [PubMed]
Fan, F.; Ji, Q.; Wu, G.; Wang, M.; Ye, X.; Mei, Q. Dynamic Barrier Coverage in a Wireless Sensor Network for
Smart Grid. Sensors 2019, 19, 41. [CrossRef]
Sensors 2019, 19, 2752
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
23 of 23
Almalkawi, I.T.; Guerrero Zapata, M.; Al-Karaki, J.N. A Secure Cluster-Based Multipath Routing Protocol for
WMSNs. Sensors 2011, 11, 4401–4424. [CrossRef]
Guermazi, A.; Belghith, A.; Abid, M.; Gannouni, S. KMMR: An Efficient and scalable Key Management
Protocol to Secure Multi-Hop Communications in large scale Wireless Sensor Networks. KSII Trans. Int.
Inf. Syst. 2017, 11, 901–923.
Md Zin, S.; Badrul Anuar, N.; Laiha Mat Kiah, M.; Khan Pathan, A. Routing protocol design for secure WSN:
Review and open research issues. J. Net. Comput. Appl. 2014, 41, 517–530. [CrossRef]
Xu, L.; O’Hare, G.M.P.; Collier, R. A Smart and Balanced Energy-Efficient Multihop Clustering Algorithm
(Smart-BEEM) for MIMO IoT Systems in Future Networks. Sensors 2017, 17, 1574. [CrossRef] [PubMed]
Musznicki, B.; Tomczak, M.; Zwierzykowski, P. Dijkstra-based Localized Multicast Routing in Wireless Sensor
Netw. In Proceedings of the 2012 8th International Symposium on Communication Systems, Networks &
Digital Signal processing, Poznan, Poland, 18–20 July 2012; pp. 1–6.
Razzaq, M.; Kwon, G.; Shin, S. Energy Efficient Dijkstra-Based Weighted Sum Minimization Routing Protocol
for WSN. In Proceedings of the 2018 Third International Conference on Fog and Mobile Edge Computing
(FMEC), Barcelona, Spain, 23–26 April 2018; pp. 246–251.
Thomas, S.; Gayathri, I.K.; Raj, A. In Joint design of Dijkstra’s shortest path routing and sleep-wake scheduling
in wireless sensor networks. In Proceedings of the 2017 International Conference on Energy, Communication,
Data Analytics and Soft Computing (ICECDS), Chennai, India, 1–2 August 2017; pp. 981–986.
Xie, W.; Zhang, Q.; Sun, Z.; Zhang, F. A Clustering Routing Protocol for WSN Based on Type-2 Fuzzy Logic
and Ant Colony Optimization. Wirel. Pers. Commun. 2015, 84, 1165–1196. [CrossRef]
Kandris, D.; Tselikis, G.; Anastasiadis, E.; Panaousis, E.; Dagiuklas, T. COALA: A Protocol for the Avoidance
and Alleviation of Congestion in Wireless Sensor Networks. Sensors 2017, 17, 2502. [CrossRef] [PubMed]
Mohamed, R.E.; Ghanem, W.R.; Khalil, A.T.; Elhoseny, M.; Sajjad, M.; Mohamed, M.A. Energy efficient
collaborative proactive routing protocol for Wireless Sensor Network. Comput. Net. 2018, 142, 154–167.
[CrossRef]
Leyva-Mayorga, I.; Rivero-Angeles, M.E.; Carreto-Arellano, C.; Pla, V. QoS Analysis for a Nonpreemptive
Continuous Monitoring and Event-Driven WSN Protocol in Mobile Environments. Int. J. Distrib. Sens. Net.
2015, 11, 471307–471322. [CrossRef]
Muhammad, Z.; Roy, A.; Ahn, C.W.; Sachan, R.; Saxena, N. A Novel Random Scheduling Algorithm based
on Subregions Coverage for SET K-Cover Problem in Wireless Sensor Network. KSII Trans. Int. Inf. Syst.
2018, 12, 2658–2679.
Boukerche, A.; Sun, P. Connectivity and coverage based protocols for wireless sensor networks. Ad Hoc Netw.
2018, 80, 54–69. [CrossRef]
Bhatti, D.M.S.; Saeed, N.; Nam, H. Fuzzy C-Means Clustering and Energy Efficient Cluster Head Selection
for Cooperative Sensor Network. Sensors 2016, 16, 1459. [CrossRef]
Heinzelman, W.B.; Chandrakasan, A.P.; Balakrishnan, H. An application-specific protocol architecture for
wireless microsensor networks. IEEE Trans. Wirel. Commun. 2002, 1, 660–670. [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/).