SPECIAL SECTION ON INTELLIGENT SYSTEMS FOR THE INTERNET OF THINGS

Received August 11, 2017, accepted September 4, 2017, date of publication September 7, 2017,
date of current version September 27, 2017.
Digital Object Identifier 10.1109/ACCESS.2017.2749606

An Efficient Centroid-Based Routing Protocol for

Energy Management in WSN-Assisted IoT
JIAN SHEN1,2 , (Member, IEEE), ANXI WANG3 , CHEN WANG3 ,
PATRICK C. K. HUNG4 , (Member, IEEE), AND CHIN-FENG LAI5 , (Senior Member, IEEE)
1 Jiangsu Engineering Center of Network Monitoring, School of Computer and Software, Nanjing University of Information Science and Technology,
Nanjing 210044, China
2 State Key Laboratory of Information Security, Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China
3 School of Computer and Software, Nanjing University of Information Science and Technology, Nanjing 210044, China
4 Faculty of Business and IT, University of Ontario Institute of Technology, Oshawa, ON L1H 7K4, Canada
5 Department of Engineering Science, National Cheng Kung University, Chia-Yi 62102, Taiwan

Corresponding author: Chin-Feng Lai (cinfon@ieee.com)

This work was supported in part by the National Science Foundation of China under Grant 61672295, Grant 61300237, and
Grant U1405254, in part by the State Key Laboratory of Information Security under Grant 2017-MS-10, in part by the 2015 Project
of six personnel in Jiangsu Province under Grant R2015L06, in part by CICAEET, and in part by PAPD.

ABSTRACT Wireless sensor networks (WSNs) distribute hundreds to thousands of inexpensive micro-
sensor nodes in their regions, and these nodes are important parts of Internet of Things (IoT).
In WSN-assisted IoT, the nodes are resource constrained in many ways, such as storage resources, computing
resources, energy resources, and so on. Robust routing protocols are required to maintain a long network
lifetime and achieve higher energy utilization. In this paper, we propose a new energy-efficient centroid-
based routing protocol (EECRP) for WSN-assisted IoT to improve the performance of the network. The
proposed EECRP includes three key parts: a new distributed cluster formation technique that enables the
self-organization of local nodes, a new series of algorithms for adapting clusters and rotating the cluster
head based on the centroid position to evenly distribute the energy load among all sensor nodes, and
a new mechanism to reduce the energy consumption for long-distance communications. In particular,
the residual energy of nodes is considered in EECRP for calculating the centroid0 s position. Our sim-
ulation results indicate that EECRP performs better than LEACH, LEACH-C, and GEEC. In addition,
EECRP is suitable for networks that require a long lifetime and whose base station (BS) is located in the
network.

INDEX TERMS Internet of Things, wireless sensor networks, energy management, cluster.

I. INTRODUCTION data in IoT [10]. With the help of cloud computing,

Internet of Things (IoT) is the inter-networking of physical
devices, vehicles, buildings, and other items that contain
embedded electronics, software, sensors, actuators and so on.
IoT enables these intelligent objects to collect and exchange
data [1]–[3] for different purposes. For example, wireless
sensor networks (WSNs) are a typical type of IoT network,
in which the sensors can detect and monitor the network
area. WSNs are developed based on microelectromechanical
systems (MEMS), system on chip (SoC), wireless commu-
nications and low-power embedded technology. At present,
WSNs are widely used in military applications [4], intel-
ligent transportation [5], civilian domains [6] and other
fields [7]–[9]. In addition, WSNs can be used for collecting
this network can provide great convenience for our daily
lives [11], [12].
WSN-assisted IoT in a wired network has the advantages
of low cost, convenient deployment and good scalability.
However, the defects cannot be ignored. One of the main
drawbacks is limited energy resources. In fact, sensor nodes
are generally powered by batteries, and it is difficult to add
energy to the sensor nodes due to the rugged environments
where they operate, which makes energy management an
important task for WSN-assisted IoT.
Due to the temporary validity of sensors and the high
price of replacement, replacing the node components is not
worthwhile. Hence, prolonging the lifetime of the network

2169-3536
2017 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 5, 2017 Personal use is also permitted, but republication/redistribution requires IEEE permission. 18469
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management
and managing the energy resources throughout the entire
network are the most important aspects when improving
the performance of WSN routing protocols. The low-energy
adaptive clustering hierarchy (LEACH) [13] protocol is one
of the most common protocols in this field. A considerable
number of new protocols have been designed to improve the
performance of LEACH [14]–[21] by reducing the energy of
the cluster head (CH) nodes or finding a multi-hop path from
CHs to the base station (BS) [22]–[28].
To simplify the management of the network, the concept
of cluster is proposed, and the CH nodes are the managers
of local clusters. A CH node takes the responsibilities of
organizing the cluster, establishing the routing table, and
collecting, compressing, and transmitting data. Because of its
high frequency of utilization, the energy consumption of a
CH node is the fastest among the entire network. Searching
for a multi-hop path from the CHs to the BS is fundamen-
tal for reducing the energy consumption of CH nodes. The
energy consumption of long-distance node communication
is very large. If the network can find a node that does not
considerably contribute to the network and that is located at
the edge of the cluster to be the forwarder, then the energy
dissipation rate of CH nodes will be significantly reduced.
This approach will not reduce the life cycle of the entire
network. As mentioned in [29], at the end of the entire sensor
network life cycle, the energy consumption of sensor nodes at
the edge of the network is only 10%. Thus, selecting an edge
node as an intermediate node for communication between a
CH and BS will not reduce the life cycle of the entire network.
To summarize, controlling the energy consumption of CH
nodes plays an important role in prolonging the lifetime of
the network and achieving energy efficiency.
Aiming at a higher energy efficiency for the entire network,
a new protocol named energy-efficient centroid-based routing
protocol (EECRP) is proposed to manage energy resources in
WSN-assisted IoT. The main contributions of this paper are
as follows:
• A clustering algorithm that operates according to the
position of the energy centroid and the residual energy
of nodes is constructed.
• An optimization algorithm that is based on the number
of dead nodes and CH nodes is added to the protocol.
• To reduce the number of long-distance communications,
a protection mechanism for EECRP to save the energy
of CH nodes is established.
Notably, the above contributions substantially broaden the
field of applications of the energy-efficient routing protocol
by applying an energy centroid with local management and
global control. Moreover, the average energy consumption
of each round is reduced without impacting the network
lifetime.
The remainder of this paper is organized as follows.
In section II, the related works and the motivation for our
paper are discussed. Our proposed EECRP is described in
detail in section III. In section IV, the energy consumption
model of EECRP is presented. In section V, the performance
18470
evaluations of our protocol and comparisons with other proto-
cols are presented. Conclusions are drawn in section VI, and
future works are also mentioned.
II. RELATED WORK
Cluster routing protocols work on the network layer, which
connects the data link layer and the transport layer. When
deployed in the network, these protocols can avoid the ‘‘hot
spot’’ problem and obtain better network performance. In pre-
vious studies, clustering algorithms play a very important
role in designing cluster routing protocols. Here, we mainly
discuss the clustering and CH selection algorithms in hierar-
chical networks.
LEACH [13] is the first hierarchical routing protocol
designed for WSNs. The main characteristics of LEACH
include local cluster generation and dynamic CH node rota-
tion combined with data fusion technology. In LEACH,
the node uses a stochastic algorithm, which is shown as
Eq. (1), to decide whether to be a CH node. The details of
this stochastic algorithm are discussed in [27].
P
, if n N
(
1
T (n) = 1−P·(r mod P ) (1)
0 , otherwise
However, the process of CH node selection leads to a
significant amount of energy consumption. The large amount
of transmitted data leads to a short lifetime of nodes and a
lack of monitoring of the area filled with dead nodes. Note
that the extendibility of LEACH is not suitable for large-scale
networks.
LEACH-C is an improved version of LEACH.
In LEACH-C, at the beginning of each round, nodes send
information of their position and residual energy to the
BS. After receiving this information, the BS calculates the
average energy of all nodes. Nodes whose energy is lower
than the average energy will not be selected to be a candidate
CH node. In this way, the energy consumption resulting
from the CH node selection process and the overhead caused
by data transmission are significantly reduced. However,
the large amount of data transmission tasks at the beginning
of each round still cause considerable overhead. In addition,
LEACH-C does not perform as well as LEACH when the BS
is located inside of the sensor network.
Lindsey et al. [30] introduced a protocol called
power-efficient gathering in sensor information sys-
tems (PEGASIS), in which a chain is formed for all nodes
to transfer data packets to the BS. Loscri et al. [31] proposed
a two-level hierarchy routing protocol (TL-LEACH), which
uses random rotations of local cluster base stations. In this
way, TL-LEACH can better distribute the energy load among
the sensors in the network, particularly when the density of
the network is higher. Wei et al. [32] proposed a distributed
clustering algorithm (EC) that determines a suitable cluster
size based on the hop distance to the BS. By applying EC,
the network makes a good balance between the cluster’s
energy consumption and the lifetime of nodes. In 2016,
VOLUME 5, 2017
J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management
Razaque et al. [33] combined the features of LEACH and
PEGASIS to improve the energy efficiency in routing. In the
same year, Razaque et al. [34] designed H-LEACH, which
is used to solve problems of energy considerations while
selecting a CH. H-LEACH considers the residual and max-
imum energies of nodes for every round while selecting a
CH using threshold conditions. Lin et al. [35] took advan-
tage of a game model to select CH nodes. In addition,
a routing protocol named game-theory-based energy-efficient
clustering (GEEC) was proposed. GEEC, which is a type
of clustering routing protocol, adopts an evolutionary game
theory mechanism to achieve energy balance and longevity.
From the above analysis, note that CH node selection
algorithms should take four aspects into consideration: the
local autonomy, network coverage, node location and the
remaining energy of the node. Previous CH node selection
algorithms mostly meet one or two with respect to three
aspects, and there is no comprehensive consideration of the
influence of all four aspects.
III. ENERGY-EFFICIENT CENTROID-BASED
ROUTING PROTOCOL (EECRP)
In this section, the centroid-based routing protocol is dis-
cussed in detail. Subsequently, the energy consumption
model is introduced. Moreover, some definitions, terminolo-
gies and assumptions are presented for a better understanding.
A. SYSTEM MODEL
1) SOME ASSUMPTIONS AND NOTATIONS
We assume that wireless sensor nodes are randomly dis-
tributed in the network. Once the arrangement of the entire
sensor network is completed, the positions of the sensor nodes
will not be changed. In addition, the location information of
the node is already loaded into the node when the network is
deployed. We also assume that each node knows the position
of the BS and its remaining energy at any time. The shape
of the entire tested area that the nodes are distributed in is
a rectangle. We create a Cartesian coordinate system with
its origin point located at the lower left corner of the entire
region. In EECRP, CH nodes have direct communication with
the BS. We set the number of CH nodes at the percentage
of 5% during the simulation.
2) SENSOR ENERGY CONSUMPTION MODEL
The sensor energy model is an important metric used to mea-
sure the performance of EECRP. The model that we choose is
a popular one mentioned in previous works [36]. The model
is shown as Eq. (2).
(
l · (er + et + fs d 2 ) if d < dTh
E= (2)
l · (er + et + mp d 4 ) if d > dTh
In Eq. (2), E is the total energy consumption when deliv-
ering a single l-bit packet from a sender to a receiver. The
baseline energy consumption levels at the sender and receiver
radios are indicated by er and et , respectively. The value
VOLUME 5, 2017
of d is the distance of the link between the sender and
receiver. The transmission energy consumption is denoted by
either fs d 2 or mp d 4 depending on the distance d and the
distance threshold dTh . For d ≤ dTh , fs is used to reflect
‘‘free-space’’ conditions, while mp represents longer links
potentially affected by ‘‘multi-path’’ fading.
B. CLUSTERING ALGORITHM SCHEME
The clustering algorithm is used to find the most appropriate
CH node for the cluster. The algorithm includes three phases:
the initialization phase, the first cluster head selection phase
and the rotate phase.
1) INITIALIZATION PHASE
A LOCATION message is first sent to the BS from every
sensor node. The format of a LOCATION message is shown
in Fig. 1. The field Message type indicates that it is a LOCA-
TION message. The field Sender ID contains the node ID of
the message sender. The X coordinate represents the abscissa
of the node0 s location. The Y coordinate provides the ordinate
of the node0 s location. The Energy level provides the energy
status of the node.
FIGURE 1. The format of the LOCATION message.
At the end of this phase, the BS will calculate the distance
between each node and the BS itself in the network. The
BS clusters based on the distances. Subsequently, the BS
will update the node table, including each node0 s position
and energy. Then, the BS broadcasts a FEEDBACK message
to the nodes in one cluster specifically. The format of the
message is shown in Fig. 2. Message type is used to specify
the type of the message to inform the receiver. The MAX-dist
field, which is calculated by the BS, delivers the maximum
broadcast range to each node in the network. The CH0 s ID
represents the ID of the CH node in each cluster. The Avg-
energy indicates the average energy of the network. The
receiver will update all the information of the FEEDBACK
message in its routing table.
FIGURE 2. The format of the FEEDBACK message.
The main task of the initialization phase is to exchange
messages between the BS and the sensor nodes. The messages
mainly contain the location and energy information of the
nodes, the average energy of the entire network, the CH
nodes chosen by the BS in the first round and the longest
transmission distance. After implementing the entire phase,
the exchanged information will be stored in the routing tables
18471
 J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management

of the BS and sensor nodes. In addition, the information of the 2) The distance from the energy centroid of the network to
routing table will update in real time as the entire network the node is less than the average distance of the entire
operates. nodes to the energy centroid.
Moreover, in the process of CH node rotation, the numbers
2) FIRST CLUSTER HEAD SELECTION PHASE of CH nodes and dead nodes are taken into consideration.
After the LOCATION message and FEEDBACK message In other words, the number of clusters decreases as the num-
are received, the BS determines the node whose energy is ber of dead nodes increases to keep the value of P unchanged.
greater than the Avg-energy. Specifically, the first round of Here, note that P represents the desired percentage, which is
CH node selection is random because the energy levels of defined in section II.
each node are identical. The selection criteria include the
most suitable percentage P and the entire network, which C. CENTROID ALGORITHM SCHEME
should be monitored by being covered by clusters. In fact, In the field of mathematics, the centroid is the center of
in the first CH node selection phase, each node in a cluster weight, which is the imaginary point of mass concentration.
checks its own ID to determine whether it is a CH node. Centroid position is important in engineering fields. In our
The first cluster head selection phase is mainly to complete study, we use the concept of the cluster energy centroid rather
the identification of the broadcast information of the BS. than the traditional weight centroid. The reasons for using
When the ID of the node is the same as that of the CH node, the cluster energy centroid are as follows. First, the weight of
the node confirms that the node itself is the CH node and sensor nodes in the network is meaningless. Second, the cen-
opens the transmitting antenna to prepare for the next phase. ter of the weight centroid of nodes for the entire cluster is
When the IDs are different, the node closes the transmitting meaningless because node location and weight do not change
antenna to save energy and opens the receiving antenna to in the operation of the network. Finally, in the entire network,
prepare for the arrival of information from the CH node. the energy of the sensor nodes is the only factor that changes.
The energy centroid can intuitively display the distribution of
3) ROTATE PHASE residual energy in the network.
After the first cluster selection phase, the first CH node Note that Eqs. (3) and (4) are used to calculate the posi-
broadcasts the schedule message to the neighbor nodes. The tion of the weight centroid in the field of mathematics.
CH node0 s ID and position are contained in this message. Eqs. (5) and (6) are used to calculate the position of the energy
All CH nodes in the network send this message. When the centroid. Clearly, we take the residual energy and the position
neighbor nodes receive this message, they will determine of node i into consideration to calculate the energy centroid.
whether it belongs to this cluster based on the CH node0 s ID in Next, the meanings of the parameters used in the equations
the FEEDBACK message and the schedule message. At this are introduced. µ is the density of nodes’ weights in the
point, the clustering is completed. The general nodes send the cluster, S is the measure of cluster area, dσ is the differential
information about their locations and energy to the associated of weight, and dMx and dMy are the static moment to the
CH node. The CH node calculates the location of the energy x-axis and y-axis, respectively. Ei_rs represents the residual
centroid of the cluster. The node that is nearest to the energy energy level of node i, E0 is the initial value of energy, and
centroid will be chosen to be the candidate CH node. X and Y are the X coordinate and the Y coordinate of node
The rotate phase is established to choose a candidate i, respectively. Xwc and Ywc are the results of the weight
CH node. In this way, the network can uniformly distribute centroid. Xec and Yec are the results of the energy centroid
the energy consumption to all nodes in the network. EECRP
x · µdσ
RR
simultaneously meets the four aspects that were mentioned My D
in section II. In the first round, the CH nodes are chosen by Xwc = = RR (3)
M µdσ
the BS. Thus, the BS has an overall view of the network.
When the network is running, the CH node is selected in the RRD
x · µdσ
local cluster, which means that the network is self-adapting. Mx D
Moreover, the selected CH node is the closest node to the Ywc = = RR (4)
M µdσ
energy centroid that can maximize the network coverage. D
From the calculation of the position of the energy centroid, n
X Ei_rs
which is described in detail below, we find that the calculation · X
E0
is based on the node location and the remaining energy of Xec =
i=0
(5)
the node. In other words, the cluster algorithm in EECRP can N
n
make a improvement to the existing algorithms. X Ei_rs
· Y
Then, a general node is added to the candidate CH node E0
i=0
sequence. The added nodes must meet the following two Yec = (6)
conditions. N
1) Its energy level is more than the average energy level If the weights of sensor nodes in the cluster are known
of the clusters. and evenly distributed, then we can use Eqs. (3) and (4) to

18472 VOLUME 5, 2017

J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management

calculate the position of the weight centroid [37]. However, Algorithm 1 The Process of EECRP
in the real case, the effect of the node weight of the nodes in BS ← LOCATION messages
the network is not important for the network lifetime. Hence, Calculate: Avg-energy and MAX-dist
with node location information and the residual energy level, Nodes ← FEEDBACK message
we can use Eqs. (5) and (6) to calculate the position of the if (Candidate CHs exist) then
energy centroid. The energy centroid can reflect the energy Clustering
distribution during smooth operation of the network. Recalculate centroid
Create schedule
D. PROTECTIVE MECHANISM SCHEME if (Alive node number > number of clusters) then
As mentioned above, the MAX-dist value is broadcast to each if (Round time is not over then
node by the BS with the FEEDBACK message. This value is CH ← data (normal nodes)
the protective mechanism of EECRP. The MAX-dist, calcu- Fusion data
lated by the BS, is the communication threshold. Note that BS ← data (CH nodes)
the MAX-dist value is calculated by taking the Avg-energy return
in Eq. (2). When CH nodes are ready to send packets to the end if
BS, the CH nodes compare their own distances to the BS with Recalculate: Avg-energy and MAX-dist
the MAX-dist sent by the BS in the FEEDBACK message. Nodes ← FEEDBACK message
If the distance is smaller than the MAX-dist, then CH nodes Select candidate CH nodes
will turn on their antennas and send packets. Conversely, if the return
distance is larger than the MAX-dist, then CH nodes will stop end if
transmitting the data to the BS and store the packets, waiting return
for the next round. Although this mechanism will lead to a end if
loss of data packets in the short term, it can avoid the long-
distance communications of CH nodes, which can reduce the
Clearly, the relationship between lloc and lfd can be described
energy consumption of the network, as shown in Eq. (16)
as lfd = 0.8· lloc . σ indicates the density distribution of sensor
in section IV. From the perspective of monitoring the entire
nodes. XB and YB are the location information of the BS.
network, it is more good than harm to utilize the protective
Xi and Yi are the location information of ordinate nodes.
mechanism.
XC and YC are the location information of CH nodes. ab
The details of EECRP are presented in Algorithm 1.
shows the size of the entire sensor network. P indicates the
desired percentage of CH nodes. ldata is the size of the packet
IV. ENERGY CONSUMPTION MODEL that has been compressed by the CH node. ldata0 is the size of
In this section, we provide the value of energy consumption the packets that are sent to the CH node by common nodes.
Eround of the network during a round period. This value is The symbols used in the following equations are the same as
needed by the EECRP algorithm to calculate the equalized in Eq. (7).
lifetime value and to determine the effect of the protective
mechanism. The total initial deployment energyPin the region
2
Ein1 = lloc {π dTh σ et
P
can be represented as E0_all = j  Ci E0 (j) = j  Ci E0 (j) · π dTh
2 σ

π ri2 σ = E0 · abσ , where E0 is the average deployment

energy and ri denotes the radius of cluster i. The lifetime
E0_all
of the entire network is L ≈ Eround . The round energy is
the sum of energy consumption values for cluster formation,
route discovery and data communication events. The round
energy is given by Eround = Ecluster + Ecomm + Ein−comm .
In the following parts of this section, the individual phases of
EECRP are separately calculated.
A. ENERGY CONSUMPTION IN CLUSTER FORMATION
The energy consumption of nodes sending a LOCATION
message to the BS is calculated in this section. The energy
consumption can be described as Eq. (7) when the distance
of the link between the sender and receiver is less than dTh .
We assume that the area is circular with the centroid as the
center and dTh as its radius.
In Eq. (7), lloc represents the size of the LOCATION mes-
sage. lfd represents the size of the FEEDBACK message.
X
+ fs [(XB − Xi )2 +(YB − Yi )2 ]} (7)
i=1
The energy consumption can be described as Eq. (8) when
the distance of the link between the sender and receiver
is larger than dTh . The energy consumption of transmitting
packages by nodes with a larger d in Eq. (8) is the exponential
time of that in Eq. (7). A conclusion can be drawn that
shorter long-distance transmission times can contribute to
lower energy consumption.
Eout1 = lloc {(ab − π dTh
2
)σ et
2 )σ
(ab−π dTh
X
+ mp [(XB − Xi )2 + (YB − Yi )2 ]2 } (8)
i=1
Eqs. (9) and (10) describe the energy consumed by nodes to
receive the FEEDBACK message from the BS. The condition
of receiving the LOCATION message is the same as that of
receiving the FEEDBACK message. Eq. (9) and Eq. (10) are

VOLUME 5, 2017 18473

 J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management

related to the receiving energy. We can determine that in this
phase, when the nodes are located far from the BS, the number
of long-distance communications is two times the number
of nodes. This result demonstrates that the conclusion drawn
from Eq. (8) holds.
2
Ein2 = lfd {π dTh σ er
π dTh
2 σ
X
+ fs [(XB − Xi )2 + (YB − Yi )2 ]} (9)
i=1
Eout1 = lfd {(ab − π dTh
2
)σ er
2 )σ
(ab−π dTh
X abPσ
mp [(XB − Xi )2 +(YB − Yi )2 ]2 } (10)
+ +
i=1
After exchanging LOCATION and FEEDBACK messages,
normal nodes join the nearest cluster. The value of l0 is
the size of the control message. The energy consumption
can be calculated as Eq. (11). In contrast to the formulas
for calculating information exchange, Eq. (11) is used to
calculate the energy consumptions of both the sender and the
receiver. This is because the BS is independent of the entire
network in our calculations. In addition, the energy of the BS
is limitless, which indicates that the broadcast and receive
energy consumptions of the BS do not need to be calculated.
Enon_ch = l0 {ab(1 − P)σ et + abPσ er
ab(1−P)σ
X
+ fs [(XC − Xi ) + (YC − Yi ) ]}
2 2
(11)
i=1
The total energy consumption in the cluster forma-
tion phase is shown in Eq. (12), which is the sum of
Eqs. (7) ∼ (11).
Ecluster = Ein1 + Eout1 + Ein2 + Eout2 + Enon−ch (12)
After a simple mathematical reduction using
l = 0.8· lloc = lfd , Eq. (12) can be simplified as Eq. (13). The
relationship between LOCATION and FEEDBACK messages
is shown in subsection III(B). Here, the problem of long-
distance communication must be mentioned again. From
Eq. (13), we can observe that the energy consumption of
the communication channel is an important component of the
energy consumption of the entire network. Thus, it is essential
to reduce the channel energy consumption.
Ecluster = l · {1.25abσ et + abσ er
π dTh
2 σ
X +
+ 2.25 fs [(XB − Xi )2 + (YB − Yi )2 ]}
i=1
2 )σ
(ab−π dTh
X abPσ
+ 2.25 mp [(XB − Xi )2 + (YB − Yi )2 ]2 }
i=1
+ l0 {ab(1 − P)σ et + abPσ er
ab(1−P)σ
X ab(1−P)σ
+ fs [(XC − Xi )2 + (YC − Yi )2 ]} (13)
i=1
B. ENERGY CONSUMPTION IN TRANSMITTING
DATA TO THE BS
In this part, the energy consumption model is very simple
because only cluster head nodes consume energy during
transmission. Moreover, the protective mechanism has a con-
siderable influence on the performance. The result is shown
in Eq. (14). Thanks to the protective mechanism, the long-
distance communication is greatly reduced in Eq. (14). The
purpose of doing so is to prolong the survival time of
CH nodes, thereby prolonging the network life cycle.
Ecomm = ECH −BS = ldata {abPσ et
X
fs [(XB − XCi )2 + (YB − YCi )2 ]} (14)
i=1
C. ENERGY CONSUMPTION IN TRANSMITTING DATA
TO THE CH NODES IN CLUSTERS
In this part, the normal nodes transmitting data packets con-
sume their own energy, and the cluster head nodes receiving
data also consume energy. Because data packets are transmit-
ted in one cluster, the value of d is smaller than that of dTh .
The energy consumption model is shown in Eq. (15). The
calculation in Eq. (15) is similar to that in Eq. (11). Sending
and receiving information in Eq. (15) is the data information
perceived by sensors rather than control information.
Ein−comm = Ein−cluster = ldata0 {ab(1 − P)σ et + abPσ er
ab(1−P)σ
X
+ fs [(XC − Xi )2 + (YC − Yi )2 ]} (15)
i=1
D. TOTAL ENERGY CONSUMPTION IN A ROUND
The total energy consumption in a round is the sum of Ecluster ,
Ecomm and Ein−comm , which is shown in Eq. (16).
Eround = Ecluster + Ecomm + Ein−comm
l · {1.25abσ et + abσ er
π dTh
2 σ
X
+ 2.25 fs [(XB − Xi )2 + (YB − Yi )2 ]}
i=1
2 )σ
(ab−π dTh
X
+ 2.25 mp [(XB − Xi )2 + (YB − Yi )2 ]2 }
i=1
+ l0 {ab(1 − P)σ et + abPσ er
ab(1−P)σ
X
fs [(XC − Xi )2 + (YC − Yi )2 ]}
i=1
+ ldata {abPσ et
X
+ fs [(XB − XCi )2 + (YB − YCi )2 ]}
i=1
+ ldata0 {ab(1 − P)σ et + abPσ er
X
+ fs [(XC − Xi )2 + (YC − Yi )2 ]} (16)
i=1

18474 VOLUME 5, 2017

J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management

As shown in Eq. (16), the energy consumption of the TABLE 1. Parameters used in simulation.
network using EECRP is the information exchange in the net-
work initialization phase. This means that when the network
enters the stable operation phase, the energy consumption
of the normal nodes can be ignored. First, a considerable
amount of energy consumption occurs in the initial phase of
the network. As mentioned in section III, the nodes need to
send the LOCATION message to the BS at the very beginning.
The distance between the node and the BS is relatively larger
than that between the node and the CH node in one cluster.
In addition, nodes also need to receive the FEEDBACK mes-
sage sent from the BS in the initialization phase. For most of
the nodes, these two communications belong to long-distance
communication; thus, the energy consumption will be very
significant. Second, the energy consumption of the nodes
in the stable operation phase is relatively low. Hierarchical
sensor networks are known to achieve local clustering; thus,
the local normal node only communicates with the CH node
of the cluster. The average transmission distance in the cluster
is highly related to the location of the CH node. In this paper,
we propose a CH node selection algorithm based on the loca-
tion of the energy centroid, where the CH node is located in
the energy center of the cluster and the most concentrated area
of the nodes. In LEACH, the CH node is selected by Eq. (1).
If the CH node is located at one edge of the network, then the
transmission distance of nodes on the other side of the net-
work will be greatly increased. As time passes, the network’s
coverage rate will be greatly reduced, which will directly lead FIGURE 3. The rounds at which the first and last nodes die versus
different values of Ps.
to a higher error rate of data. Regarding the stable operation
phase of the proposed EECRP, the CH node of the next round
is calculated by the CH node of the previous round. Note that A. SIMULATION PARAMETERS
the candidate CH node is located in the center of the energy
In the simulation part, the network is implemented in a
network; thus, it can greatly improve the ability of real-time
100 m × 100 m field with 100 nodes randomly distributed
monitoring.
in the network area. Each node is set to have 2 J as its
Moreover, when the distance between the CH node and
initial energy. Furthermore, the BS is located in the sensor
the local node is larger than the MAX-dist, the data will not
network. The detailed parameters that were used in our sim-
be sent and will be stored locally until a neighbor node is
ulation are listed in TABLE 1. In addition, the simulations
selected to be a new CH node. The number of long-distance
among LEACH, LEACH-C, GEEC and EECRP are also
communications is reduced by the protective mechanism in
implemented using the ns-2 simulator.
EECRP. In addition, the number of long-distance communi-
The following three performance measurements are con-
cations between CH nodes and the BS is also greatly reduced.
sidered: the total number of nodes that are alive, the total
The stored data packets will be sent to new CH nodes in the
energy dissipation and the number of messages received at
following rounds. The influence of the energy consumption of
the BS. The lifetime of the network is defined as the round
long-distance communication on the survival time of cluster
number when the remaining nodes are not sufficient to form
heads is remarkable. The protective mechanism plays a very
clusters.
important role, as indicated by the simulation results and
comparisons.
B. SIMULATION RESULTS
The results of part 1 are shown in Figs. 3, 4, and 5. To evaluate
V. PERFORMANCE EVALUATION the setting of the value of P in Eq. (1), we consider the number
Taking the location of the BS into consideration, the selection of nodes that are alive, the size of transmitted data and the
of the value of P in Eq. (1) and the comparison between total energy dissipation through the entire network with the
LEACH and LEACH-C are implemented using the ns-2 sim- BS located in the network area. The value of P ranges from
ulator [38]. The version of ns-2 used in our simulation is ns- 1% to 12%.
2.35. The software platform that we run the protocols on is As vividly shown in Fig. 3, when the value of P is small,
Ubuntu 12.04. which represents a small number of CH nodes, the nodes

VOLUME 5, 2017 18475

 J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management
FIGURE 4. Average data transmission versus different values of Ps.
FIGURE 5. The total energy dissipated versus different values of Ps.
begin to die in a very short time. This result might be due
to the shortage of small Ps, although the lifetime of the
entire network becomes longer. This results in large parts of
the network being not under monitoring in the later rounds.
In addition, Fig. 3 indicates that the average data transmission
of networks with fewer CH nodes is larger, which means that
the data integrity is poor.
However, when there is a large number of CH nodes in the
sensor network, all aspects of the sensor network are in poor
condition. In detail, shorter lifetime, smaller data transmis-
sion and quicker energy dissipation become the shortages of
the networks with large Ps, as shown in Figs. 3, 4, and 5.
Taking all aspects into consideration, setting the percent-
age of CH nodes to 5% leads to a good condition of the
network. The energy of nodes dissipates slowly, and as the
nodes die, network area that is not being monitored can
hardly be found. In other words, considering the influence
on lifetime and the rate of energy dissipation, 5% should be
the best setting of CH nodes.
C. COMPARISON RESULTS
EECRP and the other three protocols (LEACH,
LEACH-C and GEEC) are simulated in this part. Each of
18476
FIGURE 6. The number of sensor nodes that are alive in the four
protocols.
the four protocols is analyzed from three different aspects,
which are the number of sensors still alive, the number of
data signals received by the BS and energy dissipation.
In this part, as illustrated in Figs. 6, 7, and 8, several met-
rics are taken into consideration to evaluate the performances
of LEACH, LEACH-C, GEEC and EECRP when the BS is
located inside of the network. The detailed comparison of the
four protocols is as follows.
1) THE NUMBER OF NODES THAT ARE ALIVE
The number of nodes that are alive indicates the lifetime of
WSNs, which is the most important parameter of routing pro-
tocols. As shown in Fig. 6, there are considerable differences
in the number of sensor nodes that are alive among EECRP,
LEACH, LEACH-C and GEEC. First, the first node dies
at approximately the 100th round in LEACH-C. Regarding
LEACH, GEEC and EECRP, the first node dies at approxi-
mately the 400th round. Second, the nodes in LEACH-C die
in a very short time due to the large amount of communicating
control messages, whereas the network using EECRP has a
very long lifetime. The lifetime of EECRP is longer than that
of LEACH by approximately 100 rounds and longer than that
of GEEC by approximately 50 rounds.
2) THE NUMBER OF MESSAGES RECEIVED BY THE BS
Fig. 7 shows that LEACH-C delivers considerably fewer
messages to the BS than any other protocols. The num-
bers of messages received by the BS in the network using
EECRP are equal to those using LEACH and GEEC before
the 400th round. However, after the 400th round, EECRP0 s
packet delivery slowly increases, which is due to the setting
of MAX-dist in EECRP. When increasingly more nodes die,
the distance between two nodes in the network increases.
The node will stop sending information to the CH node
and the BS to save energy when the distance to the CH
node or BS exceeds the value of MAX-dist. Nodes will store
the data packets in their storage spaces. However, the mes-
sages that are stored will be sent in the following round
until the communication distance is less than the MAX-dist.
VOLUME 5, 2017
J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management
TABLE 2. Comparison among EECRP, LEACH and LEACH-C.
FIGURE 7. The number of messages received by the BS in the four
protocols.
The details about this protective mechanism are presented in
section III.
3) ENERGY DISSIPATION
As shown in Fig. 8, the speed of total energy dissipation
of LEACH-C remains at a very high level. In other words,
LEACH-C performs not as well as LEACH, GEEC and
EECRP in this aspect. In addition, the speed of energy con-
sumption of EECRP is slower than that of LEACH after the
400th round, which means that nodes in EECRP can monitor
the network for a long period. In addition, GEEC and EECRP
have similar performance in terms of energy consumption.
D. COMPARISON OF EECRP AND THE CONVENTIONAL
PROTOCOLS
The performance and comparison of the characteristics of
LEACH, LEACH-C, GEEC and EECRP are presented in
TABLE 2. The characteristics used to compare the protocols
are life cycle, scalability, computation and communication
overhead, path selection, location awareness and mobility.
As shown in TABLE 2, EECRP performs better than
LEACH, LEACH-C and GEEC in terms of life cycle, scal-
ability, computation and communication overhead and loca-
tion awareness.
However, with static sensor nodes and a single hop between
the BS and CH nodes, EECRP is suitable for the networks
VOLUME 5, 2017
FIGURE 8. The total energy dissipation in the four protocols.
where sensor nodes do not need to change locations. Finally,
it is very suitable for us to deploy EECRP in networks where
the BS is located inside the network.
VI. CONCLUSION
In this paper, we propose a new energy-efficient centroid-
based routing protocol (EECRP) to manage the energy of
WSN-assisted IoT by solving the problem of forming clus-
ters, which is based on the distance to the energy centroid.
We also propose an optimization algorithm based on the
number of dead nodes and the number of cluster head nodes.
From the simulation results, when the BS is located in the
network, the EECRP could transmit a considerable amount
of data with very low energy dissipation. Meanwhile, the
network lifetime of EECRP is longer than that of LEACH,
LEACH-C and GEEC. In future work, we want to improve the
protocol by finding the multi-hop path from the CH nodes to
the BS. The multi-hop path is used by CH nodes to transmit
data packets. We hope that our future protocol can perform
well when the BS is located outside of the network.
REFERENCES
[1] C. Floerkemeier, M. Langheinrich, E. Fleisch, F. Mattern, and S. E. Sarma,
‘‘The Internet of Things,’’ Electron. World, vol. 297, no. 6, pp. 949–955,
2017.
[2] M. Díaz, N. Cristian, and B. Rubio, ‘‘State-of-the-art, challenges, and
open issues in the integration of Internet of Things and cloud computing,’’
J. Netw. Comput. Appl., vol. 67, pp. 99–117, May 2016.
18477
 J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management
[3] S. Misra, M. Maheswaran, and S. Hashmi, ‘‘Securing the Internet of
Things,’’ in Security Challenges and Approaches in Internet of Things.
2017, pp. 39–51.
[4] T. Alhmiedat, A. Abu Taleb, and M. Bsoul, ‘‘A study on threads detection
and tracking systems for military applications using wsns,’’ Int. J. Comput.
Appl., vol. 40, no. 15, pp. 12–18, 2012.
[5] G. Xing, M. Li, T. Wang, W. Jia, and J. Huang, ‘‘Efficient rendezvous
algorithms for mobility-enabled wireless sensor networks,’’ IEEE Trans.
Mobile Comput., vol. 11, no. 1, pp. 47–60, Jan. 2012.
[6] A. Liu, J. Ren, X. Li, Z. Chen, and X. S. Shen, ‘‘Design principles and
improvement of cost function based energy aware routing algorithms for
wireless sensor networks,’’ Comput. Netw., vol. 56, no. 7, pp. 1951–1967,
2012.
[7] J. Shen, H. Tan, J. Wang, J. Wang, and S. Lee, ‘‘A novel routing protocol
providing good transmission reliability in underwater sensor networks,’’
J. Internet Technol., vol. 16, no. 1, pp. 171–178, 2015.
[8] C. F. Lai, S. Zeadally, J. Shen, and Y. X. Lai, ‘‘A cloud-integrated appliance
recognition approach over Internet of Things,’’ J. Internet Technol., vol. 16,
no. 7, pp. 1157–1168, 2015.
[9] J. Zhu, J. Liu, Z. Hai, and Y. G. Bi, ‘‘Research on routing protocol facing to
signal conflicting in link quality guaranteed WSN,’’ Wireless Netw., vol. 22,
no. 5, pp. 1739–1750, 2016.
[10] Q. Chi, H. Yan, C. Zhang, Z. Pang, and L. D. Xu, ‘‘A reconfigurable smart
sensor interface for industrial WSN in IoT environment,’’ IEEE Trans. Ind.
Informat., vol. 10, no. 2, pp. 1417–1425, May 2014.
[11] J. Shen, T. Zhou, D. He, Y. Zhang, X. Sun, and Y. Xiang, ‘‘Block
design-based key agreement for group data sharing in cloud
computing,’’ IEEE Trans. Depend. Sec. Comput., to be published,
doi: 10.1109/TDSC.2017.2725953.
[12] J. Shen, J. Shen, X. Chen, X. Huang, and W. Susilo, ‘‘An efficient
public auditing protocol with novel dynamic structure for cloud
data,’’ IEEE Trans. Inf. Forensics Security, to be published,
doi: 10.1109/TIFS.2017.2705620.
[13] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, ‘‘Energy-
efficient communication protocol for wireless microsensor networks,’’ in
Proc. 33rd HICSS, 2000, pp. 10–20.
[14] A. Yektaparast, F.-H. Nabavi, and A. Sarmast, ‘‘An improvement on
LEACH protocol (cell-LEACH),’’ in Proc. IEEE ICACT, Feb. 2012,
pp. 992–996.
[15] Y. Wu, S. Fahmy, and N. B. Shroff, ‘‘Energy efficient sleep/wake schedul-
ing for multi-hop sensor networks: Non-convexity and approximation
algorithm,’’ in Proc. IEEE INFOCOM, May 2007, pp. 1568–1576.
[16] D. Kandris, P. Tsioumas, A. Tzes, G. Nikolakopoulos, and D. D. Vergados,
‘‘Power conservation through energy efficient routing in wireless sensor
networks,’’ Sensors, vol. 9, no. 9, pp. 7320–7342, 2009.
[17] C. Lung and C. Zhou, ‘‘Using hierarchical agglomerative clustering in
wireless sensor networks: An energy-efficient and flexible approach,’’ Ad
Hoc Netw., vol. 8, no. 3, pp. 328–344, 2010.
[18] B. Wang, X. Gu, L. Ma, and S. Yan, ‘‘A variable threshold-value authen-
tication architecture for wireless mesh networks,’’ Int. J. Sensor Netw.,
vol. 23, no. 4, pp. 265–278, 2017.
[19] P. Raja and P. Dananjayan, ‘‘Game theory-based efficient energy consump-
tion routing protocol to enhance the lifetime of WSN,’’ Int. J. Inf. Commun.
Technol., vol. 8, no. 4, pp. 357–370, 2016.
[20] J. Shen, S. Chang, J. Shen, Q. Liu, and X. Sun, ‘‘A lightweight multi-layer
authentication protocol for wireless body area networks,’’ Future Generat.
Comput. Syst., to be published, doi: 10.1016/j.future.2016.11.033.
[21] P. Marappan and P. Rodrigues, ‘‘An energy efficient routing protocol for
correlated data using CL-LEACH in WSN,’’ Wireless Netw., vol. 22, no. 4,
pp. 1415–1423, 2016.
[22] J. Zhang, C. K. Jeong, G. Y. Lee, and H. J. Kim, ‘‘Cluster-based multi-
path routing algorithm for multi-hop wireless network,’’ Future Generat.
Commun. Netw., vol. 1, pp. 67–75, Jan. 2008.
[23] J. Shen, A. Wang, C. Wang, Y. Ren, and X. Sun, ‘‘A RFID based local-
ization algorithm for wireless sensor networks,’’ in Proc. ICCCS, 2016,
pp. 275–285.
[24] M. Chen, V. C. M. Leung, and S. Mao, ‘‘Directional controlled fusion
in wireless sensor networks,’’ Mobile Netw. Appl., vol. 14, no. 2,
pp. 220–229, Apr. 2009.
[25] J. Shen, D. Liu, J. Shen, Q. Liu, and X. Sun, ‘‘A secure cloud-assisted urban
data sharing framework for ubiquitous-cities,’’ Pervasive Mobile Comput.,
to be published, doi: 10.1016/j.pmcj.2017.03.013.
18478
[26] Y. Yuan, Z. He, and M. Chen, ‘‘Virtual MIMO-based cross-layer design
for wireless sensor networks,’’ IEEE Trans. Veh. Technol., vol. 55, no. 3,
pp. 856–864, May 2006.
[27] J. Shen, A. Wang, C. Wang, Y. Ren, and J. Wang, ‘‘Performance com-
parison of typical and improved LEACH protocols in wireless sensor
network,’’ in Proc. IEEE CCITSA, Dec. 2015, pp. 187–192.
[28] Y. Zhang, X. Sun, and B. Wang, ‘‘Efficient algorithm for k-barrier coverage
based on integer linear programming,’’ China Commun., vol. 13, no. 7,
pp. 16–23, 2016.
[29] A. Liu, Z. Zheng, C. Zhang, Z. Chen, and X. Shen, ‘‘Secure and energy-
efficient disjoint multipath routing for WSNs,’’ IEEE Trans. Veh. Technol.,
vol. 61, no. 7, pp. 3255–3265, Sep. 2012.
[30] S. Lindsey and C. S. Raghavendra, ‘‘PEGASIS: Power-efficient gath-
ering in sensor information systems,’’ in Proc. Aerosp. Conf., 2002,
pp. 1125–1130.
[31] V. Loscri, G. Morabito, and S. Marano, ‘‘A two-levels hierarchy for
low-energy adaptive clustering hierarchy (TL-LEACH),’’ in Proc. IEEE
VTC, Sep. 2005, pp. 1809–1813.
[32] D. Wei, Y. Jin, S. Vural, K. Moessner, and R. Tafazolli, ‘‘An energy-
efficient clustering solution for wireless sensor networks,’’ IEEE Trans.
Wireless Commun., vol. 10, no. 11, pp. 3973–3983, Nov. 2011.
[33] A. Razaque, M. Abdulgader, C. Joshi, F. Amsaad, and M. Chauhan,
‘‘P-LEACH: Energy efficient routing protocol for wireless sensor net-
works,’’ in Proc. IEEE LISAT, Apr. 2016, pp. 1–5.
[34] A. Razaque, S. Mudigulam, K. Gavini, M. Abdulgader, G. S. Krishna, and
F. Amsaad, ‘‘H-LEACH: Hybrid-low energy adaptive clustering hierarchy
for wireless sensor networks,’’ in Proc. IEEE LISAT, Apr. 2016, pp. 1–4.
[35] D. Lin and Q. Wang, ‘‘A game theory based energy efficient clustering
routing protocol for WSNs,’’ Wireless Netw., vol. 23, no. 4, pp. 1101–1111,
May 2017.
[36] J. Yick, B. Mukherjee, and D. Ghosal, ‘‘Wireless sensor network survey,’’
Comput. Netw., vol. 52, no. 12, pp. 2292–2330, Aug. 2008.
[37] J. Shen, C. Wang, A. Wang, X. Sun, S. Moh, and P. C. K. Hung,
‘‘Organized topology based routing protocol in incompletely predictable
ad-hoc networks,’’ Comput. Commun., vol. 99, pp. 107–118, Feb. 2016,
doi: j.comcom.2016.07.009.
[38] M. Khosrowpour, ‘‘Network simulator ns-2,’’ J. Inst. Image Inf. Television
Eng., vol. 65, pp. 946–949, Mar. 2011.
JIAN SHEN (M’09) received the B.E. degree from
the Nanjing University of Information Science and
Technology (NUIST), Nanjing, China, in 2007,
and the M.E. and Ph.D. degrees in computer sci-
ence from Chosun University, Gwangju, South
Korea, in 2009 and 2012, respectively. Since 2012,
he has been a Full Professor with the School of
Computer and Software, NUIST, Nanjing, China.
His research interests include computer network-
ing, security systems, mobile computing and
networking, ad-hoc networks and systems, and ubiquitous sensor networks.
ANXI WANG received the B.E. degree from Nan-
jing University of Information Science and Tech-
nology, Nanjing, China, in 2016, where he is
currently pursuing the M.E. degree. He focuses
on energy efficient routing protocols in wire-
less sensor network and security systems. His
research interests include ad-hoc networks and
systems, information security, and wireless sensor
networks.
VOLUME 5, 2017
J. Shen et al.: Efficient Centroid-Based Routing Protocol for Energy Management

CHEN WANG received the B.E. degree from CHIN-FENG LAI (SM’14) received the
Nanjing University of Information Science and Ph.D. degree from the Department of Engineering
Technology, Nanjing, China, in 2016, where he is Science, National Cheng Kung University, Tainan,
currently pursuing the M.E. degree. He focuses on Taiwan, in 2008. He is currently an Associate
information security and incompletely predictable Professor with the Department of Engineering
ad hoc networks. His research interests include Science, National Cheng Kung University. He
information security, ad hoc networks and sys- has authored or co-authored over 100 refereed
tems, and wireless sensor networks. papers in journals, conferences, and workshop
proceedings about his research areas within four
years. His research interests include multimedia
communications, sensor-based healthcare, and embedded systems. He is a
member of the IEEE Circuits and Systems and the IEEE Communications
PATRICK C. K. HUNG (M’94) is currently an Societies.
Associate Professor with the University of Ontario
Institute of Technology, Canada. He is a Found-
ing Committee Member of the IEEE Interna-
tional Conference of Web Services, the IEEE
International Conference on Services Comput-
ing, the IEEE Congress on Services, and the
IEEE Congress on BigData. He is an Associate
Editor of the IEEE TRANSACTIONS ON SERVICES
COMPUTING, the International Journal of Web Ser-
vices Research, and the International Journal of Business Process and
Integration Management, and also an Executive Group Member and
a Co-Ordinating Editor of the Information Systems Frontiers
by Springer.

VOLUME 5, 2017 18479