D2CRP A Novel Distributed 2-Hop Cluster Routing Protocol For Wireless Sensor Networks

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO.

20, 15 OCTOBER 2022 19575

D2CRP: A Novel Distributed 2-Hop Cluster


Routing Protocol for Wireless Sensor Networks
Chao Chen , Li-Chun Wang , Fellow, IEEE, and Chih-Min Yu , Senior Member, IEEE

Abstract—Typically, wireless sensor networks (WSNs) use I. I NTRODUCTION


limited-capacity batteries that cannot be recharged or replaced.
HE Internet of Things (IoT) is a network consisting of
In general, designing an energy-efficient routing protocol has a
significant impact on prolonging the network lifetime. In this arti-
cle, a novel distributed 2-hop cluster-routing protocol (D2CRP)
T devices that contain specific sensors to sense the moni-
toring region and transmit the collected data to the end users.
is introduced to achieve energy efficiency in WSNs. In the clus- IoT is widely used in many scenarios, such as COVID-19 pre-
ter formation phase, each node obtains the information of its vention [1], healthcare [2], [3], environment surveillance [4],
neighbor nodes within the 2-hop range to form the 2-hop clus-
ter in a fully distributed manner. The transmission distance and smart building [5], indoor localization [6], and military battle-
residual energy are jointly considered to determine the energy- field [7], [8]. Gartner reported that the worldwide IoT market
efficient cluster head (CH) in each 2-hop cluster. After the CH is is near 6 billion devices in the year 2020 [9]. McKinsey [10]
generated, each member node can transmit packets to its 1-hop estimated that IoT will be worth nearly U.S. $11,000 trillion by
neighbor or the CH directly. To reduce the overall transmission
distance for intercluster communication, multiple chains can be
2025. Furthermore, the emerging 6G wireless communication
formed among CHs via their adjacent CHs that are closer to the and mobile edge computing (MEC) contribute to promising
base station (BS). As a result, energy-efficient intracluster and growth in the demand for IoT devices [11]–[14].
intercluster routing can be achieved for packet transmission. In Wireless sensor network (WSN) is a key technology of
addition, the optimal cluster number of 2-hop clustering is for- IoT [15]. In a WSN, sensor nodes are deployed in a specific
mulated and derived to minimize the energy consumption for
both intracluster and intercluster communications. Simulation region to sense information from the external environment and
results show that the optimal cluster number of D2CRP can be are capable of establishing an ad hoc network [16]. According
achieved and D2CRP is effectively improved on the performances to specific requirements, these nodes are equipped with corre-
of network lifetime, energy consumption, and packet transmission sponding functional sensors to monitor various conditions in
than the other four state-of-the-art competitive routing protocols,
the region, such as air pressure, temperature, etc. [17]. After
including low-energy adaptive clustering hierarchy (LEACH), R-
LEACH, power-efficient gathering in sensor information system the data are collected and processed, sensor nodes send it to
(PEGASIS), and two-tier distributed fuzzy logic-based protocol the base station (BS) in a one-hop or multihop manner via
(TTDFP). wireless communication. The BS or sink node performs fur-
Index Terms—2-hop cluster, energy efficient, Internet of Things ther data processing and sends data to the client or terminal
(IoT), routing protocol, wireless sensor network (WSN). device via the Internet or satellite network.
Energy consumption has always been an important issue
of WSNs [18], which is determined by the characteristics of
Manuscript received 13 October 2021; revised 28 December 2021; accepted
19 January 2022. Date of publication 1 February 2022; date of current ver- variant WSNs. First, there are a considerable number of sensor
sion 7 October 2022. This work was supported in part by the Center for Open nodes deployed in a WSN and the monitoring environments
Intelligent Connectivity from The Featured Areas Research Center Program can be harsh and complex. Second, due to volume and cost
within the framework of the Higher Education Sprout Project by the Ministry
of Education (MOE) in Taiwan and in part by the Higher Education Sprout constraints, sensor nodes are generally fueled by equipped bat-
Project of the National Yang Ming Chiao Tung University and Ministry of teries with limited energy capacity; and considering the harsh
Education (MOE), Taiwan. The work of Li-Chun Wang was supported in part monitoring environment, battery replacement is almost impos-
by the Ministry of Science and Technology under Grant MOST 110-2634-
F-A49-006 and Grant MOST 110-2221-E-A49-039-MY3, Taiwan. The work sible. When a node dies because of energy exhaustion, there
of Chih-Min Yu was supported in part by the Natural Science Foundation will be two consequences: it directly leads to data loss of this
of Fujian Province, China, under Grant 2020J01089; in part by the Major dead node, which affects the accuracy of data collection in
Project Funding for Social Science Research of Fujian Province, China, under
Grant FJ2020MJDZ055; in part by the Fujian Key Laboratory of Spatial this area; the topology of the network changes, which affects
Information Perception and Intelligent Processing, Fuzhou, China; and in part the network stability. Furthermore, the primary objective of
by the Fujian University Engineering Research Center of Spatial Data Mining
and Application, Fuzhou, China. (Corresponding author: Chih-Min Yu.)
WSN is data collection. Although different application sce-
Chao Chen is with the Institute of Network Engineering, National narios have different requirements for network performance
Yang Ming Chiao Tung University, Hsinchu 300, Taiwan (e-mail: (QoS, security, etc.), extending the network lifetime for col-
cchen.cs07g@nctu.edu.tw).
Li-Chun Wang is with the Department of Electrical and Computer
lecting more data is a fundamental and critical requirement
Engineering, National Yang Ming Chiao Tung University, Hsinchu 300, for all WSNs. Therefore, achieving a good balance of network
Taiwan (e-mail: lichun@cc.nctu.edu.tw). energy consumption, improving energy utilization, and extend-
Chih-Min Yu is with the College of Artificial Intelligence, Yango University,
Fuzhou 350015, China (e-mail: hankycm7@gmail.com). ing the network lifetime have always been hot topics for WSN
Digital Object Identifier 10.1109/JIOT.2022.3148106 researchers.
2327-4662 
c 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
19576 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 20, 15 OCTOBER 2022

Recent research on WSNs has come up with various ideas to communication. Section VI describes the intercluster-routing
reduce energy consumption. Designing energy-efficient rout- algorithm and determines the optimal cluster number of 2-hop
ing protocols is a major challenge of these researches [18]. network configuration. In Section VII, computer simulation is
Clustering is used to divide the network into a hierarchical- conducted to demonstrate the comparisons with other methods,
based structure. Also, data acquisition and transmission power including the performances of the network lifetime, energy
control are performed in each cluster. Based on the clus- consumption, and raw packets to BS. Lastly, we present
ter structure, the network can maintain a longer lifetime by conclusion and future work in Section VIII.
scheduling the duty cycle between sensor nodes in the cluster.
Except for achieving energy-efficient in the scheduling pro-
cess, the structure can also reduce the energy consumption II. R ELATED W ORK
while transmitting data, because the member node only needs In the past two decades, a number of cluster-based and
to communicate with the cluster head (CH) node and no longer chain-based routing protocols have been explored for the pur-
needs to communicate with the farther BS. Chain structure is pose of achieving effective energy consumption in WSNs.
another approach for dividing the network. In chain-based pro- LEACH is the first work to introduce the concept of cluster-
tocols, the nodes are arranged to form a multiple chains-like ing into WSN routing and is an energy-efficient hierarchical
topology. Each chain has a chain head node that gathers data routing protocol designed for WSNs [19]. The LEACH proto-
from other nodes in the same chain, and the chain head node col randomly selects the CH nodes in a round-robin approach
finally forwards the data to the BS. and balances the network energy consumption through sensor
To achieve energy-efficient and energy-balanced WSNs, a nodes taking turns being the CH nodes. The common sensor
novel distributed 2-hop cluster-routing protocol (D2CRP) is node chooses one cluster to join based on the signal strength
proposed. With the introduction of new energy-related factors received from the CH nodes. The CH nodes act as a gateway
for CH competition and the optimal cluster number for the for the communication between common nodes of the cluster
2-hop network configuration, the energy balance for network and the BS for data transmission. The limitation of LEACH is
transmission can be attained. To the best of the authors’ knowl- that the randomly selected CH node may not be able to handle
edge, this work is the first one to deal with the 2-hop cluster the node management and data processing tasks of the entire
formation for both intracluster and intercluster communica- cluster, especially the node with low residual energy acting as
tions. The main contributions of this article are concluded as the CH node will accelerate its energy exhaustion [20].
follows. The hybrid energy-efficient distributed (HEED) clustering
1) In cluster formation, a distributed 2-hop clustering algorithm elects the CH nodes according to two designed
approach is designed to optimize the transmission dis- parameters periodically [21]. The first parameter is the resid-
tance and balance the energy consumption for CH ual energy level of each node, and the second parameter is
competition and packet transmission. the average minimum reachability power (AMRP). The HEED
2) To achieve energy-efficient transmissions in a cluster, protocol first elects candidate CH node by comparing the resid-
a distributed intracluster-routing approach is performed ual energy; if the coverage of the two candidate CH nodes
that employs 1-hop neighbor nodes as relays for trans- overlaps with each other, then the AMRP of the two candidates
mitting data in each 2-hop cluster. is compared for electing the final CH node. Common sensor
3) For energy-efficient intercluster communication from nodes will choose the CH node with the smallest AMRP within
each CH to the BS, a distributed intercluster-routing their own communication range to associate. The HEED is
approach is used for connecting the CH nodes and trans- proposed based on the LEACH, which overcomes the draw-
mitting data to the BS in a chain-based manner to reduce backs caused by the random selection of CH nodes in LEACH.
the transmission distance of each CH. The CH nodes elected with the residual energy can handle
4) To minimize the energy consumption of network trans- the task of the entire cluster without prematurely exhausting
mission, the optimal cluster number of 2-hop clus- energy. However, the HEED algorithm is highly complex, and
ters is formulated and derived. This optimal topol- too many iterations will result in a long time of cluster setup
ogy configuration achieves better performance than the and thus increase the data transmission delay [22].
other four competitive protocols, including low energy Lindsey and Raghavendra [23] proposed PEGASIS proto-
adaptive clustering hierarchy (LEACH), R-LEACH, col. This protocol uses a chain structure to connect sensor
power-efficient gathering in sensor information system nodes, which is different from the cluster structure and data
(PEGASIS), and two-tier distributed fuzzy logic-based transmission approach of LEACH. The basic idea of PEGASIS
protocol (TTDFP). is to use a greedy algorithm for node selection, making
The remainder of this article is organized as follows. locally optimal decisions, and finally forming a long chain
Section II introduces related works in WSN routing pro- structure [24]. In the communication stage, each node only
tocols. In Section III, the network model and radio model communicates with its nearest neighbor node. Except for
are examined. Section IV presents the main design crite- the chain tail node, all other nodes merge the data sent by
ria for the D2CRP, including the 2-hop cluster formation, previous-hop nodes with their own data and then transmit the
CH determination, and the transmission scheme. Section V data to their next-hop nodes. Finally, a node is selected as the
presents the intracluster routing of D2CRP by considering chain head node to transmit data to the BS. The chain head
the possible cases of energy consumption for the intracluster node is selected in a round-robin manner, which enables the

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: D2CRP: NOVEL DISTRIBUTED 2-HOP CLUSTER ROUTING PROTOCOL FOR WIRELESS SENSOR NETWORKS 19577

TABLE I
energy consumption to be evenly distributed to each node. C OMPARISON OF R ELATED ROUTING P ROTOCOLS
However, PEGASIS requires each node to obtain all other
nodes’ location information for greedy selection, which is
difficult to obtain in practice. Besides, the residual energy
is not considered. As a result, the node with lower resid-
ual energy acting as the chain head nodes will die soon like
LEACH [25], [26].
An accurate distance-based transmission scheme (ADTS) is
proposed by Xin and Liu [27]. ADTS achieves the optimum of
transmission distance for different network regions. However,
it is only for strip-based WSNs. The improved unequal-
clustering and routing (IUCR) scheme proposed in [28] uses
unequal clustering to balance the energy consumption in the
network. The clusters closer to the BS are formed in small
size to mitigate the load of the CH nodes for transmitting data.
However, IUCR is only for circular networks. The R-LEACH
protocol in [18] considers the initial energy level, current resid-
ual energy level, and an optimal number of CH nodes when
executing the CH node competition. However, the transmis-
sion distance from the BS is not considered, which may cause
the nodes that are further away from the BS to die quickly.
The improved clustering and routing (ICR) scheme proposed
by Alharbi et al. [29] treated clustering and routing as a single
unified problem to achieve energy efficiency. For electing CH
nodes, ICR takes residual energy and node degree into account,
however, it misses the importance of the distance between the
CH node and the BS as R-LEACH does. A TTDFP is proposed
by Sert et al. [30]. TTDFP contains two tiers, in the first tier,
the CH node is elected by fuzzy logic based on the residual
energy, relative connectivity, and distance-to-BS. In the second does not require centralized control overhead. Also, this pro-
tier, the optimal path from the CH nodes to the BS is found tocol should consider several critical parameters, including the
by the fuzzy logic based on the energy efficiency and distance residual energy, transmission distance, and topology configu-
to the routing path. Sert and Yazıcı [31], [32] used the mod- ration to balance the energy consumption for both intracluster
ified clonal selection algorithm (CLONALG-M) to improve and intercluster communications. To summarize the related
the performance of rule-based fuzzy routing algorithms. This works, the comparison of these routing protocols is given in
modified algorithm is utilized for finding the closest form of Table I.
the output membership functions, which can improve the over-
all performance of fuzzy routing algorithms and extend the
III. S YSTEM M ODEL
network lifetime in WSNs. Wang et al. [15] proposed the
destination-oriented routing algorithm (DORA) which divides A. Network Model
a network into a multichain structure, and derives the optimal In this article, some assumptions of the network model are
transmission distance as well as the direction to the BS. The made as follows.
network lifetime is extended with shorter chain routing. The 1) The network region is a square,
√ the network diameter is
limitation of DORA is that it is based on centralized archi- M (so the side length is M/ 2); the BS is outside the
tecture and requires global network information to achieve region and located at a distance L from one side of the
energy-efficient optimization, which is hard to deploy in real square, as shown in Fig. 1.
scenarios. 2) The sensor nodes are uniformly distributed and ran-
Although many researchers have tried to propose various domly located in the network region, the quantity of
routing protocols aiming at extending the network lifetime, nodes is n. Each sensor node is labeled with a unique
each protocol has its own advantages and limitations. When ID i (i = 1, 2, . . . , n).
the routing algorithm meets the requirements of low com- 3) The sensor nodes are static and have processing, data
plexity and practical deployment, it may consider too few fusion, communication capabilities. The communication
parameters and thus the goal of reducing energy consumption range of each sensor node is a circular area with a
is not easy to achieve; or when considering multiple param- radius of R. In the simulation, the value of R is set
eters for optimal transmission routing, the algorithm requires to 10 m.
global network information, which limits the application sce- 4) All sensor nodes have the same level of initial energy
nario of the protocol. Thus, a routing protocol is required to and know their own real-time energy level, after the node
be developed that features a fully distributed architecture and is deployed, there is no energy supply.

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
19578 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 20, 15 OCTOBER 2022

(a)

(b)

Fig. 3. Broadcast packet format. (a) Initial broadcast packet information.


(b) Updated broadcast packet information.

given in (1) and (2), respectively



kEelec + kεfs d2 , d ≤ d0
ETx (k, d) = (1)
kEelec + kεmp d4 , d > d0
ERx (k) = kEelec (2)
where Eelec denotes the per bit energy dissipation that is used
for running the transmitter and receiver electronic circuits.
εfs and εmp represent the transmitter amplifier for free-space
fading and multipath fading, respectively. d0 is the distance
Fig. 1. Considered network model.
threshold between the two fading models and the calculation
of d0 is given as follows:


εfs 10 pJ/bit/m2
d0 = = = 87 m. (3)
εmp 0.0013 pJ/bit/m4
When d ≤ d0 , it means that the transmission follows the
free-space fading model, and vice versa, adopts the multipath-
fading model.

IV. D ESIGN OF D2CRP


A. Gathering 2-Hop Neighborhood Information
To form 2-hop clusters in a distributed manner, each node
needs an approach to obtain the information of its neighbor
nodes within the 2-hop range; meanwhile, as we consider dis-
Fig. 2. Radio energy model.
tance factor and energy factor when electing CHs, we design
a broadcasting/listening scheme for the node to collect the
distance-from-BS and residual energy information of all its
5) The sensor node can measure the distance between the 1-hop neighbor nodes as well as 2-hop neighbor nodes. It is
node itself and its neighbor nodes or the BS based on assumed that each node is marked with a unique identification
the received signal strength. number, and the designed packet format is shown in Fig. 3.
6) The sensor node can adjust the transmission power in For any node i, when a new round starts, the initial broadcast
order to meet the needs of different communication packet information contains its Self ID (i), distance from the
distances. BS measured by RSSI (DBSi ), and current residual energy
7) The energy supply, processing, and storage capabilities level (REi ), as shown in Fig. 3(a). If a node i receives packets
of the BS are unlimited, the BS is able to cover the from another node j within its communication range (R), it
entire network region. means node j is a (1-hop) neighbor node of node i. Node
8) According
 to (3), the distance threshold j broadcasts its own information (Self ID: j, DBSj , REj ) as
d0 (= (εfs /εmp ) = 87 m) is much larger than node i does. After listening to the broadcast packet from node
the maximum transmission distance, so we only j, node j’s information is attached into its own packet by node
consider the free-space fading amplifier model. i, so the updated broadcast packet contains node i’s as well as
node j’s information, as shown in Fig. 3(b).
By using this approach, the packet broadcasting by node i
B. Radio Energy Model contains the information of node i and all its (1-hop) neigh-
In this article, the typical first order radio model is used [33], bors. If node i detects node k’s information from the broadcast
as shown in Fig. 2. The required energy for transmitting (ETx ) packets from node j and, meanwhile, node i does not receive
and receiving (ERx ) k bits packet over the distance of d is any packet from node k directly, it means node k is outside

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: D2CRP: NOVEL DISTRIBUTED 2-HOP CLUSTER ROUTING PROTOCOL FOR WIRELESS SENSOR NETWORKS 19579

node i’s 1-hop range but within 2-hop range (node k is a 1-hop factor ξi as shown in the following:
neighbor node of node j), and node k is a 2-hop neighbor node REi
of node i. As a result, each node can finally obtain the residual ξi =   (5)
2HopNeig
energy level as well as the distance-from-BS information of max REi
all its (1-hop) neighbor nodes and 2-hop neighbor nodes.
where REi 2HopNeig is a set of the residual energy of node i, this
set contains the residual energy of the 1-hop neighbor node and
2HopNeig
B. Distance Factor the 2-hop neighbor node of node i; max{REi } repre-
sents the maximum value in the set REi 2HopNeig , which means
After gathering the distance-from-BS and residual energy
level information of all the neighboring nodes in a 2-hop the maximum residual energy in node i’s 2-hop neighborhood.
neighborhood, the CH nodes are elected based on our designed
distance factor and energy factor. According to the radio D. Threshold Function
energy model mentioned in Section III-B, when nodes transmit By considering the distance factor and energy factor, we
data of the same amount, except for the influence of environ- improve the LEACH’s threshold function into the following:
mental factors, the energy consumption is mainly affected by ⎧
the distance between the transmitter and the receiver. As a ⎨ p
1
· ξii , i ∈ G
T(i) = 1−p r mod p (6)
result, considering the transmission distance is an important ⎩ 0, i∈/G
design factor for CH node election.
For any node i, the measured distance from the BS is DBSi . where p is the ratio of CH nodes to all the nodes. If there are
Since each node has the distance-from-BS information of all n nodes in the network, there will be np CH nodes in total,
its neighboring nodes in 2-hop range, we define a normalized and it can be concluded that the network will be divided into
distance factor i , as shown in the following: np clusters (the optimal number of 2-hop cluster is derived
in Section VI-D). r is the number of rounds that the network
DBSi runs at this time, and the elements in the set G are the nodes
i =   (4)
2HopNeig that have not successfully campaigned for CH nodes in the
max DBSi
latest rmod(1/p) round. ξi is the energy factor, and i is the
2HopNeig distance factor.
where DBSi denotes the set of the distance between
With the modified threshold function, the CH nodes are
node i and the BS, the distance between the BS and each
elected in the same way as LEACH. For each node, it decides
1-hop neighbor node of node i, as well as the distance
whether or not to become a CH node for the current round.
between the BS and each 2-hop neighbor node of node i;
2HopNeig This decision is made by the node i generating a random num-
max{DBSi } denotes the maximum value in the set of
2HopNeig ber ηi between 0 and 1. If ηi is smaller than T(i), node i
DBSi , which means the maximum distance from BS in becomes a CH node for the current round; otherwise, node i
node i’s 2-hop neighborhood. remains as a common node.
It can be observed from (6): if the distance from a sensor
C. Energy Factor node to the BS is greater, its distance factor i is greater cor-
Although the initial energy of each sensor node is the same, respondingly, and the smaller the value of T(i), the smaller
as the network data transmission processing, each node con- the probability that this node will be elected as a CH node;
sumes variant energy with different locations. When a node’s meanwhile, if the residual energy level of a node is smaller, its
energy is exhausted in advance, it will lose the ability to work energy factor ξi is smaller correspondingly, and the smaller the
and exit the network, becoming a dead node. In some cases, value of T(i), the smaller the probability that this node may
nodes closer to the BS may be assigned more communica- be elected as a CH node. In summary, the nodes farther from
tion and processing tasks than those farther from the BS. The the BS and the nodes with less residual energy level have a
nodes within the area are called hot spot nodes. Since hot spot lower probability of being elected as CH nodes. This modifi-
nodes undertake more processing and communication tasks, cation solves the shortcoming of randomly electing CH nodes
their energy consumption rate is higher than the nodes in in LEACH without considering the nodes’ energy conditions
other areas, so it may cause the nodes in the area to exit the while designing an energy-efficient routing protocol.
network prematurely due to the exhaustion of energy, caus-
ing the hot spot cannot be monitored, forming an energy hole E. Cluster Formation
problem, which reduces the quality of service in WSNs [34]. Once the CH nodes are elected for the current run-
Therefore, the election of CH nodes should take the residual ning round, they broadcast the CH election announcement
energy factor of nodes into consideration to ensure that nodes information in their communication range. For a common
can evenly bear the network task load and balance the distri- node in the communication range (1-hop neighbor node), if
bution of energy consumption. This is also an effective way it receives announcements from multiple CH nodes, it checks
for improving the lifetime of WSNs. the distance between the node itself and the CH nodes based
For any node i, its current residual energy is REi . Since each on the strength of the received broadcast signal. Then, the clos-
node has the residual energy information of all its neighboring est CH node is associated to join its cluster. After that, the CH
nodes in the 2-hop range, we define the normalized energy node broadcasts the TDMA time slot schedule to the 1-hop

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
19580 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 20, 15 OCTOBER 2022

TABLE II
N OTATIONS T HROUGH T HIS A RTICLE Algorithm 1: Cluster Formation of D2CRP
input : Node information
output: Cluster formation
forall ele ∈ SN do
broadcast(ID, DBSele , REele );
1Hop
Ni .add(1 hop neighbor node);
2Hop
Ni .add(2 hop neighbor node);
2HopNeig 1Hop 2Hop
Ni ⇐ ni ∪ Ni ∪ Ni ;
2HopNeig
forall ele ∈ Ni do
listen(DBSele , REele );
2HopNeig
DBSi .add(DBSele );
2HopNeig
REi .add(REele );
2HopNeig
i ⇐ DBSi / max{DBSi };
2HopNeig
ξi ⇐ REi / max{REi };
if ni ∈ G then
T(i) ⇐ ξi /i ∗ p/(1 − p(r mod 1/p);
G.delete(ni );
else
T(i) ⇐ 0;
if ni s random number < T(i) then
CH.add(ni );
broadcast(CH announcement);
else
listen(CH announcement);
neighbor nodes, the 1-hop neighbor nodes transmit data to the 1Hop 1Hop
if ele ∈ Ni \ (Ni ∩ CH) then
CH node in different time slots for the purpose of avoiding join the closest cluster;
collisions. relay broadcast (CH announcement);
Once the 1-hop neighbor nodes become cluster member 2Hop 2Hop
else if ele ∈ Ni \ (Ni ∩ CH) then
nodes, they also broadcast the information of the CH node join the cluster which the closest 1 hop neighbor node joins;
and themselves in their communication range. When a 2-hop start data transmission
neighbor node (of the CH node) receives announcements from r ⇐r+1
multiple 1-hop neighbor nodes, it checks the distance between
the node itself and the 1-hop neighbor nodes based on the
strength of the received broadcast signal and then chooses the
maximum number of iterations is max , the total commu-
closest 1-hop neighbor node to join the cluster. Then, the 1-hop
nication overhead of the clustering algorithm in D2CRP is
neighbor node broadcasts the TDMA time slot schedule to the
O(mn2 max ) [35], [36].
2-hop neighbor nodes, then the 2-hop neighbor nodes transmit
data to the 1-hop neighbor node in different time slots to avoid
collisions. As a result, the transmission distance for all nodes V. I NTRACLUSTER ROUTING
including the CH node in a cluster can be effectively reduced. Based on the considered network model and 2-hop cluster-
ing, the network is divided into multiple circular clusters, as
shown in Fig. 4. Each cluster contains one CH node, multiple
F. Clustering Algorithm 1-hop neighbor nodes, and multiple 2-hop neighbor nodes. For
each 2-hop neighbor node, it chooses the closest (based on the
We summarize the designed clustering algorithm in distance measured by RSSI) 1-hop neighbor node as the next
Algorithm 1. To describe the algorithm, some variables are forwarder and transmits the collected data to the next-hop node
defined beforehand in Table II. in its own time slot. The chosen 1-hop neighbor node acts as
As there are n nodes in the network, during the process a relay between the 2-hop neighbor node and the CH node, it
of CH node election, each node needs to broadcast its own merges the data received from the 2-hop neighbor node with
information. As a result, O(n) packets are broadcast. Because its own collected data and then transmits the fused data to the
m CH nodes are elected, for each CH node, it needs to broad- CH node in its own time slot. Through the relaying mecha-
cast the CH announcement information. As a result, O(m) nism, the data of all the intracluster 1-hop neighbor nodes and
packets are broadcast, then the non-CH nodes need to compare 2-hop neighbor nodes are aggregated to the CH node.
the distance after receiving the CH announcement information,
the communication overhead of each iteration of the cluster-
ing algorithm is O(mn). In addition, each node also broadcasts A. Node Distribution
the information of its 1-hop neighbor nodes after receiving Assume the number of 2-hop clusters is m, for uniformly
the information from its neighbor nodes. The communica- distributed clusters, each cluster contains (n/m) nodes. As each
tion overhead of this process is O(n). Assuming that the cluster only has one CH node, the total number of the 1-hop

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: D2CRP: NOVEL DISTRIBUTED 2-HOP CLUSTER ROUTING PROTOCOL FOR WIRELESS SENSOR NETWORKS 19581

C. Energy Consumption of Intracluster Transmission


In this article, the main power effect is concerned for the
energy consumed by the communication module in each sen-
sor. Based on the radio energy model [(1) and (2)], the energy
consumed by the communication module mainly depends on
the transmission distance and the amount of data. In each
2-hop cluster, the energy consumption of communication can
be generally divided into four cases. We assume the max-
imum transmission distance is no more than d0 (3), when
transmitting/receiving k bits data.
1) Case 1: The energy consumption of transmitting data to
the 1-hop neighbor nodes for the 2-hop neighbor nodes,
is given as follows:
E2HopTx1Hop = n2Hop · ETx (k, d)
Fig. 4. Intracluster routing.  
3 n M2
= − 1 k Eelec + εfs . (12)
4 m 8π m
neighbor nodes as well as the 2-hop neighbor nodes in each 2) Case 2: The energy consumption of receiving data from
cluster is (n/m) − 1. the 2-hop neighbor nodes for the 1-hop neighbor nodes,
According to the proportion of occupied area, the number is given as follows:
of all the 1-hop neighbor nodes is shown as follows:
E1HopRx2Hop = n1Hop · 3 · ERx (k)
π R2 n 3 n
n1Hop = · −1 = − 1 kEelec . (13)
π(2R)2 m 4 m
1 n 3) Case 3: The energy consumption of transmitting fused
= −1 . (7)
4 m data to CH node for 1-hop neighbor nodes, is given as
follows:
The number of all the 2-hop neighbor nodes is shown as
follows: E1HopTxCH = n1Hop · ETx (k, d)
 
n 1 n M2
n2Hop = − 1 − n1Hop = − 1 k Eelec + εfs . (14)
m 4 m 8π m
3 n
= −1 . (8) 4) Case 4: The energy consumption of receiving fused data
4 m from 1-hop neighbor nodes for CH node, is given as
The average number of the 2-hop neighbor nodes that each follows:
1-hop neighbor node is responsible for relay broadcasting and
ECHRx1Hop = n1Hop · ERx (k)
the value is 3 (n2Hop /n1Hop ). 1 n
= − 1 kEelec . (15)
4 m
B. Intracluster Transmission Distance The energy consumption of transmission in each 2-hop
Based on the approach in [37], we first assume a 2-hop clus- cluster is given as follows:
ter as a larger 1-hop cluster. From the perspective of the whole
EIntra = E2HopTx1Hop + E1HopRx2Hop + E1HopTxCH
network, the expectation of distance square between the CH
node and the 2-hop neighbor nodes is given as follows [37]: + ECHRx1Hop . (16)
 M2
E d2 = . (9) VI. I NTERCLUSTER ROUTING
2π m After gathering data from the associated cluster member
Then, we consider the 2-hop cluster as an observation area, nodes, each CH node merges the data and then transmits the
the expectation of distance square between the CH node and fused data to the BS in some way. In LEACH-like routing
the 1-hop neighbor nodes is given as follows: protocols, all the CH nodes transmit the data directly to the
BS. This single-hop manner may consume considerable energy
 M2
E d2 = . (10) if the CH node is far away from the BS.
8π m Aiming at improving the energy efficiency of transmitting
As a result, the expectation of distance square between the data to the BS, we propose a distributed 2-hop intercluster-
1-hop neighbor nodes and the 2-hop neighbor nodes is given routing algorithm to interconnect CH nodes based on the ideas
as follows: of the PEGASIS protocol. We assume that within each CH
node’s 2-hop range, there are other CH nodes. Through the
 M2
E (d)2 = . (11) received packets from the nodes in the 2-hop range, each sen-
8π m sor node has the knowledge of the distance-to-BS information

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
19582 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 20, 15 OCTOBER 2022

(a) (b)

Fig. 6. Distribution and transmission distance of CH nodes. (a) CH sends


data to BS directly. (b) CH sends data to BS via three possible forwarders.


Thus, the expectation of distance square for the m CH
Fig. 5. Intercluster routing. nodes is shown in the following:

E d 2 = L2 . (18)
(DBS) of each of its 1-hop neighbor nodes as well as each √
of its 2-hop neighbor nodes. For each CH node, it decides its For (m − m) CH √ nodes that are not in the row closest to
next hop based on the DBS information of other CH nodes the BS (2 ≤ RN ≤ m), if a CH node transmits data to a CH
in the 2-hop range: if another CH node is closer to the BS forwarder that are closer to the BS, there are three possible
(has a smaller value of DBS) when compared with itself, the situations, including the upward CH directly, left upward CH,
CH node uses the closer (to the BS) CH node as its next for- and right upward CH as shown in Fig. 6(b), then the expected
warder; otherwise, it sends the data to the BS directly. As a transmission distance from the CH node itself to the next-hop
result, multiple routing chains can be constructed from the far- CH node is as follows:

thest CH nodes to the BS, as shown in Fig. 5. Based on this 1 M 1 M/ 2 1 M
mechanism, for gathering data of the network in each round, E[d] = · √ + · √ + ·√
3 m 3 m 3 m
each elected CH node receives data from its 1-hop neighbor √
nodes and 2-hop neighbor nodes (by relaying) in the clus- 4+ 2 M
ter, fuses with its own data, transmits to the next hop (either = √ . (19)
6 m
another CH node or the BS) on the chain, and the data is
finally aggregated to the BS (terminal point of the chain). Therefore,
√ the expectation of distance square for the
(m − m) CH nodes is shown in the following:
A. CH Node Distribution √
 9 + 4 2 M2
Based on the assumption that nodes are uniformly dis- E d2 = . (20)
tributed and randomly located. For simplifying the derivation, 18m
here we assume that the distribution of the elected CH nodes
is completely uniform. As the number of clusters C. Energy Consumption of Intercluster Transmission
√ is√m, the
square WSN monitoring region is divided into m · m for For the energy consumption of interclusters data transmis-
small square regions, as shown in Fig. 6. From sion, because the BS has unlimited energy supply, the energy
√ the perspec-
tive of BS, we√divide these small squares into m rows, each consumption of the CH nodes is focused. The energy con-
row contains m√CH nodes and is marked with a row number sumption of the CH nodes can be concluded into three cases,
(RN, 1 ≤ RN ≤ m). The √ average
√ gap between two adjacent
and the maximum transmission distance is assumed no more
rows of CH nodes is ([M/ 2]/ m). than d0 (3), when transmitting/receiving k bits data.
1) Case 1: The energy consumption
√ of transmitting fused
B. Intercluster Transmission Distance data to the BS for all the m CH nodes in the row
√ closest to the BS (RN = 1), is given as follows:
For the m CH nodes in the row closest to the BS
√  
(RN = 1), they transmit data to the BS directly, as shown in ECHTxBS = m · ETx k, d
Fig. 6(a). Because the value of L is relatively large, we approx- √
imate the expectation of the transmission distance between = mk Eelec + εfs L2 . (21)
these nodes and the BS as L, as follows:
  2) Case 2: The energy consumption of transmitting√ fused
E d = L. (17) data to the next-hop CH nodes for all the (m − m) CH

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: D2CRP: NOVEL DISTRIBUTED 2-HOP CLUSTER ROUTING PROTOCOL FOR WIRELESS SENSOR NETWORKS 19583


nodes not in the row closest to the BS (2 ≤ RN ≤ m), Algorithm 2: Intracluster and Intercluster Routing of
is given as follows: D2CRP
 √  input : Cluster formation
ECHTxCH = m − m · ETx (k, d) output: Routing chain
⎛ √ ⎞ 2Hop 2Hop
 √  9 + 4 2 M2 forall ele ∈ Ni \ (Ni ∩ CH) do
= m − m k⎝Eelec + εfs ⎠. 1Hop
transmit data to the closest ele ∈ Ni
1Hop
\ (Ni ∩ CH);
18m
1Hop 1Hop
forall ele ∈ Ni \ (Ni ∩ CH) do
(22) receive data from ele ∈
2Hop 2Hop
Ni \ (Ni ∩ CH) it is responsible for;
3) Case 3: The energy consumption of receiving fused√
data fuse data;
from the previous-hop CH nodes for all the (m − m) transmit data to the closest ele ∈ CH;
√ nodes not in the row furthest to the BS (1 ≤ RN ≤
CH forall ele ∈ CH do
m − 1), is given as follows: receive data from cluster member nodes;
 √  receive data from previous-hop ele ∈ CH;
ECHRxCH = m − m · ERx (k) fuse data;
 √  if ele ∈ Ni
2HopNeig
∪ CH & DBSele
2HopNeig 2HopNeig
< DBSi
= m − m kEelec . (23)
then
transmit data to another ele ∈
The energy consumed for intercluster data transmission to CH closer to the BS in 2-hop range;
the BS over the whole network is given as follows: else
transmit directly to the BS;
EInter = ECHTxBS + ECHTxCH + ECHRxCH . (24)

D. Optimal Number of 2-Hop Clusters


According to (16) and (24), the total energy consumption E. Routing Algorithm
by m 2-hop clusters over the whole network is given below After the multiple chains are generated, the intercluster com-
munication for each CH node is achievable and the designed
ETotal = m · EIntra + EInter intracluster and intercluster-routing algorithm is presented in
 
M2 Algorithm 2.
= (n − m)k 2Eelec + εfs
8π m Here, the communication overhead for routing chain forma-
√ tion is presented as below. It is assumed that the maximum
+ mk Eelec + εfs L 2
number of hops for the elected CH nodes to transmit the col-
⎛ √ ⎞
9 + 4 2 M2 lected data to the BS is max , each CH node needs to compare
 √ 
+ m − m k⎝2Eelec + εfs ⎠. (25) the distances from other CH nodes to the BS to determine its
18m next forwarder. As a result, the communication overhead of
forming the transmission chain from the m CH nodes to the
It can be observed from (25) that when the number of all BS is O(m max ). Combined with the communication overhead
nodes n, the network diameter M, and the distance L from of the clustering algorithm, the total communication overhead
the BS to the monitoring area are all preset constants, the of the designed routing protocol is O(mn2 max + m max ).
total energy consumption is mainly affected by the number
of clusters m. Referring to the method used in [15], [27], and
[37], we let the first derivation of (25) equal to zero in order to VII. P ERFORMANCE E VALUATION
determine the optimal number of 2-hop clusters that minimizes In order to evaluate the proposed D2CRP, the MATLAB
the energy consumption of the network R2021a simulation tool is used to simulate the LEACH,
R-LEACH, PEGASIS, TTDFP, and the proposed D2CRP
∂ETotal
= 0. (26) protocols. Then, the experimental results are compared and
∂m analyzed from the perspective of network lifetime, network
The optimal number of 2-hop clusters m in D2CRP is given energy consumption, and raw packets to BS.
as follows: We choose the four protocols for comparison due to their
  high relevance to D2CRP. For CH election, D2CRP improves
∂ETotal
mOptimal
= m|m ∈ N, =0 . (27) LEACH’s threshold function by introducing the energy factor
∂m ξi and distance factor i ; for forming the clusters, LEACH’s
As there are n nodes in the network, the optimal probability 1-hop clusters are improved into 2-hop clusters. For inter-
of being elected as CH nodes in the network is (mOptimal /n), cluster transmission, D2CRP forms the transmission chains
thus the optimal value of p in the threshold function (6) is as PEGASIS does, but the chains are between the CHs and
redefined as follows: the BS (rather than the common nodes) and formed in a dis-
tributed manner (rather than the centralized manner). For CH
mOptimal election factors, R-LEACH considers the residual energy but
pOptimal = . (28)
n misses the distance-to-BS which also matters, we introduce

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
19584 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 20, 15 OCTOBER 2022

Fig. 7. Simulation scenario.

TABLE III Fig. 8. Finding the optimal number of 2-hop clusters.


S IMULATION PARAMETER

both the two factors in D2CRP. TTDFP takes residual energy,


relative connectivity, and distance-to-BS into account, but it
uses 1-hop clusters.
Fig. 9. Comparison of network lifetime.
A. Environment and Parameter Setting
The simulation scenario is that 100 sensor nodes are
deployed in a 100 m × 100 m square region, following uni- the number of clusters is too great, it means the distribution
form distribution and random location, and the BS is deployed of CH nodes is too dense, the CH node receives data from
50 m outside the region with coordinates (50 m, 150 m), fewer member node which leads to a lower fusion degree of
as shown in Fig. 7. The detailed parameter settings of this intercluster data aggregation, as a result, the network energy
simulation scenario are shown in Table III. consumption is higher. This curve reaches the minimum value
when the number of clusters is equal to 7. Therefore, in this
experimental scenario, the optimal number of 2-hop clusters
B. Finding the Optimal Number of 2-Hop Clusters is 7, which means the probability of the nodes to be elected
According to the derived energy consumption model (25), as the CH node is 0.07.
the relationship between the total network energy consumption
per-bit and the number of 2-hop clusters in the simulation
scenario is shown in Fig. 8. C. Comparison of Network Lifetime
It can be observed from Fig. 8 that as the number of 2-hop To demonstrate that the D2CRP protocol can effectively
clusters increases, the network energy consumption per-bit slow down the death rate of nodes and extend the network
decreases first and then increases. The reason is that when the lifetime, we show the relationship between the number of alive
number of clusters is too small, it means the distribution of nodes and the number of running rounds of the five protocols
CH nodes is too sparse and the member nodes need to transmit in Fig. 9. Additionally, the simulation results of the round of
the collected data very far to reach the CH node, resulting in first node death (FND), half node death (HND), and all node
higher network energy consumption; on the other hand, when death (AND) are presented in Table IV.

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: D2CRP: NOVEL DISTRIBUTED 2-HOP CLUSTER ROUTING PROTOCOL FOR WIRELESS SENSOR NETWORKS 19585

TABLE IV
C OMPARISON OF FND, HND, AND

It can be seen from Fig. 9 that the LEACH protocol is


the first to have the death nodes, followed by R-LEACH,
TTDFP, PEGASIS, and the proposed D2CRP. As the num-
ber of running rounds increases, the downward trend of the
D2CRP curve is relatively gentle, and the curves of LEACH,
R-LEACH, PEGASIS, and TTDFP are relatively steep. From
Table IV, we can observe that the overall network lifetime
(the round of AND) under D2CRP is approximately twice
that under LEACH, is about 38% longer than that under R- Fig. 10. Comparison of network energy consumption.
LEACH, is about 23% longer than that under PEGASIS, and
TABLE V
is about 2% longer than that under TTDFP. It is obvious that C OMPARISON OF AVERAGE N ETWORK E NERGY C ONSUMPTION R ATE
D2CRP has the best results in terms of these three indicators
since it has the lowest node death rate and thus the longest
network lifetime.
We analyze that the slow node death rate of D2CRP is
mainly due to the introduction of a new factor by taking into
account both the impacts of the residual energy and trans-
mission distance for the CH competition process. Because the
energy consumed by the CH nodes is more than that of mem-
ber nodes, the residual energy of those nodes that have been We analyze that the slow energy consumption rate of the
elected as CH nodes is usually lower after certain rounds. At network under the D2CRP protocol is mainly due to the use
this moment, the residual energy factor is becoming smaller, of 2-hop clustering and routing, and the introduction of the
and the probability of being elected as a CH node is rel- distance factor in the CH election process. For member nodes
atively smaller. On the other hand, those nodes with high (especially for 2-hop neighbor nodes), the use of node relay-
residual energy are more likely to be elected as CH nodes. ing to transmit data to the CH node within the cluster reduces
This approach achieves a well-balanced energy consumption the transmission distance to a certain extent, thereby reducing
distribution between the participating nodes, thus extending the energy consumed for transmitting data to the CH nodes.
the lifetime of each node. For CH nodes, because there are 1-hop neighbor nodes in
the cluster to help them receive and merge the data of 2-hop
D. Comparison of Network Energy Consumption neighbor nodes, the total amount of data received by CH nodes
can be effectively mitigated, thereby reducing the energy con-
The network energy consumption is used to measure the
sumed for receiving data from member nodes; meanwhile, as
performance of the routing protocols. Since the energy supply
the relaying mechanism similar to PEGASIS is adapted for
of the nodes is limited, the protocol performs better if less
intercluster communication, compared with the approach of
energy is consumed in the network after the same rounds. The
directly transmitting data to the BS, the transmission distance
network energy consumption comparison of the five protocols
between the BS and the CH nodes is also reduced to a cer-
is shown in Fig. 10.
tain extent, thus effectively improving the energy utilization
It can be observed from Fig. 10 that the LEACH protocol
for transmitting data to the BS.
is the first that completely exhausts the energy, following R-
LEACH, PEGASIS, TTDFP, and the proposed D2CRP. The
LEACH protocol consumes the most energy after the same E. Comparison of Raw Packets to BS
rounds, and the growth rate of its curve is the fastest. For The number of raw packets (before data fusion) sent to the
our proposed D2CRP protocol, the growth rate of its curve is BS can well represent the transmission efficiency of the WSN
the slowest. Although there are slight fluctuations in the mid- routing protocol. After the same time, the more raw pack-
dle rounds, the overall energy consumption curve is relatively ets sent to the BS, the more monitoring information collected
stable and slow. Table V shows the average network energy by the BS, and the higher the transmission efficiency of the
consumption rate of the five protocols. The rate of the D2CRP network. The comparison of raw packets to BS of the five
is about 52% slower than LEACH, is about 27% slower than protocols is shown in Fig. 11 and Table VI.
R-LEACH, is about 18% slower than PEGASIS, and is about It can be seen from Fig. 11 that at the beginning, as no
2% slower than TTDFP. node dies, the five protocols send the same raw packets to

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
19586 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 20, 15 OCTOBER 2022

Fig. 12. Scalability Study—Comparison of network lifetime.


Fig. 11. Comparison of raw packets to BS.

TABLE VII
TABLE VI S CALABILITY S TUDY—C OMPARISON OF FND, HND, AND
C OMPARISON OF T OTAL R AW PACKETS TO BS

that of TTDFP as well as D2CRP (1-hop variant of D2CRP).


The results are shown in Fig. 12 and Table VII.
the BS during the same operation rounds. At around 150 From the simulation results, we can note that D2CRP still
rounds, the difference begins to appear, and after 200 rounds, has a relatively lower node death rate in such a large-scale
the difference becomes significant. During the same time, the network. The 1-hop variant D1CRP has a better performance
amount of raw packets sent to the BS under LEACH is the than TTDFP in terms of FND, but later the death rate becomes
least, followed by R-LEACH, PEGASIS, TTDFP, and D2CRP. higher and dies out earlier than the other two protocols. The
The amount of raw packets under the LEACH protocol stops overall network lifetime (AND) of D2CRP is about 37%
growing around 430 rounds, while for R-LEACH, PEGASIS, longer than that of D1CRP and is about 4% longer than that
and TTDFP, the number is around 650, 740, and 890, respec- of TTDFP.
tively, while D2CRP stops growing until around 900 rounds.
According to the results in Table VI, in the entire lifetime of
WSN, D2CRP has the largest amount of raw packets sent to VIII. C ONCLUSION AND F UTURE W ORK
the BS and has increased by approximately 115%, 60%, 33%, In this article, a distributed 2-hop routing algorithm called
and 16% when compared to LEACH, R-LEACH, PEGASIS, D2CRP is proposed for energy-efficient WSNs. In the clus-
and TTDFP, respectively. ter formation, the transmission distance and residual energy
We analyze this mainly because the D2CRP protocol can are jointly considered to determine the CH node in each
well balance the energy consumption between the nodes, mak- 2-hop cluster. For intracluster and intercluster routing, the
ing the rate of death node slower, so that more nodes that transmission distance and the amount of transmitted data are
survive at the same time compared to the other four protocols, both reduced by using one-hop relay and CH chains. Each
thus the more raw packets are sent accordingly. node only needs to acquire the node information within the
2-hop range rather than the globe information. Thus, the
energy consumption is optimized from the perspective of the
F. Scalability Study network layer. Experimental results indicate that compared
For evaluating the performance of D2CRP in large-scale with the LEACH, R-LEACH, PEGASIS, and TTDFP pro-
scenarios, a scalability study is carried out. We change the tocols, the proposed D2CRP effectively reduces the rate of
network region to a 1000 m × 1000 m square, and the coor- network energy consumption, increases the lifetime of the
dinator of BS to (500 m, 1050 m). Besides, the number of network, achieves a more balanced distribution of energy con-
deployed nodes is changed to 1000. For other parameter con- sumption, and is capable to send more raw packets to BS.
figurations, it is the same as Table III. Under this scenario, the Additionally, the results show that D2CRP has good scalability
network lifetime of D2CRP is investigated and compared with for large-scale WSNs.

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
CHEN et al.: D2CRP: NOVEL DISTRIBUTED 2-HOP CLUSTER ROUTING PROTOCOL FOR WIRELESS SENSOR NETWORKS 19587

For future work, the proposed D2CRP can be evaluated [21] O. Younis and S. Fahmy, “HEED: A hybrid, energy-efficient, distributed
under scenarios with mobile nodes. Additionally, implement- clustering approach for ad hoc sensor networks,” IEEE Trans. Mobile
Comput., vol. 3, no. 4, pp. 366–379, Oct.–Dec. 2004.
ing the routing protocol based on wide used standards, such [22] Z. Ullah, “A survey on hybrid, energy efficient and distributed (HEED)
as ZigBee and Bluetooth should be investigated for practical Based energy efficient clustering protocols for wireless sensor networks,”
applications. Wireless Personal Commun., vol. 112, pp. 2685–2713 , Jan. 2020.
[23] S. Lindsey and C. S. Raghavendra, “PEGASIS: Power-efficient gathering
in sensor information systems,” in Proc. IEEE Aerosp. Conf., vol. 3,
R EFERENCES 2002, pp. 1–3.
[24] J. Kim, Y. Kim, and Y. Yoo, “Energy-efficient routing using genetic
[1] N. Pathak, P. K. Deb, A. Mukherjee, and S. Misra, “IoT-to-the-rescue: A algorithm in cluster-based wireless sensor networks,” in Proc. IEEE 90th
survey of IoT solutions for COVID-19-like pandemics,” IEEE Internet Veh. Technol. Conf. (VTC-Fall), 2019, pp. 1–5.
Things J., vol. 8, no. 17, pp. 13145–13164, Sep. 2021. [25] L. Chan, K. G. Chavez, H. Rudolph, and A. Hourani, “Hierarchical
[2] M. N. Bhuiyan, M. M. Rahman, M. M. Billah, and D. Saha, “Internet of routing protocols for wireless sensor network: A compressive survey,”
Things (IoT): A review of its enabling technologies in healthcare appli- Wireless Netw., vol. 26, no. 5, pp. 3291–3314, 2020.
cations, standards protocols, security and market opportunities,” IEEE [26] A. M. Khedr, A. Aziz, and W. Osamy, “Successors of PEGASIS pro-
Internet Things J., vol. 8, no. 13, pp. 10474–10498, Jul. 2021. tocol: A comprehensive survey,” Comput. Sci. Rev., vol. 39, Feb. 2021,
[3] K. S. Awaisi, S. Hussain, M. Ahmed, A. A. Khan, and G. Ahmed, Art. no. 100368.
“Leveraging IoT and fog computing in healthcare systems,” IEEE [27] H. Xin and X. Liu, “Energy-balanced transmission with accurate dis-
Internet Things Mag., vol. 3, no. 2, pp. 52–56, Jun. 2020. tances for strip-based wireless sensor networks,” IEEE Access, vol. 5,
[4] J. Shah and B. Mishra, “IoT-enabled low power environment monitoring pp. 16193–16204, 2017.
system for prediction of PM2.5,” Pervasive Mobile Comput., vol. 67, [28] M. A. Alharbi and M. Kolberg, “Improved unequal-clustering and rout-
Sep. 2020, Art. no. 101175. ing protocol,” IEEE Sensors J., vol. 21, no. 20, pp. 23711–23721,
[5] M. Dryjanski, M. Buczkowski, Y. Ould-Cheikh-Mouhamedou, and Oct. 2021.
A. Kliks, “Adoption of smart cities with a practical smart building [29] M. A. Alharbi, M. Kolberg, and M. Zeeshan, “Towards improved
implementation,” IEEE Internet Things Mag., vol. 3, no. 1, pp. 58–63, clustering and routing protocol for wireless sensor networks,”
Mar. 2020. EURASIP J. Wireless Commun. Netw., vol. 46, pp. 1–31, 2021,
[6] Y. Cao, C. Chen, D. St-Onge, and G. Beltrame, “Distributed TDMA doi: 10.1186/s13638-021-01911-9.
for mobile UWB network localization,” IEEE Internet Things J., vol. 8, [30] S. A. Sert, A. Alchihabi, and A. Yazici, “A two-tier distributed
no. 17, pp. 13449–13464, Sep. 2021. fuzzy logic based protocol for efficient data aggregation in multihop
[7] B. Bera, A. K. Das, S. Garg, M. J. Piran, and M. S. Hossain, wireless sensor networks,” IEEE Trans. Fuzzy Syst., vol. 26, no. 6,
“Access control protocol for battlefield surveillance in drone-assisted pp. 3615–3629, Dec. 2018.
IoT environment,” IEEE Internet Things J., early access, Jan. 4, 2021, [31] S. A. Sert and A. Yazıcı, “Optimizing the performance of rule-based
doi: 10.1109/JIOT.2020.3049003. fuzzy routing algorithms in wireless sensor networks,” in Proc. IEEE
[8] H. Peng, C. Chen, C.-C. Lai, L.-C. Wang, and Z. Han, “A predictive Int. Conf. Fuzzy Syst. (FUZZ-IEEE), 2019, pp. 1–6.
on-demand placement of UAV base stations using echo state network,” [32] S. A. Sert and A. Yazici, “Increasing energy efficiency of rule-based
in Proc. IEEE/CIC Int. Conf. Commun. China (ICCC), 2019, pp. 36–41. fuzzy clustering algorithms using CLONALG-M for wireless sensor
[9] “Gartner Says 5.8 Billion Enterprise and Automotive IoT Endpoints networks,” Appl. Soft Comput., vol. 109, Sep. 2021, Art. no. 107510.
Will Be in Use in 2020.” Gartner. [Online]. Available: https://www. [33] W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrishnan, “An
gartner.com/en/newsroom/press-releases/2019-08-29-gartner-says-5-8- application-specific protocol architecture for wireless microsensor
billion-enterprise-and-automotive-io (accessed Jun. 15, 2021). networks,” IEEE Trans. Wireless Commun., vol. 1, no. 4, pp. 660–670,
[10] “Insights on the Internet of Things,” McKinsey. [Online]. Available: Oct. 2002.
https://www.mckinsey.com/featured-insights/internet-of-things/our- [34] J. Ren, Y. Zhang, K. Zhang, A. Liu, J. Chen, and X. S. Shen, “Lifetime
insights (accessed May 3, 2021). and energy hole evolution analysis in data-gathering wireless sensor
[11] V. Gupta and S. De, “An energy-efficient edge computing framework networks,” IEEE Trans. Ind. Informat., vol. 12, no. 2, pp. 788–800,
for decentralized sensing in WSN-assisted IoT,” IEEE Trans. Wireless Apr. 2016.
Commun., vol. 20, no. 8, pp. 4811–4827, Aug. 2021. [35] C. Xu, Z. Xiong, G. Zhao, and S. Yu, “An energy-efficient region source
[12] C. Gong, F. Lin, X. Gong, and Y. Lu, “Intelligent Cooperative Edge routing protocol for lifetime maximization in WSN,” IEEE Access,
Computing in Internet of Things,” IEEE Internet Things J., vol. 7, no. 10, vol. 7, pp. 135277–135289, 2019.
pp. 9372–9382, Oct. 2020. [36] C.-M. Yu, M.-L. Ku, and L.-C. Wang, “BMRHTA: Balanced multipath
[13] F. Guo, F. R. Yu, H. Zhang, X. Li, H. Ji, and V. C. Leung, “Enabling routing and hybrid transmission approach for lifecycle maximization in
massive IoT toward 6G: A comprehensive survey,” IEEE Internet Things WSNs,” IEEE Internet Things J., vol. 9, no. 1, pp. 728–742, Jan. 2022.
J., vol. 8, no. 15, pp. 11891–11915, Aug. 2021. [37] S. Hussain and A. W. Matin, “Energy efficient hierarchical cluster-
[14] U. M. Malik, M. A. Javed, S. Zeadally, and S. Ul Islam, “Energy efficient based routing for wireless sensor networks,” Jodrey School Comput.
fog computing for 6G enabled massive IoT: Recent trends and future Sci., Acadia Univ. Wolfville, Wolfville, NS, Canada, Rep. TR-2005-011,
opportunities,” IEEE Internet Things J., early access, Mar. 22, 2021, 2005.
doi: 10.1109/JIOT.2021.3068056.
[15] K. Wang, C.-M. Yu, and L.-C. Wang, “DORA: A destination oriented
routing algorithm for energy-balanced wireless sensor networks,” IEEE
Internet Things J., vol. 8, no. 3, pp. 2080–2081, Feb. 2021.
[16] R. Krishnamurthi, A. Kumar, D. Gopinathan, A. Nayyar, and B. Qureshi,
“An overview of IoT sensor data processing, fusion, and analysis
techniques,” Sensors, vol. 20, no. 21, p. 6076, 2020.
[17] B. Montrucchio, E. Giusto, M. G. Vakili, S. Quer, R. Ferrero, and
C. Fornaro, “A densely-deployed, high sampling rate, open-source air
pollution monitoring WSN,” IEEE Trans. Veh. Technol., vol. 69, no. 12, Chao Chen received the B.E. degree from the
pp. 15786–15799, Dec. 2020. School of Optical and Electronic Information,
[18] T. M. Behera, S. K. Mohapatra, U. C. Samal, M. S. Khan, Huazhong University of Science and Technology,
M. Daneshmand, and A. H. Gandomi, “Residual energy-based cluster- Wuhan, China, in 2017, and the M.S. degree from
head selection in WSNs for IoT application,” IEEE Internet Things J., the Institute of Network Engineering, National Yang
vol. 6, no. 3, pp. 5132–5139, Jun. 2019. Ming Chiao Tung University (formerly, National
[19] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy- Chiao Tung University), Hsinchu, Taiwan, in 2021.
efficient communication protocol for wireless microsensor networks,” From January 2020 to June 2020, he was
in Proc. 33rd Annu. Hawaii Int. Conf. Syst. Sci., vol. 2, 2000, p. 10. a Research Intern with MIST Lab, École
[20] I. Daanoune, B. Abdennaceur, and A. Ballouk, “A comprehensive sur- Polytechnique de Montréal, Montreal, QC, Canada.
vey on LEACH-based clustering routing protocols in wireless sensor His research interests include networked systems,
networks,” Ad Hoc Netw., vol. 114, Apr. 2021, Art. no. 102409. wireless sensor networks, and software-defined networks.

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.
19588 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 20, 15 OCTOBER 2022

Li-Chun Wang (Fellow, IEEE) received the Ph.D. Chih-Min Yu (Senior Member, IEEE) received
degree from the Georgia Institute of Technology, the B.S. degree in electrical engineering from the
Atlanta, GA, USA, in 1996. Chung-Cheng Institute of Technology, Taoyuan City,
From 1996 to 2000, he was with AT&T Taiwan, in 1989, and the M.S. and Ph.D. degrees in
Laboratories, Florham Park, NJ, USA, where he was telecommunication engineering from National Chiao
a Senior Technical Staff Member with the Wireless Tung University, Hsinchu, Taiwan, in 1997 and
Communications Research Department. Since 2000, 2007, respectively.
he has been with the Department of Electrical He has been a Full Professor with the College
and Computer Engineering of National Chiao Tung of Artificial Intelligence, Yango University, Fuzhou,
University, Hsinchu, Taiwan, where he is currently China, since March 2019. In 2020, he was a
co-appointed with the Department of Computer Distinguished Professor of Minjiang Scholars in
Science and Information Engineering. He holds 26 U.S. patents, and has Fujian Province. From 1989 to 2007, he was a Researcher with the
published over 100 journal papers and 200 conference papers, and co- Chung-Shan Institute of Science and Technology, Taoyuan City, Taiwan.
edited a book, Key Technologies for 5G Wireless Systems, (Cambridge From 2008 to 2009, he was an Assistant Professor with the Department of
University Press, 2017). His current research interests include software- Information and Communication Engineering, Asia University, Taichung City,
defined mobile networks, heterogeneous networks, and data-driven intelligent Taiwan. In August 2009, he joined the Faculty of Chung-Hua University,
wireless communications. Hsinchu, Taiwan, where was a Full Professor with the Department of
Dr. Wang was a recipient of the Distinguished Research Award of the Electronics Engineering in 2018. His research interests include mobile com-
Ministry of Science and Technology, Taiwan, in 2012 and 2017, and a munication, ad hoc networks, wireless sensor networks, and Internet of Things
co-recipient of the IEEE Communications Society Asia–Pacific Board Best networks.
Award in 2015, the Y. Z. Hsu Scientific Paper Award in 2013, and the IEEE Dr. Yu was a recipient of the University-Level Outstanding Teaching Award
Jack Neubauer Best Paper Award in 1997. in 2014 and 2016, as well as the Outstanding Research Award in 2017 and
2019 at Chung-Hua University, and the Outstanding Talent Award of the
Ministry of Science and Technology, Taiwan, in 2018; the First Prize of
the Natural Science Outstanding Academic Paper Award, Fuzhou City, and
the Third Prize of the Natural Science Outstanding Academic Paper Award,
Fujian Providence, China, in 2020. He is currently serving as an Associate
Editor for IEEE ACCESS.

Authorized licensed use limited to: ASTON UNIVERSITY. Downloaded on February 29,2024 at 09:31:13 UTC from IEEE Xplore. Restrictions apply.

You might also like