See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/329881391

IoT enabled adaptive clustering based energy efficient routing protocol for
wireless sensor networks

Article in International Journal of Ad Hoc and Ubiquitous Computing · April 2018

DOI: 10.1504/IJAHUC.2019.10024265

CITATIONS READS

2 253

5 authors, including:

Muhammad Asad Dr-Muhammad Aslam

Nagoya Institute of Technology Wuhan University
28 PUBLICATIONS 107 CITATIONS 38 PUBLICATIONS 457 CITATIONS

SEE PROFILE SEE PROFILE

Naeem Ayoub Ehsan Munir

Queen's University Belfast COMSATS University Islamabad
17 PUBLICATIONS 87 CITATIONS 62 PUBLICATIONS 727 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Communication Efficiency and Privacy Preservation in Federated Learning View project

Visual Saliency detection View project

All content following this page was uploaded by Muhammad Asad on 16 November 2020.

The user has requested enhancement of the downloaded file.

Int. J. Ad Hoc and Ubiquitous Computing, Vol. 32, No. 2, 2019 133

IoT enabled adaptive clustering based energy efficient

routing protocol for wireless sensor networks

Muhammad Asad*
School of Computer Science and Technology,
Dalian University of Technology,
Dalian, 116000, China
Email: [email protected]
*Corresponding author

Muhammad Aslam
Department of Computer Science,
COMSATS University Islamabad,
WAH CANTT, 47040, Pakistan
Email: [email protected]

Yao Nianmin and Naeem Ayoub

School of Computer Science and Technology,
Dalian University of Technology,
Dalian, 116000, China
Email: [email protected]
Email: [email protected]

Khalid Ibrahim Qureshi

School of Software Engineering,
Dalian University of Technology,
Dalian, 116000, China
Email: [email protected]

Ehsan Ullah Munir

Department of Computer Science,
COMSATS University Islamabad,
WAH CANTT, 47040, Pakistan
Email: [email protected]

Abstract: In this paper, we propose a new internet of things (IoT) enabled multi adaptive
clustering (MAC) energy efficient routing protocol for wireless sensor networks (WSNs) to
minimise energy dissipation and improve the network performance. This new technique holds
the hybrid cluster formation algorithm in which the network topology is divided in two regions
where the first region is centralised and the second region is distributed. Both regions contain
homogeneous and heterogeneous nodes while the sink is static and located in the centre of
both networks. Specifically, proposed IoT enabled MAC routing protocol holds the major three
properties: enabling of resources to sensor nodes through IoT, hybrid cluster formation to
distribute the network load evenly among sensor nodes and a new mechanism to minimise the
energy consumption in long range data transmission. In all of our experiments MAC achieves
greater performance than state-of-the-art routing protocols. Moreover, performance evaluation
proofs that MAC outperforms in critical network environment.

Keywords: internet of things; IoT; wsnS; wireless sensor networks; energy efficient; routing
protocols.

Reference to this paper should be made as follows: Asad, M., Aslam, M., Nianmin, Y.,
Ayoub, N., Qureshi, K.I. and Munir, E.U. (2019) ‘IoT enabled adaptive clustering based
energy efficient routing protocol for wireless sensor networks’, Int. J. Ad Hoc and Ubiquitous
Computing, Vol. 32, No. 2, pp.133–145.

Copyright © 2019 Inderscience Enterprises Ltd.

134 M. Asad et al.

Biographical notes: Muhammad Asad received his Bachelor degree in Telecommunication and
Networks from COMSATS Institute of Information Technology, WAH CANTT, Pakistan in
2014. Now he is pursuing his Masters in School of Computer Science and Technology, Dalian
University of Technology, China under the supervision of Prof. Yao Nianmin. His main research
interests include wireless sensor networks, internet of things, vehicular ad hoc networks and
artificial intelligence.

Muhammad Aslam received his BS in Telecommunication System and his MS in Electrical

Engineering from BZU Multan and COMSATS IIT Islamabad in 2010, 2012, respectively. He
received his PhD from School of Computer Science of Dalian University of Technology, China.
He is currently working as an Assistant Professor at COMSATS IIT, Wah Cantt. His major
research interests are energy optimisation in WSNs, WBANs, UWSNs and SDWSNs.
Yao Nianmin is a Professor at School of Computer Science and Technology of Dalian University
of Technology. He got his Bachelor, Master and Doctor degree from Jilin University. His main
research interests include wireless ad hoc networks and artificial intelligence.
Naeem Ayoub is pursuing his PhD in Computer Science and Technology from Dalian University
of Technology, China. He received his Masters in Information Technology from the University
of Education Lahore, Multan, Pakistan. His research interests include computer vision, image
processing, IoT and WSNs.
Khalid Ibrahim Qureshi is pursuing his PhD from School of Software Engineering, Dalian
University of Technology, China. He received his Masters in Telecommunication and
Networking from Bahria University Islamabad, Pakistan. His research interests include software
defined networks, wireless communication and network security.
Ehsan Ullah Munir received his Masters in Computer Science from Barani Institute of
Information Technology, Pakistan in 2001. He completed his PhD in Computer Software and
Theory from Harbin Institute of technology Harbin, China in 2008. He is currently working
as an Associate Professor in CIIT Wah, Pakistan. His research interests include heterogeneous
parallel and distributed computing, wired and wireless networks, and information retrieval.
This paper is a revised and expanded version of a paper entitled ‘HADCC: hybrid advanced
distributed and centralized clustering path planning algorithm for WSNs’ presented at 2014
IEEE 28th International Conference on Advanced Information Networking and Applications,
Victoria, BC, Canada, 13–16 May, 2014.

1 Introduction light-weight and small size sensor nodes that are used
to supervise the environment or system. These sensor
Internet of things (IoT) provides the connectivity between nodes become more advanced and intelligent when they
physical devices such as, vehicles, building and all those are equipped with IoT enabled micro systems in them,
things that contain actuators, sensors, embedded electronics so they can knob more complicated functions in WSNs
or software (Zhang et al., 2014a). By enabling IoT, these technology (Baker et al., 2017). IoT enabled WSNs is one
intelligent items can collect and exchange information of the latest platform in today’s smart energy management
for multiple purpose more efficiently (Shen et al., 2017). systems, it provides a dedicated gateway between sensor
IoT applications and concepts are emerging now a days nodes and Base Station (BS) (Komalavalli, 2017). Once
which encourage researchers to put effort for industry and the sensor nodes are deployed in the network they
academia (Siow et al., 2018; Paul and Saraswathi, 2017; Lin cannot be recharged or replaced due to the high cost.
et al., 2017). Interwork of wireless sensor networks (WSNs) Therefore, lifetime enhancement while maintaining the
and IoT brings reliability and efficient data flow between energy resources throughout the network operations are the
both networks (Chernyshev et al., 2017). Figure 1 shows essential for improving the performance of routing protocol
the graphical view of inter-networking between WSNs and in WSNs (Zhang and Zhang, 2012).
IoT. In order to manage the network, the concept of cluster
Development of WSNs are based on micro electro and cluster-head (CH) is proposed which has an attractive
mechanical systems (MEMS), system on chip (SoC), methodology to make the network energy efficient (Haseeb
wireless communications and low power embedded systems et al., 2017; Zhang et al., 2017a; Dhiman and Singh,
(Zhang et al., 2015b, 2014b). Presently, WSNs are widely 2017; Awasthi et al., 2017; Ma et al., 2017). Classification
used in hostile region, military applications and security, of routing protocol is given in Figure 2. In clustering
remote territory, disaster areas and also in health sector routing protocol, network is divided into clusters and each
(Bala et al., 2018). Particularly, WSNs consists of cluster is managed by CH. The selection of CH variate in
 IoT enabled adaptive clustering based energy efficient routing protocol for wireless sensor networks 135

routing protocol but the major responsibilities of CH are • Experimental results show that MAC produce 204%,
the same in each routing protocol. CH is accountable for 119%, 56% and 72% higher stability period than
organising the cluster, management of routing table, collect LEACH-C, DEEC, D-DEEC and E-DEEC.
data from all the associated member nodes, compressing
and transmitting of data. Due to the more responsibilities, Particularly, the aforementioned contributions significantly
CHs consumes more energy than the computational energy extend the field of applications where the energy efficient
of whole network (Ma et al., 2017; Zhang et al., 2015a, routing protocol is required.
2012). Remainder of this paper is organised as follows: In
Section 2, the related work and motivations are briefly
Figure 1 WSN assisted with IoT explained. System models and problem statement are
defined in Section 3. Technical modelling of our proposed
MAC are detailed in Section 4. Performance evaluation
and evaluation metrics are given in Section 5. Simulation
results are compared in Section 6. Results are analysed in
Section 7. Characteristics comparison of proposed MAC
with other protocols is given in Section 8. Conclusion and
future works are drawn in Section 9.

2 Related work and motivations

Figure 2 Classification of energy efficient routing protocols
In this section, we discuss the existing work on energy
efficient schemes of WSNs, several works are proposed
in the literature to extend the network lifetime of routing
protocol in real-time scenarios of WSNs. We briefly
describe the strength and weakness of some routing
protocols in the following paragraph.
Low energy adaptive clustering hierarchy (LEACH)
(Heinzelman et al., 2000) is the pioneer of hierarchical
routing protocol in WSNs. The main contribution of
LEACH is the dynamic selection of CHs using stochastic
To summarise, managing the energy dissipation of CHs algorithm. following equation is used in stochastic
plays an important role in prolonging the network lifetime algorithm to select a new CH in each round.
of routing protocol and archive energy efficiency. {
1 , if n ∈ N ,
P
Aiming towards higher energy efficient network, a
Tn = 1−P (r×mod× P ) (1)
new protocol named, multi adaptive clustering (MAC) is 0, otherwise
proposed to maintain the resources of sensor node in IoT
enabled WSNs. The major contribution of this paper are as This scheme of replacing CHs in each round divided
follows: the network load evenly on all nodes which results
into significant amount of energy dissipation. Huge
• We propose a new technique and architecture of
amount of data transmission leads to shortage of network
MAC by enabling IoT which supports application
lifetime because the nodes consumes more energy while
specific requirement of WSNs.
transmission, this proofs the inefficiency of LEACH in large
• MAC provides better stability and better network scale networks. Furthermore, LEACH is purely distributed
lifetime due to the hybridisation of network. MAC network which lacks it the most of application specificity.
also distribute the network load evenly among all the Various version of LEACH are proposed in the past to
sensor nodes based on residual energy and distance to improve the performance of LEACH.
BS. LEACH-C (Centralised) (Heinzelman et al., 2002) is
the well-known routing protocol, which is proposed to
• Advance Cluster-head selection algorithm (CHSA)
overcome the limitations of LEACH by using the central
which can execute in a hybrid network and selects
control algorithm. The network topology is same but the
optimum number of CHs. This unique CHSA
BS is in charge of generating the CHs by the knowledge
algorithm maintain centralised and distributed cluster
of location and remaining energy of sensor nodes. BS
formation which makes sure of higher packet delivery.
calculates the average energy of network by this knowledge
• MAC offers heterogeneity which makes it more and compare the average energy with node’s remaining
suitable for large scale networks. Furthermore, energy. If a node has higher remaining energy than average
proposed routing protocol based on realistic energy then it will be selected as CH otherwise it will
environment which helps it to perform more be a member node. In this way, energy dissipation in
effectively in real-time scenarios. the selection process of CH and transmission overhead
136 M. Asad et al.

Table 1 Comparison of the state-of-the-art clustering routing protocols

Features of routing protocols MAC LEACH-C DEEC D-DEEC E-DEEC

Energy efficiency X X X
Heterogeneity-aware X X X
Scalability towards nodes X X X
Residual energy-based X X X
Application-aware X X X X
Organised Network X X
Evading rapid dissipation X X
Avoiding redundant packets transmissions X X X X
Controller assistance X X X
Complex algorithm X X X X
Controlled CHs formation X X X
Hierarchal forwarding X X X
Higher data delivery X X
Cooperative routing for long range transmission X X X
Inter-cluster communication X
Configuration through internet X

are significantly reduced. However, the overhead still 2017; Carrabs et al., 2017; Kaur and Mahajan, 2018).
exist in large amount of data transmission furthermore, Based on the above analysis, network performance can be
LEACH-C is not suitable for the network where the BS enhanced with proper network topology, network coverage
is located inside of network topology. Distributed energy and knowledge of remaining energy and location of sensor
efficient clustering (DEEC) (Qing et al., 2006) propose nodes. In this paper we propose a new routing protocol; IoT
a distributed network with heterogeneous property by Enabled Multi Adaptive Clustering energy efficient routing
utilising two-level of heterogeneity. Similar like LEACH, protocol supported by the new unique CHSA algorithm
DEEC also distributes the overall network load among all which selects optimum number of CHs in a hybrid network
the sensor nodes by dividing the CH responsibilities. During which brings better stability and improved network lifetime
the selection of CHs, DEEC does not need any global in simulations. Features comparison of proposed MAC
knowledge which is the key advantage of this protocol. The routing protocols with state-of-the-art routing protocols is
probability threshold of DEEC allows each node to take part given in Table 1.
in CH election.
{ Pi
1 , if si ∈ G , 3 System models and problem statement
Tsi = 1−Pi (r×mod× Pi ) (2)
0, otherwise
In this section, system models and network architecture
Developed-DEEC (D-DEEC) (Elbhiri et al., 2010) is are discussed in detail. Subsequently, the heterogeneous
proposed to balance the energy distribution among all the network and energy model are introduced. Moreover,
sensor nodes in the network. The method of CH selection problem definition and possible solution is described.
is similar as DEEC in D-DEEC, the only difference
in expression of selection CHs in advance nodes. The
probability of being a CH is higher in advance nodes 3.1 Problem statement
than normal nodes. Which results a drawback of routing
As described in Section 1, due to the small size of
protocol because the distributed algorithm always selects
sensor nodes WSNs has various limitations such as storage,
the advance nodes as a CH which leads to early death
computing, communicating and energy. As the requirement
of these nodes and creates energy-hole in the network.
of longer network lifetime, most of the existing literature
Enhanced-DEEC (E-DEEC) (Saini and Sharma, 2010)
considered energy resources by minimising the energy
proposed as an extension of DEEC routing protocol.
dissipation through different techniques (Yetgin et al.,
E-DEEC propose three different types of nodes in the
2017). By allowing WSNs to be an integral part of IoT,
distributed network normal, advance and super nodes.
these sensor nodes can be controlled for all these resources.
Initial energy of these nodes increases in each type
In this paper, we mainly focus on energy consumption and
respectively, while the CH selection probability is in reverse
tried to extend the network lifetime with the help of IoT.
order.
Each routing protocol proposed in the past tries to
overcome the energy issue and improved some network 3.2 Heterogeneous network model
lifetime (Zhang et al., 2016; Zhang, 2012; Zhang et al.,
2013). Each routing protocol proposed different cluster MAC routing protocol considers two types of sensor nodes
formation with proper functional mathematical analysis in order to sense the environment; normal nodes and super
(Zhang et al., 2017b; Misra et al., 2017; Fei et al., nodes. In the proposed model, normal nodes are referred
 IoT enabled adaptive clustering based energy efficient routing protocol for wireless sensor networks 137

Figure 3 Network design and architecture of multi adaptive clustering

Figure 4 Clustering mechanism and topology of multi adaptive clustering

as homogeneous nodes and super nodes are referred as In equation (4), ET x−elec represents the transmission
heterogeneous nodes. The initial energy of normal nodes is energy while ET x−amp represent the links on a long
set to 0.5J whilst the initial energy of advanced nodes is set distance potentially affected by ”multi-path” fading.
to 1.0J. This extra property of heterogeneity brings more {
jEelec + Ef s jd2 , if d < dT h .
stability and robustness during selection of CHs because ET x (j, d) = (5)
CHSA selects super nodes more often than normal nodes jEelec + Eamp jd4 , if d ≥ dT h .
due to the higher remaining energy. The IoT brings advance In equation (5), ET x (j, d) is the total energy consumption
scalability in MAC and allows to add more super nodes for transmitting a j − bit data packet from a source to
anytime during the network operations in order to increase destination d. The energy consumption in transmission is
the stability and network lifetime. To calculate the total denoted by either Ef s jd2 or Eamp jd4 depending on the
initial energy of network, heterogeneous network model distance threshold dT h and the distance d while d ≥ dT h
applies: reflects the condition of ‘free-space’.
In order to receive this data packet the model applies:
EIE = Nn (1 − a)Eo + Ns Nn Eo (1 + α)
(3) ERx (j) = ERx−elec (j) = jEelec (6)
= Nn Eo (1 + αNs )
In equation (6), ERx (j) represents the total energy
consumption for receiving a j − bit data packet.
In equation (3), EIE is the total initial energy of the
network, Ns represents the super nodes with initial energy
Eo(1+α) and (1 − a) while Nn represents the normal nodes 3.4 Network architecture and design
with initial energy Eo .
The network architecture of MAC is divided into two
regions; first region is centralised and the second region
3.3 Energy consumption model is distributed. Normal and super nodes are deployed in
both regions while the BS is located in the centre of
Energy consumption model is used to measure the the network topology represented in Figure 3. Whereas,
performance of MAC. Depending on the distance between Figure 4 represents clustering topology of MAC, both
transmitter and receiver, first order radio model and free regions have their own clusters and CHs.
space model are used for fair comparison with other routing
protocols (Martinez-Sala et al., 2005; Yick et al., 2008).
Parameters of radio models are Eelec = 50 nJ/bit, Ef s = 4 Proposal of IoT enabled multi adaptive clustering
2 4
10 pJ/bit/m andEamp = 0.0013 pJ/bit/m . Thus, to send a routing protocol
j − bit message to a distance d, the model applies:
In this section, the MAC routing protocol is discussed in
ET x (j, d) = ET x−elec (j) + ET x−amp (j, d) (4) detail along with all major phases.
138 M. Asad et al.

4.1 Proposed model and ordinate of node’s location. The last field Energy level
indicates the remaining energy of sensor node.
The proposed MAC features centralised and distributed
clustering algorithms which brings the hybrid cluster Figure 6 Format of initialisation message
formation. As we have discussed earlier, the network
topology of MAC is divided into two regions and static
sink is located in the centre of the network. MAC
performs the cluster formation simultaneously in both
regions; centralised and distributed. In order to execute and Similarly, the same message is broadcasted in the
forming the clusters, MAC algorithm is divided into five distributed region and each node gets aware from other
major phases; network initialisation phase (NIP), network node’s location and energy level.
set-up phase (NSP), nodes association phase (NAP),
data transmission phase (DTP) and network termination
4.3 Network set-up phase (NSP)
phase (NTP). All major phases of MAC along with
communication flow is shown in Figure 5. In NSP of MAC, BS calculates the distance between sensor
node and BS itself and update the nodes table including
Figure 5 Communication flow of multi adaptive clustering node’s location, and energy level. Then BS clusters the
routing protocol (see online version for colours)
centralised region based on the distance and broadcast a
response message to the nodes in each cluster. Figure 7
shows the format of response message. The field M essage
type indicates that this message is an association message
for member nodes. The Distance field is used to inform
the maximum broadcasting range for all the sensor nodes
in centralised region. The field CH ′ s ID is used to specify
the ID of CH whilst the last field Average Energy
contains the network average energy.

Figure 7 Format of response message

After receiving this response message, all sensor nodes in

centralised region update this information in their routing
table. After formation of clusters in the centralised region,
the MAC starts cluster formation in the distributed region.
When the nodes received the initialisation message, each
node broadcast a message to become a CH, node with
the highest energy level will be selected as CH and all
the remaining sensor node will associate with the nearest
CH and forms the clusters. Detailed CHs Selection Process
(CHSP) for both regions is given below.

CHSP in centralised region. If there are M CHs and

each cluster have N member nodes. Then each CH
dissipates energy while receiving and transmitting the
information. Therefore dissipated energy of each CH during
4.2 Network initialisation phase (NIP) communication will be calculated as:

In the NIP of MAC, first the cluster formation is initialise MCH (j, d) = jEelec Ni + jEDA (Ni ) + 1 + jEelec
in the centralised region where the BS controls the central +Eamp jdtoBS (7)
algorithm while sensor nodes in the distributed region holds
the control of algorithm and initialise the cluster formation. In equation (7), jEelec Ni represents the received signal from
In the centralised region, a message containing node ID member nodes, jEDA (Ni ) represents the data aggregation
and location is sent to BS from each sensor node. The signal, jEelec shows the transmission of aggregated signal
format of initial message is shown in Figure 6. The field and Eamp jdtoBS represents the distance from CH to BS. In
M essage T ype specifies that it is an initialisation message, order to localise member nodes BS broadcast initialisation
Source ID contains the node ID of transmitting node. The message IM to request member nodes for their location.
X-coordinate and Y -coordinate represents the abscissa Nodes response to BS and the message contains ID,
 IoT enabled adaptive clustering based energy efficient routing protocol for wireless sensor networks 139

position, and the remaining energy of sensor node. By the The above mention equations showing that the nodes
knowledge of initial message, BS will calculate the energy having the slighter value of epoch and have more energy
consumption ratio (ECR) of each node by the following with slighter distance to sink. These nodes have the higher
equation: chance to become a CH.
Eo When CHs are chosen they will assign TDMA schedule
ECRNi = (8) to member nodes for communication.
Eo − Er
∑
n
In equation (8), Eo represents the initial energy and jEelec + Eamp j(dtom (m)) + jEelec (14)
Er represents the residual energy of sensor node. After m=1
calculating the ECR, BS will able to calculate the suitability
for final CH selection. In order to calculate the suitability, BS assigns TDMA schedule to the CHs and when CHs
algorithm applies: receives data from member nodes they will transmit that
data packet to the BS including their ID and remaining
Suitability = Er /(ECR × distoBS ) (9) energy.
After calculating the suitability, BS will choose the perfect EmCH = jEelec + jID Eelec + jdtBS (15)
CHs by the desired percentage (10.0% in MAC) of being a
CHs. Nodes with higher suitability will be selected as CHs. Algorithm 1 is proving that CH is taking the responsibility
of load balancing which clearly leads it to increasing in the
CHSP in distributed region. After selection of CHs in network lifetime.
centralised region, CHSA performs cluster formation in
distributed region. When the formation executes in second Algorithm 1 Cluster head selection algorithm (CHSA)
region, the distributed property of CH selection will utilise
and deploy all the nodes. The CHSA will adjust in the
environment because the nodes are deployed away from
the sink. Residual energy is an important metric for
CH’s selection in the second region. All nodes in the
network broadcast the initialisation message in order to
communicate with each other. Number of iteration and
number of nodes to become CHs are symbolised as i ,Mi
respectively and the CHSA rotates as an epoch.
When all the energy levels are calculated; than the final
decision of CH will based upon threshold value. In order to
calculate the threshold value, CHSA performs the following
equation:
{ PMi
(1−PMi (rand× P 1 )×d)
, if n ∈ Υ ,
Threshold = Mi (10)
0, otherwise

In equation (10), Υ represented as a set of nodes, d is a

distance between nodes and BS, and r is current round.
If Mi is not selected as a CH in the running round
than n belongs to Υ. In order to become a CH Mi will
choose a random number by itself from 0 to 1. CHSA 4.4 Nodes association phase (NAP)
will compare the random number with threshold, if the
threshold is greater than the random number then Mi will When the CHs are chosen for the current round than
be chosen as a CH otherwise it will be a member node of selected CHs will broadcast advertisement for node
that cluster. To calculate the energy level according to the association using CSMA MAC (Medium Access Control)
desired percentage, the algorithm applies: protocol. All CHs broadcast the association message in
their cluster range and all the nodes in the network will
El = Eoptimum /1 + α (11) stay active during the NAP, in this way sensor nodes will
be able to choose the suitable CH. In order to associate
Energy level of super nodes which contains some extra
with the CH, each node calculates the range by following
energy than normal nodes will be calculated as:
equation:
PEE = Popt 1 + α/1 + αM, (12)
Selectcriteria = RSSI/DisInRangeof CH , (16)
In the first centralised region the desired percentage will be
calculated to become a CH is: In equation (16), RSSI represents the Received Signal
Strength Indication and DisInRangeof CH denotes the
∑
N
PCR = Popt N (1 + αi )/N + αi, (13) distance of CHs to the member nodes. When the nodes
i=1
receive the advertisement from CHs they will transmit the
140 M. Asad et al.

association request to their appropriate CH using CSMA Algorithm 2 Local energy aware algorithm
medium access control (MAC) protocol. After receiving the
association request CH will broadcast TDMA schedule to
its corresponding nodes. When nodes get TDMA schedule
they will consider it as a confirmations information that
this node is associated to the particular cluster. CHs assigns
TDMA slots to each node; each node will only transmit its
sensed information on that time slot otherwise it will go the
sleep mode.

4.5 Data transmission phase (DTP)

According to the MAC network topology the algorithm

transmits data packets into time based iterations. So the
node can only send its sensed data in the given time
slots. Energy consumed by the node in one round can be
calculated as:
5 Performance evaluation
EM N −CH = jEelec + jEf s d2toCH (17)
In this section, we simulate our proposed model and
Nodes are randomly distributed in region X × Y and the
compare it with state-of-the-art routing protocols namely;
occupied area by each cluster is n2 /m, in order to measure
LEACH-C, DEEC, D-DEEC and E-DEEC.
the distance from member nodes to CHs we use the
following equation:
∫ ∫
5.1 Simulation parameters and methodology
DistoCH = (x2 + y 2 )r(x, y)dxdy =
∫ ∫ (18) We evaluate the performance of MAC routing protocol
τ 2 r(τ, θ)τ dτ dθ implemented with MATLAB v9.2. For fair comparison,
we use the same probability of signal collision and
After sending the data packet or when the node has no interference in wireless channel as in existing routing
schedule of transmitting the data packet it will turn-off protocols. Furthermore, we adopt the same Media Access
its transmitter and receiver; this is another property of Control (MAC) as in LEACH-C, DEEC, D-DEEC and
saving energy. Until the working time of current running E-DEEC.
round; each CH will remain active all the time during the
operations of network. Using multiple signal processing Table 2 Characterisation of simulation parameters
techniques CHs will compress the received data packets
Parameter Value
from their corresponding nodes and send the useful
information to the BS. Energy consumed by the CH during Initial Energy of homogeneous nodes .3J
the transmission of this message will be: Initial Energy of heterogeneous nodes 1.0J
(n ) Percentage of being Cluster Head 10.0%
n EDA Energy cost 50pJ/bitJ
ECH = jEelec − 1 + jEDA + jEelec
m m Number of nodes 100
+jEamp d2toBS (19) Number of rounds 9999
packet size 4000 bit
In order to minimise the hot-spot area problem, we use ET X 50nJ/bit
super nodes act as relays for sensors which are located far ERX 50nJ/bit
from CHs. These nodes sensed their data and transmit it to Ef s 10pJ/bit/m2
super nodes. Super nodes will add this data with its own Eamp 100pJ/bit/m2
data and aggregate to the CH. To calculate the distance from
the node to super node and CH, we use the simple local
In order to explain the major simulation parameters,
energy aware algorithm which is detailed in the following
we dispersed 100 sensor nodes in 100 × 100 m network
Algorithm 2.
dimensions which are connected through internet with
single static BS located in the centre of the network
4.6 Network termination phase (NTP) and considered 9999 number of rounds. Initial energy
of homogenous sensor nodes is set to 0.5 J while
After transmitting all the sensed data to BS, NTP checks heterogeneous sensor nodes are set to 1.0 J. Both
the rounds status if the current running round is completed homogeneous and heterogeneous sensor nodes are
than NTP checks the alive nodes in the network otherwise distributed randomly in centralised and distributed regions
algorithm go back to the NSP. If there is no alive node left of network. Size of data packet per transmission is 4000
in the network than the algorithm will terminate itself. bits and the percentage to become CHs per round is
 IoT enabled adaptive clustering based energy efficient routing protocol for wireless sensor networks 141

set to 10.0% in both regions. The detailed performance 6.1 Number of alive nodes
parameters that are used in our simulations are given in
Table 2. Number of alive nodes is the most important measurement
metric for any routing protocol in WSNs. Figure 8. shows
the evidence of improvement in proposed MAC protocol in
5.2 Evaluation metrics the comparison of existing routing protocols. The network
lifetime of proposed MAC protocol is improved by 7124
The following performance metrics are considered for rounds, 5600 rounds, 3339 rounds and 5013 rounds from
comparison: LEACH-C, DEEC, D-DEEC and E-DEEC respectively.
Nodes in LEACH-C die very early because of large amount
• Network lifetime or DLN (Death of Last Node): of communication control messages, whilst DEEC and
During the operation of network, sensor nodes utilises E-DEEC perform quite better than LEACH-C but due to
their energy for transmission and collection of data. the lack of resources and long range communication these
These sensor nodes eventually starts dying after protocols do not survive for longer period. Whereas, the
consuming all the energy. Network lifetime is sensor nodes in D-DEEC survived till the 5128 rounds
measured in rounds and the current running round at which is best among other routing protocols but the
the time of DLN will be considered as the maximum network using MAC has a very long network lifetime which
lifetime of network. performs efficiently till 8467 rounds.
• Network stability or DFN (Death of First Node):
Network stability is directly proportional to Figure 8 Number of alive nodes (see online version
for colours)
networklifetime and it is the most important factor in
energy efficient routing protocols. Current running
round at the time of DFN is considered as the
network stability, after the DFN network instability
period starts.
• Network instability Period (NIP): Total number of
rounds/time from the DFN to DLN is the instability
period of routing protocol.
• Packets transmitted to CHs (Cluster Heads): After
collecting the data from sensing field, sensor nodes
transmit this data towards CHs. Maximum received
data packets by CHs strengthen the routing protocol.
• Packets transmitted to BS (Base Station): After
receiving the data from sensor nodes, CHs compress
all the data and forward it to BS. Maximum received
Figure 9 Number of dead nodes (see online version
data packets by BS shows the robustness of routing for colours)
protocol.
• CHs per round: Minimum generation of CHs in
single round while maintaining the requirement of
network helps in prolonging the network lifetime.
• Packet delivery ratio (PDR): Communication ability
of routing protocols can be measured by calculating
the PDR. Ratio of the successfully received data
packets from the total number of transmitted data
packets is the PDR of any routing protocol.

6 Comparison results

In this section, we compare the MAC routing protocol

with other four protocols (LEACH-C, DEEC, D-DEEC 6.2 Number of dead nodes
and E-DEEC). All five protocols are deeply analysed with
aforementioned evaluation metrics. Figures 6–10 shows the Number of dead nodes is used to measure the stability
comparison of proposed protocol with above mentioned period of any routing protocol in WSNs. Figure 9 shows the
other protocols. The detailed comparison of all the five number of dead nodes with respect to number of rounds.
protocols is as follows. The difference of network stability between MAC are 1995
142 M. Asad et al.

rounds, 1618 rounds, 1075 rounds and 1245 rounds in the is large then the whole network performs poorly such as,
comparison of LEACH-C, DEEC, D-DEEC and E-DEEC sudden energy dissipation and smaller data transmission
respectively. The DFN in proposed protocol is at 2972 which results into shorter lifespan of network. Taking all
round which is 204%, 119%, 56% and 72% better than these aspects into consideration, percentage of being a CHs
other protocols respectively. in proposed MAC protocol is set to 10.0% which generates
perfect CHs to lead the network. This percentage of CHs
covers the whole network area and distribute the energy
6.3 Number of packets transmitted towards base-station resources equally among all the sensor nodes.
(BS)
Figure 11 Packets transmission to CHs (see online version
Figure 10 shows the number of total data packets for colours)
transmitted and received by the BS. Figure shows
that D-DEEC delivers considerable data packets in the
comparison of LEACH-C, DEEC and E-DEEC. But the
data packets received in the network of MAC is higher than
D-DEEC. This is due to the local energy aware algorithm
in the proposed model, when the distance between CH
is greater than the super nodes then the super node will
act as a CH which helps to deliver the maximum data
packets without delay. However, after the 2972 rounds,
MAC packet delivery decreases due to the settings of local
energy aware algorithm. When more nodes starts dying
in the network, the distance between sensor nodes will
increase then the packets received by BS faces considerable
amount of delay.

Figure 10 Packets transmission to BS (see online version

for colours) Figure 12 CHs generated per round (see online version
for colours)

6.4 Number of packets transmitted towards

cluster-heads
7 Results analysis in different size of networks
Figure 11 shows that the packet delivered to CHs by
MAC protocol is better than other protocols; MAC protocol
In this section, we analyse the proposed model in the
consumes less energy which results into maximising in
comparison of LEACH-C, DEEC, D-DEEC and E-DEEC.
the lifespan of network and also increasing in the packet
To improve the quality of graphs, first we simulate all the
delivery towards CHs.
routing protocols in MATLAB v9.2 and then generate the
graphs in OriginLab for deep analysis.
6.5 Number of cluster-heads generated The simulation in this section, consists of 100 sensor
nodes which are dispersed in 100 × 100 m, 200 × 200 m
Figure 12 vividly shows the number of CHs generated with and 300 × 300 m, respectively. The rest of the simulation
respect to number of rounds. When the percentage of being parameters are same as mentioned in Table 2. The reason
a CH in network setting is small then the sensor nodes of this simulation is to prove the performance of proposed
begin to die very early while when the percentage of CHs protocols in large scale networks.
 IoT enabled adaptive clustering based energy efficient routing protocol for wireless sensor networks 143

Table 3 Characteristic comparison of proposed protocol with state-of-the-art routing protocols

Routing protocols LEACH-C DEEC D-DEEC E-DEEC MAC

Death of first node (DFN) 977 rounds 1354 rounds 1897 rounds 1727 rounds 2972 rounds
Death of last node (DLN) 1343 rounds 2867 rounds 5128 rounds 3454 rounds upto 8467 rounds
NIP 366 rounds 1513 rounds 3231 rounds 1727 rounds upto 5495 rounds
Packet Delivery Ratio Bad Better Better Good Best
Scalability Bad Bad Good Better Best
Localisation-Awareness Yes Yes No No Yes
Communication Path Single-Hop Single-Hop Single-Hop Single-Hop Single-Hop
BS Location Fixed at centre Fixed at center Fixed at centre Fixed at centre Fixed at centre

Figure 13 vividly shows the network lifetime of proposed ratio (Zhou et al., 2009). The probability of packet delivery
protocol in different size of networks. Due to the small fluctuate due to the change of distance between CHs and
amount of sensor nodes, network lifetime of all five BS because the CHs are changing with respect to rounds.
protocols decreases gradually in larger networks. Similarly, To measure the accurate packet delivery ratio we use the
Figure 14 shows the network stability period of all five following equation:
routing protocols. It can be seen clearly that proposed MAC

routing protocol perform much better in comparison of 
1 if DCH,BS > 100
LEACH-C, DEEC, D-DEEC and E-DEEC. Figures proved P DR = 1/20×(DCH,BS - 80) if 80 ≤ DCH,BS ≤ 100
that MAC routing protocol is suitable for large scale of 

networks in terms of network lifetime and network stability. 0 if 0 ≤ DCH,BS < 80
(20)
Figure 13 Network lifetime (see online version for colours)
where DCH,BS shows the distance between CH and BS,
if the probability is larger than PDR, than the packet
will successfully delivered otherwise it will be dropped.
Figure 15 shows the number of packets which are delivered
successfully. Figure proves that proposed MAC routing
protocol has the highest packet delivery ratio not only in
standard network dimensions but also in larger network.

Figure 15 Packet delivery ratio (see online version for colours)

Figure 14 Stability of network (see online version for colours)

8 Characteristics comparison of MAC and existing

conventional routing protocols

The characteristics comparison over performance of

During the transmission of data packets towards BS, it LEACH-C, DEEC, D-DEEC, E-DEEC and MAC
is assumed that all data packets are successfully received are presented in Table 3. The characteristics used to
by the BS but in reality, due to the noise, reflection, compare the performance are DLN, DFN, NIP, PDR,
interference and attenuation some data packets are dropped. scalability, localisation-awareness, communication path
In order to perform the fair comparison, we use the uniform and BS location. As shown in Table 3, MAC performs
random distribution model to calculate the packet delivery better than LEACH-C, DEEC, D-DEEC and E-DEEC
144 M. Asad et al.

in terms of DLN, DFN, NIP, PDR, scalability, and Symposium on I/V Communications and Mobile Network
localisation-awareness. (ISVC, 2010), September, IEEE, pp.1–4.
However, for the case of communication path and BS Fei, Z., Li, B., Yang, S., Xing, C., Chen, H. and Hanzo, L.
location, MAC routing protocol is suitable for the network (2017) ‘A survey of multi-objective optimization in wireless
which contains the single-hop communication and where sensor networks: metrics, algorithms, and open problems’,
the BS is static and located in the centre of the network. IEEE Communications Surveys and Tutorials, Vol. 19, No. 1,
pp.550–586.
Haseeb, K., Bakar, K.A., Abdullah, A.H. and Darwish, T. (2017)
‘Adaptive energy aware cluster-based routing protocol for
9 Conclusion and future work
wireless sensor networks’, Wireless Networks, Vol. 23, No. 6,
pp.1953–1966.
In this paper, we propose a new technique of WSNs
Heinzelman, W. R., Chandrakasan, A. and Balakrishnan, H.
namely; multi adaptive clustering (MAC) that is assisted (2000) ‘Energy-efficient communication protocol for wireless
with IOT to enable the resources in sensor nodes. Proposed microsensor networks’, Proceedings of the 33rd Annual
routing technique consists of a new hybrid network Hawaii International Conference on System Sciences, 2000,
architecture which contains centralised and distributed January, 10-pp, IEEE.
network where the homogeneous and heterogenous nodes Heinzelman, W.B., Chandrakasan, A.P. and Balakrishnan, H.
are deployed and the BS is located in the centre of hybrid (2002) ‘An application-specific protocol architecture for
network. MAC also propose an advanced CH selection wireless microsensor networks’, IEEE Transactions on
algorithm (CHSA) which efficiently generates optimum Wireless Communications, Vol. 1, No. 4, pp.660–670.
number of CHs in both network by calculating the residual Kaur, S. and Mahajan, R. (2018) ‘Hybrid meta-heuristic
energy of sensor nodes. We compare the performance optimization based energy efficient protocol for wireless
of MAC with state-of-the-art routing protocols such as; sensor networks’, Egyptian Informatics Journal.
LEACH-C, DEEC, D-DEEC and E-DEEC. Simulation Komalavalli, C. (2017) ‘Convergence of wireless sensor networks,
results and performance evaluation of MAC provide the internet of things, big data: challenges’, Convergence, Vol. 6,
evidence of significant improvement in terms of network No. 6.
lifetime, network stability and packet delivery ratio. In Lin, J., Yu, W., Zhang, N., Yang, X., Zhang, H. and Zhao, W.
our future work, we would like to implement MAC (2017) ‘A survey on internet of things: architecture, enabling
in NS-3 simulation environment which will give the technologies, security and privacy, and applications’, IEEE
Internet of Things Journal, Vol. 4, No. 5, pp.1125–1142.
more accurate picture of performance improvement. The
real-time implementation of MAC can also be considered Ma, Z., Zhang, D.G., Chen, J. and Hou, Y.X. (2017)
‘Shadow detection of moving objects based on multisource
for future work.
information in internet of things’, Journal of Experimental
and Theoretical Artificial Intelligence, Vol. 29, No. 3,
pp.649–661.
References Martinez-Sala, A., Molina-Garcia-Pardo, J.M., Egea-Ldpez, E.,
Vales-Alonso, J., Juan-Llacer, L. and Garcia-Haro, J. (2005)
Awasthi, K.K., Singh, A.K. and Tahilyani, S. (2017) ‘Study of ‘An accurate radio channel model for wireless sensor
cluster based routing protocols in wireless sensor networks’, networks simulation’, Journal of Communications and
Optimization, Vol. 9, p.10. Networks, Vol. 7, No. 4, pp.401–407.
Baker, T., Asim, M., Tawfik, H., Aldawsari, B. and Buyya, R. Misra, S., Chatterjee, S. and Obaidat, M.S. (2017) ‘On theoretical
(2017) ‘An energy-aware service composition algorithm for modeling of sensor cloud: A paradigm shift from wireless
multiple cloud-based IoT applications’, Journal of Network sensor network’, IEEE Systems Journal, Vol. 11, No. 2,
and Computer Applications, Vol. 89, pp.96–108. pp.1084–1093.
Bala, T., Bhatia, V., Kumawat, S. and Jaglan, V. (2018) ‘A Paul, P.V. and Saraswathi, R. (2017) ‘The internet of things:
survey: issues and challenges in wireless sensor network’, a comprehensive survey’, International Conference
International Journal of Engineering and Technology (UAE), on Computation of Power, Energy Information and
Vol. 7, No. 2.4, pp.53–55. Commuincation (ICCPEIC, 2017), March, IEEE,
Carrabs, F., Cerulli, R., D’Ambrosio, C. and Raiconi, A. pp.421–426.
(2017) ‘Prolonging lifetime in wireless sensor networks with Qing, L., Zhu, Q. and Wang, M. (2006) ‘Design of a distributed
interference constraints’, International Conference on Green, energy-efficient clustering algorithm for heterogeneous
Pervasive, and Cloud Computing, May, Springer, Cham, wireless sensor networks’, Computer Communications,
pp.285–297. Vol. 29, No. 12, pp.2230–2237.
Chernyshev, M., Baig, Z., Bello, O. and Zeadally, S. (2017) Saini, P. and Sharma, A. K. (2010, October) ‘E-DEEC-enhanced
‘Internet of things (IoT): research, simulators, and testbeds’, distributed energy efficient clustering scheme for
IEEE Internet of Things Journal, Vol. 5, No. 3. heterogeneous WSN’, 1st International Conference on
Dhiman, S. and Singh, O. (2017) ‘Energy efficient routing Parallel Distributed and Grid Computing (PDGC, 2010),
protocol for wireless sensor networks: a review’, IJASRE, IEEE, pp.205–210.
p.3. Shen, J., Wang, A., Wang, C., Hung, P.C. and Lai, C.F. (2017)
Elbhiri, B., Saadane, R. and Aboutajdine, D. (2010) ‘Developed ‘An efficient centroid-based routing protocol for energy
dISTRIBUTED energy-efficient clustering (DDEEC) for management in WSN-assisted IoT’, IEEE Access, Vol. 5,
heterogeneous wireless sensor networks’, 5th International pp.18469–18479.
 IoT enabled adaptive clustering based energy efficient routing protocol for wireless sensor networks 145

Siow, E., Tiropanis, T. and Hall, W. (2018) ‘Analytics for Zhang, D.G., Wang, X., Song, X.D., Zhang, T. and Zhu, Y.N.
the internet of things: a survey’, ACM Computing Surveys (2015) ‘A new clustering routing method based on PECE for
(CSUR), Vol. 51, No. 4, p.74. WSN’, EURASIP Journal on Wireless Communications and
Yetgin, H., Cheung, K.T.K., El-Hajjar, M. and Hanzo, L.H. (2017) Networking, Vol. 2015, No. 1, p.162.
‘A survey of network lifetime maximization techniques in Zhang, D-G., Zheng, K., Zhang, T. and Wang, X. (2015b) ‘A
wireless sensor networks’, IEEE Communications Surveys novel multicast routing method with minimum transmission
and Tutorials, Vol. 19, No. 2, pp.828–854. for WSN of cloud computing service’, Soft Comput., Vol. 19,
Yick, J., Mukherjee, B. and Ghosal, D. (2008) ‘Wireless sensor No. 7, July, pp.1817–1827.
network survey’, Computer Networks, Vol. 52, No. 12, Zhang, D. G., Zheng, K., Zhao, D.X., Song, X.D. and Wang, X.
pp.2292–2330. (2016) ‘Novel quick start (QS) method for optimization of
Zhang, D-g. and Zhang, X-d. (2012) ‘Design and implementation TCP’, Wireless Networks, Vol. 22, No. 1, pp.211–222.
of embedded un-interruptible power supply system EUPSS Zhang, D.G., Liu, S., Zhang, T. and Liang, Z. (2017a) ‘Novel
for web-based mobile application’, Enterp. Inf. Syst., Vol. 6, unequal clustering routing protocol considering energy
No. 4, November, pp.473–489. balancing based on network partition & distance for mobile
Zhang, D., Kang, X. and Wang, J. (2012) ‘A novel image education’, Journal of Network and Computer Applications,
de-noising method based on spherical coordinates system’, Vol. 88, pp.1–9.
EURASIP Journal on Advances in Signal Processing, Zhang, J., Tang, J., Wang, T. and Chen, F. (2017b)
Vol. 2012, No. 1, p.110. ‘Energy-efficient data-gathering rendezvous algorithms with
Zhang, D.G. and Liang, Y.P. (2013) ‘A kind of novel method of mobile sinks for wireless sensor networks’, International
service-aware computing for uncertain mobile applications’, Journal of Sensor Networks, Vol. 23, No. 4, pp.248–257.
Mathematical and Computer Modelling, Vol. 57, Nos. 3–4, Zhang, D.G. (2012) ‘A new approach and system for attentive
pp.344–356. mobile learning based on seamless migration’, Applied
Zhang, D., Wang, X., Song, X. and Zhao, D. (2014a) ‘A Intelligence, Vol. 36, No. 1, pp.75–89.
novel approach to mapped correlation of ID for RFID Zhou, Q., Cao, X., Chen, S. and Lin, G. (2009) ‘A solution to
anti-collision’, IEEE Transactions on Services Computing, error and loss in wireless network transfer’, International
Vol. 7, No. 4, pp.741–748. Conference on Wireless Networks and Information Systems,
Zhang, D., Li, G., Zheng, K., Ming, X. and Pan, Z.H. (2014b) 2009. WNIS’09, IEEE, December, pp.312–315.
‘An energy-balanced routing method based on forward-aware
factor for wireless sensor networks’, IEEE Transactions on
Industrial Informatics, Vol. 10, No. 1, pp.766–773.

View publication stats