This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2019.2940988, IEEE Internet of
Things Journal
IEEE INTERNET OF THINGS JOURNAL, VOL. …, NO. …, MONTH 2019

I-SEP: An Improved Routing Protocol for

Heterogeneous WSN for IoT based
Environmental Monitoring
Trupti Mayee Behera, Sushanta Kumar Mohapatra, Senior Member, IEEE, Umesh Chandra Samal,
Mohammad. S. Khan, Member, IEEE, Mahmoud Daneshmand, Life Member, IEEE, and
Amir H. Gandomi, Senior Member, IEEE

Abstract— Wireless Sensor Networks (WSN) is a virtual layer in to record and monitor various physical and environmental
the paradigm of the Internet of Things (IoT). It inter-relates conditions with low-cost data acquisition. The sensor nodes
information associated with the physical domain to the IoT drove are usually energy deprived in nature which leads to the
computational systems. WSN provides ubiquitous access to
formulation of innovative techniques to limit any unnecessary
location, the status of different entities of the environment, and
data acquisition for long-term IoT monitoring. Since energy is a energy dissipation mounting to shortening of lifetime [3]. The
major constraint in the design process of a WSN, recent advances source nodes consume a lot of energy in communicating data
have led to project various energy-efficient protocols. Routing of directly to BS. Hence they have to depend on intermediate
data involves energy expenditure in considerable amount. In recent nodes to operate. A comparative study of routing protocols
times, various heuristic clustering protocols have been discussed to [4][5][6] indicates different methods to enhance network
solve the purpose. This article is an improvement of the existing
lifetime. Khalil et al. [7] have designed a dynamic clustering
Stable Election Protocol (SEP) that implements a threshold-based
cluster head selection for a heterogeneous network. The threshold scheme to reduce energy consumption while transferring
maintains uniform energy distribution between member and information thereby maintaining a trade-off between stability
cluster head nodes. The sensor nodes are also categorized into period and lifetime of the network. The method assumes only
three different types called normal, intermediate and advanced single-hop data transfer which may not be feasible for a large
depending on the initial energy supply to distribute the network scale network in an IoT system. Dynamic clustering used in
load evenly. The simulation result shows that the proposed scheme
[8] is based on multi-hop communication, where sensors
outperforms SEP and DEEC protocols with an improvement of
300% in network lifetime and 56% in throughput. transfer data via intermediate nodes to sink, but reducing the
Index Terms— WSN, Heterogeneous Network, CH selection, energy at node level is still a problem.
Network lifetime. Clustering algorithms have emerged as the most energy-
efficient communication protocol that groups the sensor nodes
I. INTRODUCTION in clusters. Each cluster is headed by a cluster head (CH)
responsible for data collection from sensing nodes. The CH
O VERthe last few years, the IoT paradigm has evolved as
one of the biggest technological advances of modern
science. With the evolving era of WiFi and 4G-LTE wireless
fuses the data to remove any redundancy and then transfers to
the sink node or BS. Hence election of CH should be done
access of Internet [1] [2], IoT enabled devices like computer, judiciously to maintain proper network balance for energy
tablets, mobile phones are able to access information about the management. The existing routing methods that select CHs
environment and other objects without human intervention. optimally may not be suitable for large scale environments
The two key enablers of IoT are Radio frequency where WSN is integrated with IoT.
identification (RFID) and WSN. Any ad-hoc system can be either heterogeneous or
A WSN constitutes spatially dispersed sensor nodes meant homogeneous. The network where sensor nodes are supplied
with equal amount of energy termed as homogeneous, is
Manuscript received XXX XX, XXXX; accepted XXX XX, XXXX. Date of shown in Fig. 1(a). Heterogeneous networks as shown in Fig.
publication XXXX XX, XXXX; date of current version XXXX XX, XXXX. 1(b) and (c) have uneven initial energy distribution. A group
T. M. Behera, S. K. Mohapatra, and U. C. Samal are with School of of nodes called advanced nodes has higher energy in
Electronics Engg., KIIT University, Bhubaneswar, 751024, India. (e-mail:
[email protected] ; [email protected] ; [email protected] ). comparison to normal nodes. The proposed network model
M. S. Khan is with Department of Computer & Information Sciences, East introduces intermediate nodes, along with normal and
Tennessee State University, Johnson City, USA (e-mail: intermediate nodes, that have energy in between normal and
[email protected] ). advanced nodes. When all sensors in the network start with a
M. Daneshmand is with the School of Business, Stevens Institute of
constant energy level, the nodes die out randomly within a
Technology, Hoboken, NJ 07030, USA and he is Steering Committee Chair of
IEEE IoT J (e-mail: [email protected] ) short span of time. Heterogeneous network structures are
A.H. Gandomi (corresponding author) is with the Faculty of Engineering gaining importance because it delivers better network
& Infromation Technology, University of Technology Sydney, Australia (e- performance without demanding much increase in cost [9].
mail: [email protected]).
One of the challenging IoT application domains is
Digital Object Identifier XXXXXXXXXXXXXXXX
"Copyright (c) 2012 IEEE. Personal use of this material is permitted. However, permission to use this material
for any other purposes must be obtained from the IEEE by sending a request to [email protected] "

2327-4662 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2019.2940988, IEEE Internet of
Things Journal
IEEE INTERNET OF THINGS JOURNAL, VOL. …, NO. …, MONTH 2019

environmental monitoring, where the sensors are deployed in level in the CH selection process that can be applied in a WSN
harsh operating conditions [10] [11]. Fig. 2 depicts a scenario based IoT network.
where sensor nodes are placed over soil, air, and water to Starting with the introduction, a brief analysis of related
deliver feasible or even optimal solution to monitor moisture, work is presented in the next section. Section III enlightens the
humidity, pH level, temperature. network model and the proposed scheme is provided in section
IV. Section V discusses simulation results with issue related
analysis. And finally, the conclusion is drawn.

II. RELATED WORK

Any clustering algorithm functions by segmenting the field
into clusters headed by a CH. The member nodes, ie, non-CH
nodes communicate their data to CH, where the data is
processed and aggregated to remove redundancy and sent to
(a) (b) the BS. As energy consumption is distributed evenly
throughout the network, the overall energy consumption is
said to be reduced [13].
LEACH is an energy constrained protocol [14]. The initial
CH selection is done randomly such that every node has the
chance of becoming CH once in every 1/p epoch [15]. In
subsequent rounds, a random number is generated in the range
[0,1] and only if the number is less than threshold Tn,
formulated by (1), the node functions as CH [16]. The next
(c) CH is chosen from the set of non-CH nodes G.
Fig. 1. (a) Homogenous (b) Heterogeneous (c) Proposed Network Model  p
 ; for n  G
Tn  1  p(r mod 1 p)
(1)
After deployment, the sensors are expected to keep sensing

the environment for a longer period and have no scope of 0; Otherwise
recharging the node battery. Furthermore, the cost and After the cluster formation stage, all the elected CHs
difficulty of accessing the field physically for deployment and broadcast a TDMA schedule for non-CH nodes. The sensing
maintenance [1] become a challenging task. The WSN nodes transmit the data during the specific allotted time slots.
platform should offer low-cost nodes with long unattended Once the transmission process completes, the frame repeats.
service time and minimal maintenance to overcome these Although LEACH is a distributed protocol, it may not be
issues. This is possible only if the network comprises sensors beneficial for large-scale application due to energy constraint.
with two or more initial energy levels, which is one of the In a heterogeneous network [9] of n nodes, a fraction of
important criteria considered in the proposed method. Hence it nodes has the extra energy of factor α that are termed
can be inferred that heterogeneous networks perform better advanced nodes. The SEP algorithm focuses on the weighted
when applied in the IoT environment. election probabilities of each node for the election of CHs
according to their respective energies. It ensures a more
extended stability period with better performance than that of
LEACH protocol. The DEEC protocol [17] was proposed for
networks with different energy level, where the selection of
CH can be decided by both initial as well as residual energy
[18]. An enhanced SEP was proposed in [19] that deployed
three categories of sensors based on energy levels; named
normal, intermediate and advanced nodes. Due to the three-
tier node scenario, the network lifetime is enhanced; however,
the quantity of CHs in each cluster could not be controlled.
Fig. 2. Environmental Monitoring using WSN EEHC is an energy-efficient heterogeneous clustering
scheme that elects CHs considering the weights of each sensor
LEACH (Low Energy Adaptive Clustering Hierarchy) is a [20]. The residual energy of each node decides the set of
fully distributed routing algorithm based on TDMA designed probable CHs in the heterogeneous network. MATLAB results
for homogeneous networks [12]. SEP (Stable election indicate enhanced network lifetime in comparison to LEACH.
protocol) is a variant of LEACH where certain populations of However, no result analysis is done with any heterogeneous
nodes (advanced nodes) have some additional energy than algorithms. Threshold sensitive SEP (T-SEP) is a reactive
other nodes (normal) within the same network [9]. This article protocol introduced in [21], where data is transmitted by
discusses an extension of SEP that intends to maximize the sensors only when the explicit threshold is reached. Three
network lifetime and throughput by introducing a threshold
level heterogeneous nodes were deployed to study the lifetime

2327-4662 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2019.2940988, IEEE Internet of
Things Journal
Behera et al.: I-SEP: An improved routing protocol for heterogeneous WSN for IoT based Environmental Monitoring

and stability period of the network. Another modification of III. SYSTEM MODEL
SEP for fog-supported WSN discussed in [22] maintains a Considering energy efficiency and energy balancing [28] as
balanced energy dissipation to prolong the network lifetime. the most vital parameter in the design process of any routing
Both types of nodes have equal probabilities to be elected as algorithm in WSN, an advanced technique to route data in a
CHs. As the CHs are chosen judiciously, the node death rate heterogeneous network is introduced. To control the energy
decreases in comparison to LEACH-DCHS [12] and other dissipation, three-level heterogeneity with respect to initial
modifications of SEP. node energy is considered. All the nodes are static in nature.
I-LEACH was proposed in [23] to introduce threshold Advance nodes have the maximum, and normal nodes have
based CH selection, where the LEACH protocol was modified the lowest level of energy. Intermediate nodes are the ones
to obtain better results for IoT based applications. The with higher energy than normal nodes and lower than
simulation result showed better performance for different advanced nodes. Let b be the section of nodes that are
scenarios in comparison to energy efficient routing protocols assigned an intermediate energy level with β times more
like LEACH, EECS [24], CPCHSA [25] and Mod-LEACH power than normal ones, where β=α/2.
[26]. However, the algorithm was designed only for E0 represents the initial energy given to normal nodes. The
homogeneous networks and cannot be implemented for advanced and intermediate nodes have E0(1+α) and E0(1+β)
heterogeneous scenarios. To overcome high system energies respectively. Hence, the total energy of each type of
complexity CREEP scheme was proposed in [27] that selects node can be summed up as:
numerous CHs to improve the network lifetime by modifying EN  nE0 (1  a  b) (2)
threshold value in a 2-level heterogeneous WSN. Unbalanced
EI  nbE0 (1   ) (3)
energy consumption near CHs limit the network lifetime;
particle swarm optimization based CH selection was proposed E A  naE0 (1   ) (4)
in [28] that enhances lifetime by identifying energy holes. The Where, EN, EI, and EA are the energies for normal, intermediate
approach assumes homogenous network where nodes die out and advanced nodes respectively. Therefore, the overall
randomly. energy of the three types of nodes written as
ETotal  nE0 (1  a  b)  nbE0 (1   )  naE0 (1   ) (5)
In SEP routing algorithm, the election of new CH with the
formation of new clusters is done regularly for each round.  nE0 (1  a  b )
This in return leads to unnecessary energy utilization The CH election process is similar to that of LEACH and SEP.
generated due to routing overhead which will affect the The threshold value for CH selection is formulated for each
performance of IoT devices [29] connected to the sensor type of node by considering their probabilities. Let G1, G2 and
network. According to the classical SEP algorithm, a CH in G3 represent the set of nodes in each category that had not
the current round will not be able to participate in the CH performed as CH in former epochs and r represents the current
round. Considering p(N), p(I) and p(A) as the probabilities of
election process in the next round [27]. However, there can be
normal, intermediate and advanced nodes to be elected as CHs
cases where a CH has not utilized an ample amount of energy
respectively. For normal nodes,
in the preliminary round and is eligible for the CH election p
process in the next round. It can also happen that a sensor with p(N)  (6)
1  a  b
a comparatively lesser amount of energy gets elected as CH in
 p( N )
the subsequent selection process [26] that leads to the  ; if n N  G1
T(n N )  1  p( N ) (r mod 1 p )
(7)
untimely death of the network. Also, new CH requires new (N )

cluster formation in each round, which consumes the node 0; Otherwise
power in sending messages like ADV (advertisement) and For intermediate nodes,
ACK (acknowledgment) to CHs back and forth. The above p(1   ) (8)
p(I) 
limitation in SEP motivates to investigate and establish an 1  a  b
efficient CH replacement method.  p(I)
 ; if n I  G2
The key contributions of the proposed work are: T(n I )  1  p(I) (r mod 1 p )
(9)
i. The article aims to enhance the fundamental SEP algorithm 
(I)

by incorporating a unique threshold strategy for CH 0; Otherwise

selection. For advance nodes, p(A)  p(1   ) (10)
ii. The proposed method aims to reduce extra power 1  a  b
consumption by avoiding unnecessary clusters and CHs  p(A)
 (11)
formation in each round.   mod 1
p(A) ; if n A  G3
T(n A ) 1 p(A) ( r )
iii. After CH selection, the proposed algorithm assigns a high 
0; Otherwise
level of energy to the node. For the subsequent rounds,
when the node again becomes a sensing node, the low level Now, from equation (6), (8) and (10), we can find the average
of energy will be assigned. This variation of energy level for total CHs per round as:
different nodes will be beneficial for maintaining proper n(1  a  b) p( N )  nap(A)  nb( I )  np (12)
energy distribution in the network. It can be inferred from (12) that the resultant of CHs in a
heterogeneous environment is equal to that in case of LEACH

2327-4662 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2019.2940988, IEEE Internet of
Things Journal
IEEE INTERNET OF THINGS JOURNAL, VOL. …, NO. …, MONTH 2019

protocol. However, the energy dissipation is controlled in a aggregation and fusion, hence it is desirable that they should
better way owing to the heterogeneous energy level [21]. be equipped with higher energy level in comparison to those
nodes meant for intra-cluster communication. This will
additionally save power and packet drop ratio. Hence, the
modified algorithm assigns a high energy amplification level
only for the selected CH. In the next round, if the CH switches
to a normal sensing node, the modified algorithm assigns a
Fig. 3. The first-order radio communication system low energy level [31][23] to the corresponding node.
The communication in the heterogeneous network follows With n number of total nodes in the network and C be the
the model shown in Fig. 3 [30]. If d (Euclidean distance percentage of clusters and R is the CH replacement count. PkTx
between sending and receiving nodes) is less than or equal to a and PkRx are the packet size at transmission and reception.
Let N=nC represents the size of each cluster.
reference distance d0 (where d 0  E fs Emp ) then energy The CH replacement process for new cluster formation also
dissipation is calculated using multi-path fading model utilizes some energy given as PHR , such that
otherwise free-space model is used. Assuming symmetrical PHR  PkTx PTx  PkRx PRx (nC  1)  RN (16)
communication channel where the energy expended by a
Where PTx= Energy spent in transmitting 1 Byte of data and
sensing node in transmitting ‘k’ bits per packet can be given as
PRx= Energy spent in the receiving 1 Byte of data.
in [31] [32]: The power utilization of each cluster PWEC can be estimated by
ETx (k , d )  ETx _ elec (k )  ETx _ mp (k , d ) (13) the multiplying the initial energy supplied to each category of

 Eelec  k  E fs  k  d , d  d0
2 a node with the cluster size, that is,
E Tx (k , d )   (14) PWEC ( N )  E0  nC (17)

 E  k  E  k  d 4
, d  d
PWEC (I)  E0 (1   )  nC
elec mp 0
(18)
Emp and Efs are the amplifier parameters of transmission for
multi-path fading model and free-space model respectively PWEC (A)  E0 (1   )  nC (19)
[33]. If the transmitter or receiver expends Eelec amount of The power consumption in each cluster i for a round can be
energy per bit, then to receive a packet of k bits, a sensor node found by estimating the energy cost of a node in both cases,
expends ERx(k) energy given as: i.e. when it acts as a sensing node and as a CH. Consequently,
ERx (k )  ERx _ elec (k )  kEelec (15) P HR (i )  ( Ni  1) PkTx PTx  PkRx PRx  
The energy dissipation is estimated in each round for the
calculation of threshold boundary for CH election. The
( Ni  1) PkRx PRx  ( Ni  1) PkTx PTx  (20)
proposed scheme aims to estimate a threshold energy value to =n(5 Ni  3) PTx
be maintained by all types of nodes to preserve energy for the The sensing node expands the nPTx amount of energy during
network longevity. transmission to the respective CHs. When not transmitting, the
nodes move to sleep mode by switching the radio off till the
IV. PROPOSED WORK next TDMA slot. The CHs consumes n(N-1)PRx energy in the
Once the CH is selected, the sensing nodes join the CH process of data fusion and aggregation. The CH then transmits
according to the information in the ADV message sent by the fused data to the BS expending n(N-1)PTx energy. To
them. Hence for each round, the CHs and clusters keep estimate the threshold value for CH replacement, the
changing. In case, some CHs (either normal, intermediate or information regarding the number of rounds must be known.
advanced) especially those near to the BS, need not use much CountRnd represents the total number of iterations in the
of its energy in transmitting data packets. These CHs can network which can be calculated for the three types of nodes.
continue to transmit with the same group of member nodes in PHR
Count Rnd ( N )  100 (22)
the next round also. But according to conventional SEP, these PWEC ( N )
nodes will not be able to partake in the CH election process
PHR
for the next 1/p epochs. CountRnd (I)  100 (23)
The proposed work introduces a threshold energy value for PWEC (I)
each type of node in the SEP algorithm, now termed as I-SEP PHR
(IoT-SEP). In I-SEP, the threshold energy value decides CountRnd (A)  100 (24)
PWEC (A)
whether the CH and the corresponding cluster should change
or continue transmitting in the next round. After each round, From equation (21) and (22-24), the threshold power level
the CH node residual energy is evaluated. If the residual can be calculated as:
energy is less than the estimated threshold, the CH election PTh( N )  CountRnd ( N ) ( PkTx  PkRx ) PTx (25)
process initiates and new clusters are formed. This controls the PTh(I)  CountRnd (I) ( PkTx  PkRx ) PTx (26)
energy wasted in the unnecessary transfer of routing
information for new CH and also reduces the extra energy PTh(A)  CountRnd (A) ( PkTx  PkRx ) PTx (27)
consumed in new cluster formation. PTh(N), PTh(I) and PTh(A) are the threshold values for normal,
The energy requirement of a sensing node and a CH node will intermediate and advanced node respectively. The introduction
never be the same. CHs perform extra functions like of a threshold value of CH replacement for each type of node

2327-4662 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2019.2940988, IEEE Internet of
Things Journal
Behera et al.: I-SEP: An improved routing protocol for heterogeneous WSN for IoT based Environmental Monitoring

in the modified SEP algorithm improves WSN lifetime by for the proposed I-SEP.
minimizing the total network energy. The algorithm for the
proposed algorithm I-SEP is included in Table I.
TABLE I
I-SEP ALGORITHM
PTh: Threshold for CH election
ERes: Residual energy of existing CH
CH: CH counter
AF: Amplification factor
Gi: Set of non-CH nodes
(a) (b)
begin Fig. 5. Throughput for a = 0.1 in comparison to (a) SEP (b) DEEC
for r=1 to rmax do
CH==0;
calculate p(N),p(I),p(A) by using Eqs. (7),(9),(11);
calculate T(nN), T(nI),T(nA) by using Eqs. (6),(8),(10);
CH=CH+1;
if node==CH then
AF=high;
else
AF=low;
end if
for i = 1 to n do
update ERes for each node by using Eqs. (14-15); (a) (b)
calculate PTh(N) , PTh(I), PTh(A) by using Eqs. (25-27); Fig. 6. Network Lifetime for a = 0.1 in comparison to (a) SEP (b) DEEC
if (ERes < PTh(N) & ERes < PTh(I) & ERes < PTh(A)) then
ni € Gi is selected as new CH; The efficient threshold based CH replacement saves energy
else owing to the dual power level assignment for CH node and
previous CH is retained for next round;
sensing nodes. Introducing a threshold to retain the CH with
end if
end for high residual energy helps to conserve energy for each
end for category of nodes. The nodes are thereby able to communicate
end more data over a longer period of time. When α is kept
constant at 1, the throughput of I-SEP increased by 50% and
V. SIMULATION RESULT & DISCUSSIONS 240% as that of SEP and DEEC respectively. Similarly for α
Simulations were carried out in MATLAB with 100 sensor changed to 2, the maximum throughput further increases by
nodes deployed in a network of 100×100 m2. The BS is 56% and 300% in comparison to SEP and DEEC respectively.
positioned at the center with unlimited energy. The network
parameters used for simulation are enlisted in Table II. To
analyze the behavior of the proposed model in comparison to
SEP [9] and DEEC [17] that follows two level heterogeneity,
the values of ‘a’ and ‘α’ are varied while ‘b’ maintains a
constant value of 0.3. For the first instance α = 1, a = 0.1, for
second case α = 2, and a = 0.1, for third case α = 1, a = 0.2 and
last case α = 2, a = 0.2.
TABLE II (a) (b)
SIMULATION PARAMETERS Fig. 7. Throughput for a = 0.2 in comparison to (a) SEP (b) DEEC
Parameters Value
PTx, PRx (The total energy of the network) 50J Fig. 7 shows the throughput achieved for SEP, I-SEP, and
Emp (Energy dissipation: receiving) 0.0013pJ/bit/m4 DEEC for the case when the percentage of advanced nodes is
Efs (Energy dissipation: free space model) 10pJ/bit/m2 increased to 20%. With more number of advanced nodes, the
Eamp (Energy dissipation: power amplifier) 100pJ/bit/m2 number of CHs from this category of nodes also increases
EDA (Energy dissipation: aggregation) 5nJ/bit which indicate more data transfer. Hence, the maximum
d0 (Reference distance) 87 meters packets send to BS for I-SEP increases considerably by 53%
k (Packet size) 4000 bits and 67% than SEP for α value of 1 and 2 respectively. On a
ggkgg

similar manner, when compared to DEEC, the rise is 32% and

The efficiency of any routing algorithm can be analyzed by 80% for α=1 and α=2 respectively.
estimating the number of data packets communicated to the The instant node starts sensing in the network until the
sink node or BS with minimum packet drop ratio. This is death of the last node is termed as stability period [17]. In
called network throughput. With 10% advanced nodes in the subsequent rounds, the sensor nodes deplete energy and die
network, the throughput is analyzed for SEP and DEEC as eventually. Owing to the availability of sensors with additional
shown in Fig. 5. It can be found the throughput, increases levels of energy, I-SEP performs better in context of lifetime
considerably for I-SEP as compared to both protocols. This than SEP. Since the transmission rate is less in I-SEP, so
improvement results due to the limitation in data transmission

2327-4662 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2019.2940988, IEEE Internet of
Things Journal
IEEE INTERNET OF THINGS JOURNAL, VOL. …, NO. …, MONTH 2019

energy consumption will be less resulting in extended network the CH count as shown in Fig. 10.
lifetime.

(a) (b)

(a) (b)
Fig. 8. Network Lifetime for a = 0.2 in comparison to (a) SEP (b) DEEC

Also, I-SEP considers three categories of nodes whereas

SEP and DEEC have only two types of nodes based on initial
energy. This heterogeneity in energy level contributes to
extend the network longevity to more number of rounds as
shown in Fig. 6 and 8. The lifetime metrics in terms of Last (a) (b)
Node Dead (LND) and First Node Dead (FND) for different Fig. 10. CH count (a) Case 1 (b) Case 2 (c) Case 3 (d) Case 4
values of α and a is shown in Table III.
CHs are responsible for aggregating and routing of data to
TABLE III the BS. Fig. 10 shows the CH formation in each round. In all
LIFETIME METRICS the cases, the CH count is much higher for I-SEP than SEP
a Protocol α FND LND which indicates more data transfer to BS resulting in increased
1 1028 2023
SEP throughput. When 10% of total nodes are advanced nodes, the
2 1125 3348
1 1308 1809 CH count for I-SEP reaches almost 50 and 60 in each round
0.1 DEEC
2 1371 2002 for α value of 1 and 2 respectively as compared to only 20 for
1 2846 5287
I-SEP
2 2876 7600
SEP. The strength starts decreasing beyond 5000 rounds for
1 1141 3289 α=1, and 6000 rounds for α=2 which indicates an
SEP
2 1259 4161 enhancement in network lifespan. Similarly, for a network
1 2997 4579 with 20% advanced nodes, CH count reaches 70 to 80 for I-
0.2 DEEC
2 3388 4938
1 2997 6258 SEP as compared to 20 to 25 for SEP. The CH strength
I-SEP decreases beyond 6000 rounds for α=1, and 7000 rounds for
2 3268 <8000
α=2. The increased CH count is a result of heterogeneity.
Since the energy requirement for CH and sensing nodes Since the network has nodes of three different level of energy
are not the same, I-SEP assigns different power levels for and CHs are chosen from each type of node for equal energy
these categories of nodes. From the table data, it is clear that distribution, the overall network energy depletes at a slower
by increasing the advanced nodes count, the stability period rate. As a result, CHs continue sending data packets to BS for
and network lifetime can be increased, owing to the more rounds, thereby enhancing network performances.
segregation of nodes into different power levels. The average
number of nodes dead in each category is shown in Fig.9. VI. CONCLUSIONS AND FUTURE SCOPE
Stability period, network lifetime and throughput are the key
factors in the design of a routing protocol for WSN. To evaluate
the proposed modified algorithm, extensive simulations have
been conducted to confirm the advantages of implementing the
protocol practically. A distributed routing algorithm is
illustrated that will be well suited for a heterogeneous network
where sensors are deployed with more than one energy level.
The proposed algorithm is a modification of SEP, and well
suited for IoT based environmental monitoring. The simulation
Fig. 9. Dead Nodes in I-SEP
result shows I-SEP outperforms protocols like SEP and DEEC
Since the highest energy level is supplied to advanced in terms of lifetime and throughput for various values of node
nodes, the number of dead nodes is less in this category as density. The protocol also switches energy levels between CH
compared to intermediate and normal nodes. With time, the nodes and member nodes, which also contribute to saving
normal nodes tend to die out at a faster rate. Henceforth, the energy of the network. Implementing the modified algorithm for
intermediate and advanced nodes get elected as CH that a mobile network where nodes move from one point to another
stretches the lifetime to more number of rounds and increases with a constant speed can be explored in the future.

2327-4662 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2019.2940988, IEEE Internet of
Things Journal
Behera et al.: I-SEP: An improved routing protocol for heterogeneous WSN for IoT based Environmental Monitoring

REFERENCES energy-limited heterogeneous fog-supported wireless sensor networks,”

J. Supercomput., vol. 73, no. 2, pp. 733–755, 2017.
[1] M. T. Lazarescu, “Design of a WSN platform for long-term
environmental monitoring for IoT applications,” IEEE J. Emerg. Sel. [23] T. M. Behera, U. C. Samal, and S. K. Mohapatra, “Energy Efficient
Top. Circuits Syst., vol. 3, no. 1, pp. 45–54, 2013. Modified LEACH Protocol for IoT Application,” IET Wirel. Sens. Syst.,
vol. 8, no. 5, pp. 223 – 228, 2018.
[2] J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami, “Internet of
Things (IoT): A vision, architectural elements, and future directions,” [24] M. Ye, C. Li, G. Chen, and J. Wu, “EECS: an energy efficient clustering
Futur. Gener. Comput. Syst., vol. 29, no. 7, pp. 1645–1660, 2013. scheme in wireless sensor networks,” in Performance, Computing, and
Communications Conference, 2005. IPCCC 2005. 24th IEEE
[3] A. A. Bara’a and E. A. Khalil, “A new evolutionary based routing International, 2005, pp. 535–540.
protocol for clustered heterogeneous wireless sensor networks,” Appl.
[25] Y. R. Tsai, “Coverage-preserving routing protocols for randomly
Soft Comput., vol. 12, no. 7, pp. 1950–1957, 2012.
distributed wireless sensor networks,” IEEE Trans. Wirel. Commun.,
[4] S. K. Singh, P. Kumar, and J. P. Singh, “A Survey on Successors of vol. 6, no. 4, pp. 1240–1245, Apr. 2007.
LEACH Protocol,” IEEE Access, vol. 5, pp. 4298–4328, 2017.
[26] D. Mahmood, N. Javaid, S. Mahmood, S. Qureshi, A. M. Memon, and
[5] D. Goyal and M. R. Tripathy, “Routing protocols in wireless sensor T. Zaman, “MODLEACH: a variant of LEACH for WSNs,” in
networks: a survey,” in Advanced Computing & Communication Broadband and Wireless Computing, Communication and Applications
Technologies (ACCT), 2012 Second International Conference on, 2012, (BWCCA), 2013 Eighth International Conference on, 2013, pp. 158–
pp. 474–480. 163.
[6] T. M. Behera, U. C. Samal, and S. K. Mohapatra, “Routing protocols,” [27] S. Dutt, S. Agrawal, and R. Vig, “Cluster-head restricted energy
in Studies in Computational Intelligence, vol. 776, 2019. efficient protocol (CREEP) for routing in heterogeneous wireless sensor
[7] E. A. Khalil and A. A. Bara’a, “Energy-aware evolutionary routing networks,” Wirel. Pers. Commun., vol. 100, no. 4, pp. 1477–1497, 2018.
protocol for dynamic clustering of wireless sensor networks,” Swarm [28] J. Wang, Y. Gao, W. Liu, A. K. Sangaiah, and H.-J. Kim, “An improved
Evol. Comput., vol. 1, no. 4, pp. 195–203, 2011. routing schema with special clustering using PSO algorithm for
[8] S. Wang, J. Yu, M. Atiquzzaman, H. Chen, and L. Ni, “CRPD: a novel heterogeneous wireless sensor network,” Sensors, vol. 19, no. 3, p. 671,
clustering routing protocol for dynamic wireless sensor networks,” Pers. 2019.
Ubiquitous Comput., vol. 22, no. 3, pp. 545–559, 2018. [29] A. P. Abidoye and I. C. Obagbuwa, “Models for integrating wireless
[9] G. Smaragdakis, I. Matta, and A. Bestavros, “SEP: A stable election sensor networks into the Internet of Things,” IET Wirel. Sens. Syst., vol.
protocol for clustered heterogeneous wireless sensor networks,” 2004. 7, no. 3, pp. 65–72, 2017.
[10] A. Whitmore, A. Agarwal, and L. Da Xu, “The Internet of Things-A [30] B. Elbhiri, R. Saadane, D. Aboutajdine, and others, “Developed
survey of topics and trends,” Inf. Syst. Front., vol. 17, no. 2, pp. 261– Distributed Energy-Efficient Clustering (DDEEC) for heterogeneous
274, 2015. wireless sensor networks,” in I/V Communications and Mobile Network
[11] D. M. Swati D., Rajasekhara M., Gandomi A.H., Patan R., “Internet of (ISVC), 2010 5th International Symposium on, 2010, pp. 1–4.
Things Mobile - Air Pollution Monitoring System (IoT-Mobair),” IEEE [31] J. Jia, Z. He, J. Kuang, and C.-X. Chen, “An energy-efficient adaptive
Internet Things J., vol. In press, 2019. clustering routing algorithm for wireless sensor networks,” in Cross
[12] M. J. Handy, M. Haase, and D. Timmermann, “Low energy adaptive Strait Quad-Regional Radio Science and Wireless Technology
clustering hierarchy with deterministic cluster-head selection,” in Mobile Conference (CSQRWC), 2011, 2011, vol. 2, pp. 964–969.
and Wireless Communications Network, 2002. 4th International [32] D. S. Kim and Y. J. Chung, “Self-Organization Routing Protocol
Workshop on, 2002, pp. 368–372. Supporting Mobile Nodes for Wireless Sensor Network,” in Computer
[13] S. Selvakennedy, S. Sinnappan, and Y. Shang, “A biologically-inspired and Computational Sciences, 2006. IMSCCS ’06. First International
clustering protocol for wireless sensor networks,” Comput. Commun., Multi-Symposiums on, 2006, vol. 2, pp. 622–626.
vol. 30, no. 14–15, pp. 2786–2801, 2007. [33] D. Kumar and R. B. Patel, “Multi-hop data communication algorithm for
[14] T. Gao, L. Zhang, Y. Gai, and X. Shan, “Load-Balanced cluster-based clustered wireless sensor networks,” Int. J. Distrib. Sens. networks, vol.
cooperative MIMO transmission for wireless sensor networks,” in 7, no. 1, p. 984795, 2011.
Wireless Communication Systems, 2007. ISWCS 2007. 4th International
Symposium on, 2007, pp. 602–606.
[15] W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrishnan, “An
T. M. Behera got her B.Tech degree in
application-specific protocol architecture for wireless microsensor
networks,” IEEE Trans. Wirel. Commun., vol. 1, no. 4, pp. 660–670, Electronics & Communication. Engineering
2002. from the BPUT in 2007. She received her
M.Tech in Communication System from
[16] R. Saravanakumar, S. G. Susila, and J. Raja, “An energy efficient cluster KIIT University in 2012. She has over 11
based node scheduling protocol for wireless sensor networks,” in Solid- years of teaching experience and is
State and Integrated Circuit Technology (ICSICT), 2010 10th IEEE currently an Assistant Professor in SOEE at
International Conference on, 2010, pp. 2053–2057. KIIT University, Bhubaneswar. Her
[17] L. Qing, Q. Zhu, and M. Wang, “Design of a distributed energy-efficient research area broadly includes WSNs and its application in IoT.
clustering algorithm for heterogeneous wireless sensor networks,”
Comput. Commun., vol. 29, no. 12, pp. 2230–2237, 2006.
[18] T. M. Behera, S. K. Mohapatra, U. C. Samal, M. S. Khan, M.
Dr. S. K. Mohapatra received his Ph.D.
Daneshmand, and A. H. Gandomi, “Residual Energy Based Cluster-head
from NIT, Rourkela, in the year 2016. He is
Selection in WSNs for IoT Application,” IEEE Internet Things J., vol. 6,
no. 3, pp. 5132–5139, 2019. currently an Assistant Professor in SOEE,
KIIT University, Bhubaneswar. His
[19] F. A. Aderohunmu, J. D. Deng, and others, “An enhanced stable election research interests include Modeling and
protocol (sep) for clustered heterogeneous wsn,” Dep. Inf. Sci. Univ. Simulation of Nanoscale Devices and its
Otago, New Zeal., 2009. application in IoT. Energy-efficient WSN,
[20] D. Kumar, T. C. Aseri, and R. B. Patel, “EEHC: Energy efficient Adhoc Networks, Metamaterial absorbers in
heterogeneous clustered scheme for wireless sensor networks,” Comput. THz application, UWB-MIMO and
Commun., vol. 32, no. 4, pp. 662–667, 2009. Reconfigurable Antenna. He has been a part
[21] A. Kashaf, N. Javaid, Z. A. Khan, and I. A. Khan, “TSEP: Threshold- of committee member of various international conferences, Editorial
sensitive stable election protocol for WSNs,” in Frontiers of Information Board Member and Reviewer of international journals. He is a life
Technology (FIT), 2012 10th International Conference on, 2012, pp. member of ISTE, IETE, CSI, OITS, and Senior Member of IEEE.
164–168.
[22] P. G. V. Naranjo, M. Shojafar, H. Mostafaei, Z. Pooranian, and E.
Baccarelli, “P-SEP: A prolong stable election routing algorithm for

2327-4662 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JIOT.2019.2940988, IEEE Internet of
Things Journal
IEEE INTERNET OF THINGS JOURNAL, VOL. …, NO. …, MONTH 2019

Dr. U. C. Samal received his M. Tech. in

Electronic Systems and Communication
Engineering from NIT, Rourkela, in 2006.
He obtained his Ph.D. from the Department
of Electrical Engineering, IIT, Kanpur,
India in 2015. His area of interest lies
in wireless communication, signal
processing, and WSNs. Currently, he is
working as Assistant Professor at KIIT
University, Bhubaneswar, Odisha.

Dr. Mohammad S. Khan is currently an

Assistant Professor of Computing at East
Tennessee State University and the director
of Network Science and Analysis Lab
(NSAL). He received his M.Sc. and Ph.D. in
Computer Science and Computer
Engineering from the University of
Louisville, Kentucky, USA, in 2011 and
2013, respectively. His primary area of
research is in ad-hoc networks, network
tomography, connected vehicles, and
vehicular social networks. He currently serves as Co-Editor-in-Chief of
International Journal of Grid and High-Performance Computing (IJGHPC)
and as an associate editor of IEEE Access. He has been on technical
program committees of various international conferences and technical
reviewer of various international journals in his field. He is a member of
IEEE.

Dr. Mahmoud Daneshmand is Co-Founder

and Professor of Department of Business
Intelligence & Analytics; and Professor of
Department of Computer Science at Stevens
Institute of Technology. He has more than
40 years of Industry & University experience
as Professor, Researcher, Assistant Chief
Scientist, Executive Director, Distinguished
Member of Technical Staff, Technology
Leader, Chairman of Department, and Dean
of School at: Bell Laboratories; AT&T
Shannon Labs–Research; University of California, Berkeley; University of
Texas, Austin; Sharif University of Technology; University of Tehran;
New York University; and Stevens Institute of Technology. He received his
Ph.D. and M.S. degrees in Statistics from the University of California,
Berkeley; M.S. and B.S. degrees in Mathematics from the University of
Tehran.

Prof. Amir H. Gandomi is a Professor of

Data Science at Faculty of Engineering &
Information Technology, University of
Technology Sydney. Prior to joining UTS,
Dr. Gandomi was an Assistant Professor of
Analytics & Information Systems at School
of Business, Stevens Institute of Technology,
and a distinguished research fellow at
BEACON NSF center, Michigan State
University. Prof, Gandomi has published over one hundred and fifty
journal papers and four books. Some of those publications are now among
the hottest papers in the field and collectively have been cited more than
12,000 times (h-index = 54). He has been named as Highly Cited
Researcher (top 1%) and one of the world’s most influential scientific
minds for two consecutive years, 2017 and 2018. Prof. Gandomi is
currently ranked 19th in GP bibliography among more than 12,000
researchers. He has also served as associate editor, editor and guest editor
in several prestigious journals and has delivered several keynote/invited
talks. Dr. Gandomi is part of a NASA technology cluster on Big Data,
Artificial Intelligence, and Machine Learning. His research interests are
global optimization and (big) data mining using machine learning and
evolutionary computations in particular.

2327-4662 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.