Telecommunication Systems (2024) 85:103–124

https://doi.org/10.1007/s11235-023-01068-4

EELCR: energy efficient lifetime aware cluster based routing technique

for wireless sensor networks using optimal clustering and compression
N. Nisha Sulthana1 · M. Duraipandian2

Accepted: 16 October 2023 / Published online: 11 November 2023

© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2023

Abstract
Wireless sensor networks (WSNs) offer a multitude of advantages and find applications across various domains, garnering
substantial research interest. However, a notable drawback in these networks is the energy consumption, which can be
mitigated through compression techniques. Additionally, the limited lifespan of sensor batteries remains a concern. Even
when incorporating renewable energy sources, ensuring energy efficiency in WSNs is imperative. One prevailing issue is the
disregard for spatial data correlation in existing data clustering methods within WSNs. Addressing these challenges necessitates
effective modeling and the acquisition of event source locations in the proposed scheme. In this work, we propose an energy-
efficient lifetime-aware cluster based routing (EELCR) for WSN. In EELCR technique, modified giant trevally optimization
(MGTO) algorithm is introduced for efficient balanced clustering which minimizes energy consumption. An optimal squirrel
search (OSS) algorithm is used to selects the best optimal node, named as cluster head (CH) for prolonging the lifetime in the
sensor networks. Each CH nodes compress clustering data using optimal selective Huffman compression to achieve maximum
compression ratio which overcomes inefficiency of area overhead problem in existing Huffman compression. Furthermore,
we develop a hybrid deep learning technique which combines deep neural network (DNN) with Granular neural network
(GNN) (named as DGNN) to find optimal way for data broadcast from CH to base station (BS). Finally, we assess the
efficacy of the proposed EELCR approach through various simulation scenarios, demonstrating its effectiveness concerning
Quality of Service (QoS) parameters. The outcomes reveal a notable enhancement in our coding scheme, with an average
compression rate improvement of 9.346% when compared to state-of-the-art coding techniques. Furthermore, our proposed
EELCR technique significantly outperforms existing routing methods, exhibiting an average network lifetime improvement
of 51.88% in node density considerations and 52.625% in simulation rounds, respectively.

Keywords Wireless sensor network · Clustering · Huffman compression · EELCR · Cluster head

1 Introduction objects that are able to easily connect to the physical environ-
ment and other devices in order to carry out pre-defined or
Through millions of networked devices, people all over the programmed tasks [2]. Ad hoc network technology known as
world are connected by the Internet. As a result, the Inter- wireless sensor networks (WSNs) was developed more than
net of Things (IoT) has advanced even faster innovative, two decades ago for use in military surveillance [3, 4].WSNs
low-priced devices [1].The future’s most significant global are essentially collected large number of lymphatic vessels
technology platform will be the IoT and cloud comput- with limited resources able to connect to other network nodes
ing, including big data. The IoT generally consists of smart in order to share the data they have gathered. Each node can
function as a repeater or a data combination hub, and its
B N. Nisha Sulthana primary task is to monitor the climate with its internal sen-
[email protected]
sors. By acting as router, each node can send information
1 Department of Computer Science and Engineering, about its neighbors to a receiver or base station. Through a
Vivekanandha College of Technology for Women, network gateway, it can send data to remote servers or pro-
Tiruchengode, Namakkal, Tamilnadu 637205, India cess data locally. Over the course of the past two periods, a
2 Department of Computer Science and Engineering, number of applications for WSNs have been planned, some
Hindusthan Institute of Technology, Coimbatore, Tamilnadu of which include environmental monitoring, healthcare, the
641 032, India

123
104 N. N. Sulthana, M. Duraipandian

smart home, and smart factories [5]. Data prediction, which pieces. Leveraging clustering technique reduces the number
uses predicted values rather than actual values to avoid data of spatial communication exchanges and network routing
transfer, is an effective method for reducing the amount of become more efficient [19]. For efficient data acquisition,
data in a WSN. Accurate measurements for each sampling an industrial WSN (IWSN) uses layered adaptive compres-
period are frequently unnecessary and costly in situations sion (LACD-EDC) [20]. Multilayered network architecture
that occur in the real world. The area of a prediction method is using spatio-temporal correlations, particularly spatial cor-
is provided to reduce the number of measurements provided relations between distinct groups for efficient routing.
by the sensor nodes continuous monitoring [6].
1.1 Our contributions
But an important aspect is that the accuracy of the esti-
mation should not exceed the error specified by the user.
For further enhancement in data aggregation and transfer-
In general, power-saving data transmission is considered to
ring, an energy efficient lifetime aware cluster based routing
be an important issue in WSN applications because power
(EELCR) technique is proposed for WSN. The key contri-
resources are limited and limited, and battery capacity is
butions involved in this paper are organized as follows:
limited [7]. Therefore, we argued that energy consumption
should be controlled in order to significantly increase the life- • MGTO for efficient clustering: The utilization of the
time [8]. On the other hand, most WSN routing algorithms MGTO algorithm in the EELCR technique plays a piv-
require reliable real-time transmission with rich communica- otal role in achieving efficient and balanced clustering. By
tion between receiving nodes such as power saving and QoS minimizing energy consumption through this method, we
based routing. The greatest in WSN and trade-off between ensure that sensor nodes operate optimally, thereby con-
performance and QoS parameters and clustering algorithms tributing to the overall goal of extending network lifetime.
has always been a popular research area. Data transmission • Optimal squirrel search (OSS) for CH Selection: It is a
is a significant portion of energy consumption [9]. The sen- crucial component as it aids in the selection of optimal
sor node sends the collected data directly to the base station. nodes referred to as CH. This selection is based on multiple
So that nodes far from BS consume more power and age design constraints, and it directly impacts the network’s
much faster this greatly reduces the lifespan of the network. longevity by ensuring that CHs are strategically chosen to
Clustering technology reduces the size and distance of trans- perform their role effectively.
mission by selective CH [10, 11].Another way to understand • Optimal selective Huffman compression: Each CH
data reduction is to use compression techniques that reduce employs optimal selective Huffman compression to max-
the data size reduces the amount of data transfer. Returns imize the compression ratio of cluster data. This step
the original data from the compressed data is necessary for reduces the amount of data transmitted through the net-
lossless data compression [12]. In contrast, decompression work, which in turn conserves energy and extends the
of lossy data may result in the loss of some original data fea- network’s operational lifespan.
tures. Despite their excellent data collection performance, • Hybrid DGNN for optimal routing: It plays a pivotal role
WSNs do not require lossy algorithms due to their limited in computing the optimal routes for transmitting data from
resources. CHs to the BS. This optimization is crucial in minimizing
Image density enhancement using K clustering in low- energy expenditure during data transmission, further con-
power embedded devices has been applied to picture WSNs tributing to the network’s energy efficiency and longevity.
[13]. Distributed similarity-based clustering and collapsed
The respite of this paper is prepared as follows: Sect. 2
forwarding (DSCCF) [14] is used to reduce model load and
defines the recent works related to routing techniques for
realize distributed clustering of high-fidelity data. A multi-
WSN. In Sect. 3, we discuss the problem organization and
level lossless data compression method is used to reduce data
system design of our planned routing technique. The detailed
loss between groups. Hierarchical abstract data collection
working process of our proposed routing technique is dis-
clustering algorithm is designed by mathematical analysis
cussed in Sect. 4. Then, we analyzed the simulation results
and simulation environments [15].Noise-tolerant distributed
and compassion in Sect. 5 with statistical measures. Lastly,
image compression (NDIC) design is based principal compo-
we conclude this paper in Sect. 6.
nent analysis (PCA) (NDIC-PCA) delivers compressed data
to CH nodes [16]. Hierarchical adaptive spatial temporal data
compression (HASDC) [17] algorithm uses temporal corre- 2 Related works
lation of real sensory data using discrete cosmos transform
and adaptive threshold compression algorithms. In WSNs, In this segment, we present a brief appraisal based on our
neighborhood indexing sequence (NIS) [18] uses informa- contribution to vigor effectual routing technique for WSNs
tion to assign short codewords to each individual in the using data compression. Table 1 summarizes the research
information string by utilizing the presence of bordering gaps gleaned from recent available studies.

123
 Table 1 Research gap summary

Ref Methodology Clustering CH Compression type Findings Research gaps

[21] Disaster rescue telemedicine NCNBGF Genetic High gain MDC (HGMDC) Compression ratio, PSNR Not deal with temporal and
scheme algorithm spatial relationships
simultaneously
[22] FGAF-CDG GAF Fuzzy logic CDG Energy consumption Limited lifespan due to low
battery power
[23] Low latency PSO ACO ABC QoS parameters Not concentrate on traffic
EELCR: energy efficient lifetime aware cluster based routing …

andenergyefficientrouting transmission load reduction

[24] Multi-level compression and DTS NA Multi-level image compression Compression ratio, throughput Don’t consider the effect of
transmission sensor compression matrix
[25] EMCA-CS Hexagonal Grey Wolf Compressive sensing (CS) Lifetime, energy consumption Area overhead problem is not
cells algorithm solved properly which limits
the system performance
[26] Energy efficient data Slender NA Compressive slender penetrative Throughput, packet loss, SNR, Network deployment affects
transmission and compression penetration etiquette PSNR the node redundancy
[27] Efficient compressive sensing Seed Chain Compressive sensing (CS) Lifetime, energy consumption Higher computational cost
routing estimation construction
[27] Lossless compression delta NA NA Delta encoding and 2occurrence Compression ratio These workers are not
encoding and T-RLE character solving (T-RLE) considered, ie. Local
relationships between sensor
nodes
[28] Compressive sensing-based data Grid-based Multiple Spatio-temporal compression Lifetime, energy consumption Do not talk about power,
gathering HCDG constraints corruption, late business
[29] Routing and compressive sensing MPAR EMPAR Hybrid compressive sensing (CS) Lifetime, energy consumption Not concentrate on energy
efficiency

123
105
106 N. N. Sulthana, M. Duraipandian

Sheeja et al. [21] have proposed technique for WSNs to data, and Gray Wolf-based algorithms are used to design
plan recovery telemedicine will help the network last longer the best route for transmitting compressed data to the BS.
and save energy. These systems increase the resilience of The process of creating matrices can be made easier by uti-
the network by managing critical power nodes, clustering, lizing the CSMO-GWO algorithm. The reenactment results
and compressing medical data. The method’s performance is demonstrate that EMCA-CS outperforms previous bench-
evaluated through simulations with a variety of parameters. mark methods by increasing WSN lifetime, decreasing power
The study’s findings show that the method can significantly consumption, and decreasing the root mean squared error.
extend a network’s lifespan. Ghaderi et al. [22] have proposed Reddy et al. [26] have proposed the protocol for abstract
a compressive data gathering (CDG)-based Fuzzy based rout- thin routing has been developed for efficient data transfer in
ing protocol. The sensor region is organized geologically and WSNs, which provides high data rates by reducing interfer-
partitioned into virtual hexagonal lattice cells.CH sensors are ence and attenuation. The results showed that this method
chosen for each grid cell using a CDG-based fuzzy CC logic can provide a data path that is safe from intrusion and uses
algorithm. The simulation results show that the FGAF-CDG less energy. Aziz et al. [27] have proposed a compression-
method performs better than other competing GAF methods. sensitive routing scheme for energy-efficient data collection
This design outperforms the fuzzy logic based two-level geo- in IoT-based WSNs. The best measurement matrix is selected
graphic routing (FTGAF-HEX) approach in terms of power using the seed prediction algorithm to select the best predic-
consumption by approximately 50%, according to measure- tion seed. During the transmission and reception process,
ments of sensor area. Singh et al. [23] have proposed a particle the network nodes are arranged using the circuit generation
swarm optimization (PSO), Artificial Bee colony (ABC) and algorithm.
ant colony optimization (ACO) algorithm to improving sen- Saidani et al. [27] have proposed a two-event write reso-
sor network performance using receiver node positioning lution (T-RLE) algorithm and delta encoding-based lossless
path construction and computational methods inspired by compression technique. T-RLE is an optimization of the RLE
nature. Probability propagation uses opportunistic coding to procedure aimed at improving the compression ratio. They
reduce the number of transitions. Therefore, the application used real-life data from a range of sensor projects to evaluate
combines three systems of each to significantly improve data the effectiveness of the T-RLE, yielding good results com-
transmission. The location of the receiving node is reso- pared to other conventional methods. Ghaderi et al. [28] have
lute by the PSO and the devices, and the proposed cluster investigated a various data collection techniques in CS-based
is strong-minded by the minimum wiener spanning tree, WSNs which used to improve the QoS of overall system
which is further enhanced by ABC and opportunistic packet performance. To evaluate this model Analysis and compari-
scrambling before sending it to the neighbors. The results son of the energy consumption of the two main acquisition
effectively found that previous methods provided signifi- techniques used in CDG and HSS, CC-based data acquisi-
cant package delivery speeds. JamunaRani et al. [24] have tion systems. The total energy consumption of the models
proposed efficient multi-level transform-based image com- was analyzed using the data collection method. Consider
pression technique. It checks the network conditions and lists signal sampling. Jari et al. [29] have proposed optimiza-
the possible routes to reach the destination. This method spec- tion technique to solve clustering, multi-pool placement, and
ifies a list of available spacers for each generated path. This load-balanced unicast routing issues by combining multi-
way the flow is calculated. Choose a route by calculating the sink placement and anycast routing (MPAR) and extended
ratio of bandwidth usage and power consumption as a single (EMPAR). Ant swarm optimization algorithm generates
data transfer for each. This method determines the amount high-level unicast routing tree for each cluster. Therefore,
of payload required for the selected route based on the trust compared to the cluster average MPAR algorithm, EMPAR
degree. This method compresses the bandwidth and sends is better than the MPAR because of the selection strategy of
it to the destination. This method delivers effective video the theme CH algorithm. Energy efficient CS data collection
compression depending on the desired compression level and EMPAR reduces energy consumption by 5.98% and 12.20%.
improves the QoS quality of WSN.
Aziz et al. [25] have proposed well-organized multi-
hop group-based coordination arrangement using hybrid CS 3 Problem statement
(EMCA-CS) for different WSNs on IoT. EMCA-CS effec-
tively increases network lifetime and reduces reconfiguration 3.1 Research gaps
errors between CH and routing BS. EMCA-CS divides the
area into a number of hexagonal cells and selects a CH node Molk et al. [30] have proposed a data transmission is founded
from each group based on a number of factors. To collect on fuzzy clustering and lossless cascode Huffman density.
cluster data, each CH makes use of hybrid CS strategy. A The facts of the cluster master bulge and the traceability
reverse recovery algorithm is used to recover the original records of each node are compared during inter-node data

123
EELCR: energy efficient lifetime aware cluster based routing … 107

transfer. After the excess data has been removed, the encoded ing and compression. The main objectives of our proposed
data is sent to the host node. Fluffy rationale is utilized to EELCR technique are given as follows:
choose CH hubs in light of the hub’s excess energy and dis-
tance from the Mamdani and Takagi–Sugeno getting hubs.
Using a type 1 fuzzy system, CH selectivity was assessed 1. The joint optimization of data compression and transfer-
using three characteristics: number of enabled nodes, incom- ring is used to enhance the energy efficiency and nets
ing packets, CH, optimum delivery rate, other LEEC protocol lifetime
parameters, and lifetime. Network work and main power 2. To utilize the multiple design constraints to selects the
Increases data transfer rate by 40% without losing power. Run CH for each cluster which optimizes through searching
Length Coding (RLE) and Adaptive Huffman Coding (AHE) algorithm
data density methods applied in sensor nodes. H-RLEAHE 3. To reduce the amount of data that is redundant, choose
[31] can balance compression ratio and energy consump- the best compression method that keeps only the most
tion. It starts with summarizing using the RLE technique, important data packets for the reconstruction.
which is based on statistics from data sources and aggre- 4. Hybrid deep learning process is used to choose the best
gates data. Additionally, compressed data are used as the route for CHs to send compressed data to BS.
AHE algorithm’s input. A decompression algorithm is used
to compress the beaten data before it is sent to the BS.
Energy conservation is frequently a major concern in
WSNs due to the battery-powered sensor nodes. Because
radio communication uses energy, lowering the cost of trans- 3.2 Network design of proposed EELCR technique
mission can extend the lifespan of the sensor node [21].
Message is the primary energy overwhelming component The network model of our proposed EELCR technology
of the sensor nodes. Nodes can save energy by reducing the is shown in Fig. 1, where the sensor nodes are spread out
amount of data they transmit. However, the limited memory across the sensor field at random. Each sensor node trans-
and communication bandwidth offered by WSN platforms mits local data to the CH during sensing, and the BS then
is a major drawback in such complex applications [22, 23]. multicasts the data via the gateway node; it depends on how
However, data abstraction alleviates such shortcomings, the far CH and BS are from one another. The issue is resolved
lack of objective and thorough evaluation of relative methods, by compressing the data before transferring it. In wireless
and even more specific methods for specific demand needs. sensor networks, energy consumption has emerged as a prob-
Records compression is one approach that could be used to lem. This is because of the moves these organizations face
cut down on the amount of data that is sent between WSN because of little sensor hubs, restricted memory, little battery
nodes because the communication unit of a WSN uses a lot limit, and restricted handling power and transmission capac-
of energy [24]. Recently, many compression techniques have ity. Compression of data is used before transmission from
been proposed with various WSN implementations. How- sensor nodes to compressed network nodes to improve energy
ever, recent research to reduce network traffic and extend consumption and network performance. There have been
network life has been dominated by methods based on net- numerous proposals for data compression, but the majority
work data compression [25]. On the other hand, the power of them are not suitable for sensor nodes because of memory
consumption of WSN networks has always been a major chal- constraints, energy consumption, and processing speed.
lenge in the design and implementation of these networks. Based on recent research findings, network clustering is
Although WSN nodes use time-consuming abstraction tech- a widely popular topology technique in which CHs and
niques, this causes the nodes to have limited resources to nodes work as a group of nodes to manage more objec-
understand and communicate other functions [26, 27]. The tive tasks. This clustering design mainly emphasizes efficient
compact approach reduces the likelihood that these nodes consumption performance to achieve efficient and reliable
will enter deep sleep, which is essential for improving energy networks. Data aggregation and bonding processes mainly
efficiency. Although most of these algorithms rely on data reduce total network energy by eliminating some transmit-
correlation from sensor nodes, they can compress data to ted data and increasing network scalability and lifetime. To
high levels at low cost. To overcome above gathered prob- maintain control over the dynamic behavior of the network,
lems, an energy efficient lifetime aware cluster based routing it is necessary to use efficient mechanisms to establish a reli-
(EELCR) technique is future for WSN using optimal cluster- able and efficient sensor network. CH selection is usually
done by selecting the desired node near the desired location.
Traditional location-based CH selection adds computational
complexity to find relevant sensor nodes, resulting in poor
selection accuracy and duplicate node selection.

123
108 N. N. Sulthana, M. Duraipandian

Fig. 1 Network model of our proposed EELCR technique

4 Proposed methodology defense automatic and random features are common in hid-
den objects. Conclusion six common traits of monsters of
In this segment, we describes the working procedure of our different ages inspire this model. During each phase, giants
planned energy efficient lifetime aware cluster based rout- move as follows:
ing (EELCR) technique which consists following process
iter , age iter , age (iter −1), age
are clustering, CH selection, data compression and routing. pM  vel M + pM , age  α, β, γ δ
(1)

4.1 Cluster formation where p M

iter , age
denotes the M-th describes the risk threshold
iter , age
for each giant, i.e. vel M is the velocity vector of the
In clustering, each node sends data from the nearest route to (iter −1), age
all neighbors until the data reaches the destination node or giant, its age, and p M is the number of current itera-
the destination node reaches the data despite the fact that tions. Given the accompanying qualities, each pattern of the
the transmission of datagrams varies in some ways. The example can be addressed as a movement vector representing
algorithm’s fundamental concept is that the night process a different age.
is carried out continuously and sub-intelligence nodes are
,α ,α ,α
randomly selected with equal probability and sub-intelligent vel iter
M  gra iter
M + de f meciter
M (2)
nodes jump according to their good probabilities. A network
aggregate must the cluster process need to be configured peri- iter , β iter , β iter , β
,β iter , β
odically during its execution. Each round is divided into two vel M  gra M + h iter
m + socm + de f mec M
phases: cluster setup and data transfer phase. As the cluster (3)
reorganization process consumes more energy, in general,
the migration period is longer than the stock cluster period iter , γ iter , γ iter , γ iter , γ
to reduce the overhead of the cluster system. In this paper, a vel M  gra M + hm + socm
giant path optimization (MGTO) algorithm is designed effi- iter , γ iter , γ iter , β (4)
+ imtm + r om + de f mec M
cient symmetric clustering with low energy consumption.
The basis of this study is the behavior patterns of monsters in
,δ ,δ
their natural habitats. Herding, hierarchy, sociality, mimicry, vel iter
M  gra iter
M + imtmiter , δ + r om
iter , δ
(5)

123
EELCR: energy efficient lifetime aware cluster based routing … 109

iter , β
These are the main stages of personal gra M and social cycle In the Jth iteration this large face block is denoted by
iter , β w. Giants N is the number of giants and the age of each giant.
intelligence socm of monsters. Monsters they are Herbi-
vores eat plants and other materials. They feed for 16 to 20 h a From the estimation of these factors, t coefficients for γ and β
day and rest for a few hours continuously feeding is the term coefficients are obtained. Monsters get positive and negative
iter , β conditions when describing themselves and find better pas-
for this type of continuous feeding de f mec M . All major
plant components are manufactured using MGTO technol- tures. The operation time of cluster formation is described
ogy. In terms of G coefficient, each is huge in a specific area. by Algorithm 1 MGTO.
Monsters of any age graze throughout their lives.
iter , age (iter −1)
gra M  giter (low+ R ∗upp)( p M ), age  α, β, γ δ
(6)

iter , age (iter −1)

gM  wg × (g M ), (7)

iter , age
Here,gra M represents the movement parameter of the j-
th giant and describes the grazing ability of the corresponding
giant. The grazing variable decreases linearly in i will repeat
the details. The values of the variable "R" can be anything
from 0 to 1, and the "lower" and "upper" limits of the diet
field.0.95 and 1.05 are recommended. The h coefficient is set
to 1.5 for each age group.They follow the leader, which is
what people usually do. According to the rule of dominance,
older classes or groups are responsible for control of wild Algorithm 1 Cluster formation using MGTO
giants. Therefore, studies have found that giants follow a
stepwise law between β and γ.
4.2 CH selection
iter , age iter , age (iter −1) (iter −1)
hM  HM ( pL B H − pM ), (8)
Cluster head (CH) selection is an important step in clus-
iter , age (−1+iter ), age tering because it is primarily responsible for efficient data
hM  HM × wg (9)
transmission and data aggregation in WSNs. Recently, CH
iter , age selection has been the focus of many papers in the literature
where h M is the position of the ideal giant with variable because accurate CH selection can improve overall network
(iter −1)
speed. The p L B H value represents the ideal giant state. lifetime and reliability. In the heuristic MET approach, CH
Monsters need social interaction and can associate with other selection is based on established selection criteria for spe-
animals. Since wild giants are vulnerable to being hunted cific applications contexts. In this work, an optimal squirrel
by predators that live in groups Provide security. Diversity search (OSS) algorithm is used to selects the best optimal
improves their chances of survival and makes escape eas- node called as CH for prolonging the lifetime in the sen-
ier. Monsters often fight because of their social nature and sor networks by utilizing a multiple design constraints are
personality due to anger. Some rats enjoy being with other energy consumption, distance between sensor node to base
animals such as cows and goats; don’t like to be alone. This station, congestion rate and received signal strength. A set
behavior is represented by the s factor as a movement toward of heterogeneous nodes S  [S1,S2,S3,…,Sn] with com-
the mean position of the other giants. Between the ages of 5 munication radius CR  [CR1,CR2,CR3,…CRn] associated
and 15, giants mostly prefer to be with livestock, as shown with node numbers and initial battery capacities E  [E1,E2,
by the following formulas: E3,..,En}.Energy feasting is an significant issue in the design
   of routing protocols. Due to the internal location of the sen-
iter , age iter , age 1 n (−1+ier ) sor node, it is difficult to charge or replace the batteries of
Soc M  Soc M pi age  β, γ
n i1 the sensor nodes in these networks. The main answer for
(10) supplanting the inward sensor network is minor or complex
iter , age (−1+iter ), age medical procedure. While working out the energy level, the
soc M  Soc M × wsoc (11)
sensor hub with the highest power will be the next forward
iter , age data propagator. The energy level of sensor node is described
where Soc M the vector of social mobility given by the
iter , age
j-th giant soc M is described. The reduced group in each

123
110 N. N. Sulthana, M. Duraipandian

as follows: 
l 
m
C dist ( p)  || d j − Sinki ||+||Sinki −At || (21)
T nr −energy  T n B−energy − T n c−energy (12) j1 i1


l 
m
where, T nr −energy presents the senor node residual energy, C dist (q)  || d j − di || (22)
T n B−energy average energy and C-energy consumed energy. j1 i1

T n c−energy  S Ab × T n s−energy + r Aa × T nr −energy Specifies the coverage region, where the coverage must
(13) be in the interval [0, 1]. Plummeting the amount of nodes in
net also reduces the resulting latency.
where S Ab , r Aa are transmitted and received bits in sen-
sors also, T n s−energy mean the all out worth of received and M
transmitted energy. max(Sinki )
i1
C REG  (23)
l
T n S−energy  T n Sn−energy + T n S B−energy × dis (14)
The Received signal strength (RSS) is the customary han-
T nr −energy  T n Sn−energy (15) dover choice model in all current even handover systems.
RSS is the most by and large used measurement since it
where T n Sn−energy and T n S B−energy present the energy is easy to evaluate and directly associated with organization
required for transmitter. The energy means the transmitter quality. There is a close by connection between’s RSS scruti-
enhancer with required energy Sn and energy SB, and d is nizing and the partition from the flexible terminal to its place
the distance between the following forwarder and the source of association.
node. The mathematical model of the distance parameter
F dist ( p) a node represents the common distance between Ntotal  Y (m, C) + Z (m) (24)
two nodes common nodes of the CH, CH, and AS F dist (q)
network. The value must be in the range [0, 1]. where Y (m, C) and Z (m) are energy utilization of com-
municating and getting hub.
F dist ( p)
F dist  (16) 
F dist (q) m × T n Sn−energy + m × η F1 × C 2 ; i f C < C0
Y (m, C) 
m × T n Sn−energy + m × ηnq × C 4 ; i f C ≥ C0

l 
m
F dist ( p)  || d j − C Hi ||+||C Hi − At || (17) (25)
j1 i1
Z (m)  m × T n Sn−energy (26)

l 
m
F dist (q)  || d j − di || (18) The thought about all assaults are answerable for the
j1 i1
energy utilization. Not entirely settled by the distance and
The data transmission node delay is specified, where the transmission energy, in the event that the hub communicates
delay value must fall within the [0, 1] range. As the number bundle with energy, the nodes got signal asset RSS, with the
of cluster nodes decreases, latency decreases significantly. distance of D, can be imparted as shadows:
Represents the CH count value in WSN and represents the Y (m, C)
total node count class value. RSS  + Yx, x1 /x2 (27)
4πCi2
M
max(C Hi ) At last, using above design metrics is used to formulate
i1
F delay  (19) the trust degree (TD) as follows:
l

The numerical model of congestion rate, where C dist ( p) T Di  T nr −energy + F delay + C R E G + RSS ; i  1, . . . , n
shows the coldness between the typical device nodes and sink (28)
bulge of network and C dist (q) indicates the distance among
two ordinary nodes. The worth of C dist ( p) ought to be among The Optimal Squirrel Search (OSS) algorithm is a novel
in the scope of [0, 1]. biologically inspired optimization algorithm that it can be
used to solve optimization issues in unimodal, multimodal,
C dist ( p) and multidimensional settings. However, like other swarm
C dist  (20) intelligence-based algorithms, the OSS algorithm also has
C dist (q)

123
EELCR: energy efficient lifetime aware cluster based routing … 111

its drawbacks. n is randomly generated to identify individual through g remain in their updated states, and all individuals
squirrels: travel through fa without changing conditions as follows:

f t j  f sl + rand(1, D) × ( f tu − f sl), ( j  1, 2, ...N )

(29) f t new  f tl + levy(N ) × ( f tu f tl ) (36)

where f t j For a Squirrel, The problem’s dimension is rep-

The Lévy circulation can be designed as follows:
resented by the 1D random matrix rand(1, d), which spans
the range [0, 1]. The GC weight of this work is 1.9. gd is
calculated as a random sliding distance follows: Ra × ξ
levy(y)  0.01 × 1
(37)
gh |Ra | β
Dh  (30)
tan(φ) × t f
where Rb and Ra normal random numbers in [0,1] and β
Here gh the default value is 8; t f default value is 18; The
are distributed with a certain value 0.5. In the original OSS
slip angle can be calculated by using the formula for the
algorithm, current generation optimal and suboptimal solu-
character Tan ().
tions direct the search agent so that the location is updated.
d Nevertheless, ignoring the potential of the surrounding popu-
tan(φ)  (31) lation leads to a poor chance of finding universal energy. The
l
OSS algorithm is based on the Advanced Whale Optimiza-
Here d gives the force and power of life, which can be tion (IWO) algorithm, which introduces the behavior of weed
understood by formulations generation to direct the development of potential solutions.
First, according to the reproductive capacity of the squir-
1 rel, it determines the ability of the squirrel to reproduce
d (32)
2ρv 2 tcd and produce offspring best habitat conditions. For example,
1 squirrels have abundant food sources in hickory and horn-
l (33) beam trees, which are ideal for survival and reproduction.
2ρv 2 tcl
The fit value determines the number of iterations. The par-
where (ρ  1.204 kgm−3 ) is the air den- ent squirrel’s fitness correlates linearly with the number of
−1 offspring produced.
sity (V = 5.25 ms ), is the squirrel’s rate of
sliding,(S = 154 cm2 ) is the body’s surface area,(CD = 0.6)
Between 0.675 and 1.5, the coefficient of friction, or CL, can F( f t j ) − Fmin
be any number. The entire population needs to be for the OSS nt  (Tmax − N ) + N ( j  1, 2, ....N )
Fmax − Fmin
algorithm to work. dormant, meaning that all individuals (38)
are new. As all people renew, times change calculated as
follows:
 The F( f t j ) is fitness value of the parent squirrel Fmax and
D Fmin Large number of people large and small size. After that,
Tc 
s
( f Tg,s k − f tbi,
s )2
k i  1, 2, ..., n f t
K 1 the mean and standard deviation of the number of offspring
(34) in RSSA are distributed at random equivalent to The mean
10E −6 difference between the random number decreases with each
Tmin  (35) iteration. The standard deviation of the population distribu-
(365)s/(S/2.5)
tion f tσI ter is defined as:
The current number of iterations is s, and T is the maxi-
mum number of iterations. Winter will give way to summer if (I termax − I ter )g
the temperature is low. All people change over time traveling f tσI ter  g (σmax − σmin ) + σmin (39)
I termax

where σmax is The non-linear covariance index is g, along

with the maximum, minimum, and maximum number of
iterations, as well as the standard deviation. Algorithm 2 pro-
vides a description of the CH selection procedure based on
the OSS algorithm.

123
112 N. N. Sulthana, M. Duraipandian

i.e. e  {0, 1}, e ∈ 1 A continuous state transition and E 

0 is the stopping state stable P jiE (s) e ∈ E, i, j ∈ S, τ denotes
the time interval from state i to state j. The level transition of
an event trigger is a gain set r  {R(T , E) }

G(t), E  1, t ∈ T
R(T , E)  (40)
0, other s

In the above formula G(t), E  0 is the gain when the

system is at s and is triggered by an event. For the semi-
Markov Stochastic optimization techniques aim to produce
similar results in a model described by Markov processes
with different possible outcomes states that maximize the set
of transition state gains from the event-causing state. Calcu-
late the stochastic optimization strategy’s anticipated profit:
Algorithm 2 CH selection using OSS ∞

g π (t)  etπ E −αs R(t N , D N +1 (TN )) (41)
N 0
4.3 Data compression
Among them, DN (N = 1, 2, ...) denotes the specific trig-
After CH selection, each CH nodes compress clustering data gering event, π = { d1 , d2 , ...} denotes the event frequency,
using optimal selective Huffman compression to achieve and the penalty factor α > 0. The expected gain when the
maximum compression ratio which overcomes inefficiency initial condition is optimal is defined as:
of area overhead problem in existing Huffman compression.
Consequently, this study proposes a method for data fusion g ∗ (T )  sup g π (t) (42)
π
and compression to maximize the use of available sensors
Original quality is preserved. The primary objective is to The appropriate Then the triggering event is received. The
reduce this unnecessary data by removing some data files and better the expected profit, the better the similar amounts pro-
rebuilding the main files. Sensor data packets are lost during cessing events is π ∗ obtained as follows:
data aggregation, which slows transmission over the cellular
link and reduces packet collisions. Aggregated data storage g(t)  max G(t) + g( ), e[g(i)E −αs ] (43)
redundancy this is reduced by data compression, which is
⎧ ⎫
induced by the storage space used by the sensor and trans- ⎨  ⎬
port nodes. Since this discovery, the network has been around g(s)1  max G(t) + g( ), pti0 (α)g(i) (44)
forever increased significantly. ⎩ ⎭
i≥t
Then, an overview of the optimal Huffman algorithm-
based stochastic optimization technique is provided. The data where t goes to any non - state in state set t, G(t) + g( )
indicate that distinct conditions are brought on by distinct embodies the gain gained when the state change event occurs
events. The system’s current state is determined by the count- at state s, and optimal gain g ∗ ≥ 0 is the minimum solution.
able number of objects calculated. Let be the arrangement of Therefore, g ∗ ( )  0 G ∗ ( )  0, there is optimal static
uncountable states and e be the arrangement of countable solution set D∗ = { D, D, · · · }
occasions to such an extent that when the nth occasion hap- ⎧ ⎫
pens, the time span t ∈ T from the point at which the nth ⎪
⎪ ⎪
⎪
⎨  ⎬
event to Sn of the N + 1st event and the induced αMN+1 D(t)  arg max E G(t) + (1 − E), 0 ∗
pti (α)g (i) (45)
state will occur randomly. The state number eN+1 is inde- ⎪
⎪ ⎪
⎪
⎩ i≥t ⎭
pendent historical conditions and timing of the Nth event. E∈e
With based on this assumption, the four-part definition of
the Markov process model computation stage computational where s fits to any non - state in state set t,d( )  0. Then
stimulus events, the transition function of the state of the there are choice criteria for D∗ = { D, D, · · · } :
stimulus, and the set of transition states P jiE (s) of the stim- 
ulus. Among them, the set of stimulus events represent two 1, t < t∗
D(t)  (46)
cases of continuous state transition or stop state transition, 0, t ≥ t∗

123
EELCR: energy efficient lifetime aware cluster based routing … 113

Therefore, the search for the optimal expected profit trans-

forms into a search t ∗ for the state solution. The gain that
results from the stop state transition being carried out at the
right moment is:

X N (t N , s N )  G(t N )E −αs N (47)

The predictable stop state change event for any state is

a gain for execution incomplete, and since α > 0, there is
XN→∞  (tn , sn ) with likelihood 1 equal to zero.


G(t) ≥ pti0 (α)g(i) (48) Algorithm 3 Data compression using optimal selective Huffman
i≥ compression

When the anticipated gain is as shown, 4.4 Optimal routing for data transmission

The WSN relay protocol is responsible for generating mutual

⎧
⎨ information for autonomous nodes. Due to specific applica-
g̃(t)  lti (α)G(i + t ∗ − 1), t < t ∗ G(t), t ≥ t ∗ (49) tion characteristics and limited support of WSN protocols,
⎩ designing a WSN streaming protocol requires adapting exist-
i≥1
ing streaming protocols to address, transform, manage, and
 implement a set of machine learning tasks. Suitable for smart
p 0ji (α), 1 ≤ j ≤ i ≤ t∗ battery pack use. By analyzing the implementation of the
A ji  (50)
0, other s wide data comparison method for battery, WSN network
mode is mainly used for battery storage and auxiliary box
padding mode, following the characteristics of WSN node
c ji  P ji0 (α), 1 ≤ j < t ∗ , i ≥ t ∗ (51) nodes. A hybrid deep learning technique for optimal routing
is proposed here, which combines a granular neural network
(GNN), also known as a DGNN, and a deep neural network
The final best solution is obtainable: (DNN) to determine the best way to transmit data from CH
to BS. The deep learning framework’s goal is to find the best
⎧ ⎫ model parameters, which are defined as:
⎨  ⎬
t ∗  min t ≥ 1 : G(t) ≥ pti0 (α)g(i) (52)
⎩ ⎭ θ ∗  arg Min l(θ ) (53)
i≥t θ

where θ ∗ are the parameters of An improved model based on

In view of the above rationale, the estimation equation to weighted gradient optimization. Six neural network layers
ascertain the ideal state s ∗ is acquired, and when it changes make up the deep neural network (DNN) model that we pro-
to the stop state, it gets the best benefit. When applying the pose including noise emission layer, GNN layer, maximum
advanced Huffman coding algorithm to the data that has been layer, hidden layer, and effect layer. The term exponent (E)
sent to the base station node, when the order of the data means the vector name. In the granular layer, each noise vec-
changes, the frequency of each piece of data in the Huffman tor’s features are extracted by a feature extractor (L); He can
coding tree must be concatenated. The frequency occurs in
accordance with the stochastic optimization step that came
before the data is added. The stochastic method of plan-
ning technique transforms the Huffman index into a dynamic
range needed to optimize. Algorithm 3 is described the pro-
cess of data compression using optimal selective Huffman
compression.

123
114 N. N. Sulthana, M. Duraipandian

read the sentence in any direction. Granular Beam is a fea- The DGNN A table in the framework of our project
ture extraction tool that uses a fixed window size to compress predicts the probabilities of various sensitivity parameters,
the DGNN’s hidden lines and capture key specific features in which are defined as follows:
a single view. The need commitment capability (P) chooses
the main part of every improvement component to get the x̂  σ (wout g F1 + aout ) (63)
main part. A DGNN stowed away layer is utilized to apply
the greatest secret component to the new secret vector. Last
where x̂ ∈ R l the expression label is the prediction probabil-
but not least, the evaluation layer ought to be used to predict
ity, r is the correlation parameter, and σ is the function of a
the likelihood of various emotion labels. The design of the
sigmoid. In the previous section, we learned how to construct
granular layer is calculated as follows:
a DGNN model and how to incorporate some DGNN param-
eters.As a result, we train on the classification of patterns
Fsl  σ (w f [h ls−1 , ws ] + a F ) (54)
containing multiple emotion labels using the log-likelihood
approach. In this instance,a DGNN model is used to train a
Osl  σ (wo [h ls−1 , ws ] + ao ) (55) separate binary classifier for each label of the multi-sensory
label.The proposed neural network model’s parameters are
C̃sl  tan g(wc [h ls−1 , ws ] + ac ) (56) represented by the negative log likelihood function L:

csl  Fsl ⊗ cs−1

l
+ (1 − jsl ) ⊗ c̃sl (57) 
N
l(θ )  − x ( j) log x̂ ( j) + (1 − x ( j) ) log[1 − x̂ ( j) ] (64)
j1
gsl  ols ⊗ tan g(csl ) (58)
where θ is all the parameters in our DGNN model. Com-
where [h ls−1 , ws ] ∈ r wt+gt is a link vector of the before dis- pared with traditional neural networks, DGNN has a feature
guised state h ls−1 and using the current word as input es , w F , extraction tool consisting of an optimal rule layer and an
w j , wo , wt ∈ r gs(gs+wt) , and a F , a j , ao , at ∈ r gs are vari- undersampling layer. From a data analysis perspective, our
ables that can be learned Both are element-wise development streaming framework is designed to perform inference oper-
and a logistic sigmoid function, respectively. The tan g is an ations on each cell as it is first captured without interpolation
activation job. DGNN layer because different expressiveness to create a data processing set. Compared to the traditional
can be captured by different filter sizes features. A window performance-optimized DGNN model, which often operates
of hidden features, for instance, yields an intricate hidden on blocks of input data, the single data inference operation
feature.h lj: j+k−1 . avoids the need for a block buffer, so it is more suitable for
our latency-optimized design.
C cj  σ (wcnn · h lj: j+k−1 + acnn ) (59)
n a + (n a − 1) + (n a − 2)... + 2 + 1 1
simg  (n a + 1)simg
where C cj ∈ r is a convoluted feature. σ is a rectified linear na 2
unit (ReLU) function. (65)

C c  [C1c , C2c , ....C S−K

c
−1 ] (60) The no-pipeline congestion constraint implies that after
the processing delay of stage j, all output feature maps are
with C c ∈ r S−K +1 . We apply the maximum overtime pool- served by stage j + 1 in the same time period. Therefore, q j−1
ing function to take the most important hidden features as a the q j following equation is satisfied:
feature of each relevant filter. The framework uses different
filters to capture multiple filter functions of DGNN Max- C j−1 G j−1 w j−1 CjG jwj
Pooling:  (66)
q j−1 qj
g c  [Ĉ1 , Ĉ2 , ....Ĉ M ] (61)
Note that both G j and w j were reduced to (1/tj) of their
Here he refers to the latent traits that pervade all lines of original size G j−1 , w j−1 after passing through stage j. We
the individual. We use a fully connected neural network to obtained optimal solution to describe the relation between
extract the hidden function from the hidden DGNN. function q j−1 and q j
to predict emotion labels.
Cj 1
q j  q j−1 . · 2 (67)
g F1
 w F1 g + a F1
c
(62) C j−1 S j

123
EELCR: energy efficient lifetime aware cluster based routing … 115

According to the specification of the DGNN model, the Table 2 Simulation parameters
data rate of each stage of the pipeline is iteratively calculated
Parameters Values
using the input pixel rate of the DGNN, all proportional to
q. Therefore, we used qj as the hardware parallel indicator. Number of sensor nodes 100, 200, 300, 400 and 500
Based on the value of qj, we can estimate the amount of Network size 100 × 100 m2
reconfigurable hardware resources required to implement a
Channel type Bidirectional
particular DGNN. For example, the activation coefficient of a
Standard IEEE 802.11
convolutional level with weight j is the number of multipliers.
Propagation model Two-way propagation
 2 model
Kj
C j−1 q j C j (68) Traffic type CBR
tj
Maximum communication range 100 m
where K j and t j are the kernel size and stride of the Granu- Average sensor initial energy 2J
lar layer. Since q j was proportional to q j , overall hardware Data packet size 6400 bits
resources usage which is proportional to q j . Commonly used Control packet size 100bytes
activation functions include sigmoid function. Algorithm 4 Number of simulation rounds 500, 1000, 1500, 2000 and
describes the working process involved in the optimal path 2500
selection for routing using DGNN technique. Simulation time 200 s

on a 500-sensor node network set with a 100 × 100 m area.

All of the sensor nodes are mobile and have GPS tracking
capabilities. At position (100, 100), the BS is outside the
network. The simulation environment uses a random number
of sensor nodes. The network area has 100, 200, 300, 400,
and 500 nodes in total. The initial energy consumption of
each sensor node in the network is 2 J. The communication
range of each sensor node is 100 m. Data packet size is 6400
packets are 100 bytes in size. The CH node stores information
based on the size of data packets it collects in a variable area;
6400 bits make up its size. The power consumption of the
transmitting and receiving nodes is 50 nJ/bit and 5 nJ/bit,
respectively. We consider a simulation run time of 200 s to
run the entire simulation. We carried out the simulation five
times (cycles 500, 1000, 1500, 2000, and 2,500) in order to
Algorithm 4 Optimal routing using DGNN technique
evaluate the network’s performance by altering the number
of CH in sensor nodes. The results are calculated by taking
4.5 Results and discussion the average of these data. The various simulation parameters
used to evaluate the proposed EELCR method using existing
The simulation results and comparison of proposed and exist- protocols are listed in Table 2.
ing routing methods are discussed in this section. to confirm
the performance of our energy efficient lifetime aware cluster 4.7 Comparative analysis
based routing (EELCR) technique using different simula-
tion environments are varying number of nodes and number 4.7.1 Comparative analysis of compression techniques
of simulation rounds. Our EELCR technique was imple-
mented in network simulator (NS2) simulation tool. The best In this subsection, we analyzes the performance of proposed
instrument for studying communiqué networks is NS2, an and existing techniques have been used for data compres-
event-driven, open-source simulation. sion/coding. The compression ratio of proposed optimal
selective Huffman coding is compared with the exiting cod-
4.6 Simulation setup ing techniques are conventional Huffman coding and cascade
Huffman coding [30]. Table 3 describes the compression
A simulation system is used to evaluate the effectiveness of rate comparison of proposed and the exiting Huffman coding
the proposed EELCR method. The simulation is carried out techniques. For 200 bits of aggregated data, the compression

123
116 N. N. Sulthana, M. Duraipandian

Table 3 Compression rate compassion of proposed exiting coding tech- 4.7.2 Comparative analysis of routing techniques
niques

Data (bits) Huffman compression In this subsection, we analyzes the performance of pro-
posed and existing techniques have been used for routing.
Conventional Cascode Optimal selective The different simulations measures are used to analyze the
performance of proposed and existing routing techniques
200 194 178 165
are alive nodes, energy consumption, packet delivery ratio,
400 215 203 193
throughput, and network lifetime. The simulation results of
600 399 378 324 proposed EELCR technique is compared with the existing
800 641 638 547 routing techniques such as without compression (WC), run
1000 798 754 709 length encoding (RLE), adaptive Huffman encoding (AHE),
Hybrid-RLEAHE (H-RLEAHE) [31], Type-2 fuzzy cluster-
ing and LEACH clustering [30].For this simulation, we vary
the number of sensor nodes as 100, 200, 300, 400 and 500
ratio of our optimal selective Huffman coding is 14.948% and
with the fixed network size as 100 × 100 m2 area. Table 4
7.303% efficient than the existing state-of-art coding tech-
describes the simulation results comparison of proposed and
niques, conventional Huffman coding and cascade Huffman
existing routing techniques with respect to impact of node
coding [30], respectively. For 400 bits of aggregated data, the
density.
compression rate of our optimal selective Huffman coding
The number of alive nodes during the simulation of
is 10.233% and 4.926% efficient than the existing state-
proposed EELCR technique is 9.378%, 7.815%, 6.252%,
of-art coding techniques, conventional Huffman coding and
4.689%, 3.126% and 1.563% efficient than the existing
cascade Huffman coding [30], respectively. For 600 bits of
WC, RLE, AHE, H-RLEAHE, Type-2 fuzzy clustering
aggregated data, the compression rate of our optimal selective
and LEACH clustering, respectively. Figure 3 shows the
Huffman coding is 18.797% and 14.286% efficient than the
results of alive node comparison with respect to impact of
existing state-of-art coding techniques, conventional Huff-
nodes. The energy consumption of proposed EELCR tech-
man coding and cascade Huffman coding [30].For 800 bits of
nique is 41.320%, 36.980%, 31.947%, 26.040%, 19.010%
aggregated data, the compression rate of our optimal selective
and 10.503% efficient than the existing WC, RLE, AHE,
Huffman coding is 14.665% and 14.263% efficient than the
H-RLEAHE, Type-2 fuzzy clustering and LEACH clus-
existing state-of-art coding techniques, conventional Huff-
tering, respectively. Figure 4 shows the results of energy
man coding and cascade Huffman coding [30], respectively.
consumption comparison with respect to impact of nodes.
For 1000 bits of aggregated data, the compression rate of our
The packet delivery ratio of proposed EELCR technique is
optimal selective Huffman coding is 11.153% and 5.968%
21.683%, 18.069%, 14.455%, 10.841%, 7.228% and 3.614%
efficient than the existing state-of-art coding techniques, con-
efficient than the existing WC, RLE, AHE, H-RLEAHE,
ventional Huffman coding and cascade Huffman coding [30],
Type-2 fuzzy clustering and LEACH clustering, respectively.
respectively. Figure 2 shows the graphical representation of
Figure 5 shows the results of packet delivery ratio comparison
compression rate comparative analysis with respect to pro-
posed and existing coding techniques.

Fig. 2 Graphical representation

of compression rate comparative
analysis

123
 Table 4 Comparative analysis of simulation results with respect to impact of nodes

Routing techniques Alive nodes (%) Energy consumption (J) Packet delivery ratio (%)

100 200 300 400 500 100 200 300 400 500 100 200 300 400 500

WC 75.860 75.000 73.450 72.000 68.000 19.879 25.745 34.987 45.78 51.487 75 72 70 68 65
RLE 77.117 76.257 74.707 73.257 69.257 17.429 23.295 32.537 43.33 49.037 78.23 75.23 73.23 71.23 68.23
AHE 78.373 77.513 75.963 74.513 70.513 14.979 20.845 30.087 40.88 46.587 81.46 78.46 76.46 74.46 71.46
EELCR: energy efficient lifetime aware cluster based routing …

H-RLEAHE 79.630 78.770 77.220 75.770 71.770 12.529 18.395 27.637 38.43 44.137 84.69 81.69 79.69 77.69 74.69
Type-2 fuzzy cluster 80.887 80.027 78.477 77.027 73.027 10.079 15.945 25.187 35.98 41.687 87.92 84.92 82.92 80.92 77.92
LEACH clustering 82.144 81.284 79.734 78.284 74.284 7.629 13.495 22.737 33.53 39.237 91.15 88.15 86.15 84.15 81.15
EELCR 83.400 82.540 80.990 79.540 75.540 5.179 11.045 20.287 31.08 36.787 94.38 91.38 89.38 87.38 84.38

Routing techniques Throughput (Bps) Network lifetime (%)

100 200 300 400 500 100 200 300 400 500

WC 1666 1466 1360 966 466 84.86 82.188 81.48 77.08 74.877
RLE 3255 3055 2949 2555 2055 86.238 83.566 82.858 78.46 76.255
AHE 4844 4644 4538 4144 3644 87.616 84.944 84.236 79.84 77.633
H-RLEAHE 6433 6233 6127 5733 5233 88.994 86.322 85.614 81.21 79.011
Type-2 fuzzy cluster 8022 7822 7716 7322 6822 90.372 87.7 86.992 82.59 80.389
LEACH clustering 9611 9411 9305 8911 8411 91.75 89.078 88.37 83.97 81.767
EELCR 11,200 11,000 10,894 10,500 10,000 93.128 90.456 89.748 85.35 83.145

123
117
118 N. N. Sulthana, M. Duraipandian

Fig. 3 Alive node comparison

with respect to impact of nodes

Fig. 4 Energy consumption

comparison with respect to
impact of nodes

Fig. 5 Packet delivery ratio

comparison with respect to
impact of nodes

123
EELCR: energy efficient lifetime aware cluster based routing … 119

Fig. 6 Throughput comparison

with respect to impact of nodes

Fig. 7 Network lifetime

comparison with respect to
impact of nodes

with respect to impact of nodes. The throughput of pro- Figure 8 shows the results of alive node comparison with
posed EELCR technique is 88.947%, 74.122%, 59.298%, respect to simulation rounds.
44.473%, 29.649% and 14.8245 efficient than the exist- The energy consumption of proposed EELCR technique
ing WC, RLE, AHE, H-RLEAHE, Type-2 fuzzy clustering is 38.428%, 34.214%, 29.382%, 23.784%, 17.221% and
and LEACH clustering, respectively. Figure 6 shows the 9.422% efficient than the existing WC, RLE, AHE, H-
results of throughput comparison with respect to impact of RLEAHE, Type-2 fuzzy clustering and LEACH clustering,
nodes. The network lifetime of proposed EELCR technique respectively. Figure 9 shows the results of energy consump-
is 9.357%, 7.797%, 6.238%, 4.678%, 3.119% and 1.559% tion comparison with respect to impact of simulation rounds.
efficient than the existing WC, RLE, AHE, H-RLEAHE, The packet delivery ratio of proposed EELCR technique is
Type-2 fuzzy clustering and LEACH clustering, respectively. 22.294%, 18.578%, 14.863%, 11.147%, 7.431% and 3.716%
Figure 7 shows the results of network lifetime comparison efficient than the existing WC, RLE, AHE, H-RLEAHE,
with respect to impact of nodes. Type-2 fuzzy clustering and LEACH clustering, respectively.
For this simulation, we vary the number of simulation Figure 10 shows the results of packet delivery ratio compari-
rounds as 500, 1000, 1500, 2000 and 2500 with the fixed num- son with respect to impact of simulation rounds. The through-
ber of nodes as 500. Table 5 describes the simulation results put of proposed EELCR technique is 90.260%, 75.217%,
comparison of proposed and existing routing techniques with 60.173%, 45.130%, 30.087% and 15.043% efficient than the
respect to impact of simulation rounds. The number of alive existing WC, RLE, AHE, H-RLEAHE, Type-2 fuzzy cluster-
nodes during the simulation of proposed EELCR technique ing and LEACH clustering, respectively. Figure 11 shows the
is 10.052%, 8.377%, 6.701%, 5.026%, 3.351% and 1.675% results of throughput comparison with respect to impact of
efficient than the existing WC, RLE, AHE, H-RLEAHE, simulation rounds. The network lifetime of proposed EELCR
Type-2 fuzzy clustering and LEACH clustering, respectively. technique is 9.590%, 7.992%, 6.393%, 4.795%, 3.197%

123
 120

123
Table 5 Comparative analysis of simulation results with respect to impact of simulation rounds

Routing techniques Alive nodes (%) Energy consumption (J) Packet delivery ratio (%)

500 1000 1500 2000 2500 500 1000 1500 2000 2500 500 1000 1500 2000 2500

WC 70.471 69.611 68.061 66.611 62.611 22.557 28.423 37.665 48.458 54.165 72.55 69.55 67.55 65.55 62.55
RLE 71.728 70.868 69.318 67.868 63.868 20.107 25.973 35.215 46.008 51.715 75.78 72.78 70.78 68.78 65.78
AHE 72.984 72.124 70.574 69.124 65.124 17.657 23.523 32.765 43.558 49.265 79.01 76.01 74.01 72.01 69.01
H-RLEAHE 74.241 73.381 71.831 70.381 66.381 15.207 21.073 30.315 41.108 46.815 82.24 79.24 77.24 75.24 72.24
Type-2 fuzzy cluster 75.498 74.638 73.088 71.638 67.638 12.757 18.623 27.865 38.658 44.365 85.47 82.47 80.47 78.47 75.47
LEACH clustering 76.755 75.895 74.345 72.895 68.895 10.307 16.173 25.415 36.208 41.915 88.7 85.7 83.7 81.7 78.7
EELCR 78.011 77.151 75.601 74.151 70.151 7.857 13.723 22.965 33.758 39.465 91.93 88.93 86.93 84.93 81.93

Routing techniques Throughput (Bps) Network lifetime (%)

500 1000 1500 2000 2500 500 1000 1500 2000 2500

WC 1510 1310 1204 810 310 82.71 80.038 79.33 74.929 72.727
RLE 3099 2899 2793 2399 1899 84.088 81.416 80.708 76.307 74.105
AHE 4688 4488 4382 3988 3488 85.466 82.794 82.086 77.685 75.483
H-RLEAHE 6277 6077 5971 5577 5077 86.844 84.172 83.464 79.063 76.861
Type-2 fuzzy cluster 7866 7666 7560 7166 6666 88.222 85.55 84.842 80.441 78.239
LEACH clustering 9455 9255 9149 8755 8255 89.6 86.928 86.22 81.819 79.617
EELCR 11,044 10,844 10,738 10,344 9844 90.978 88.306 87.598 83.197 80.995
N. N. Sulthana, M. Duraipandian
EELCR: energy efficient lifetime aware cluster based routing … 121

Fig. 8 Alive node comparison

with respect to impact of
simulation rounds

Fig. 9 Energy consumption

comparison with respect to
impact of simulation rounds

Fig. 10 Packet delivery ratio

comparison with respect to
impact of simulation rounds

and 1.598% efficient than the existing WC, RLE, AHE, H- 5 Conclusion
RLEAHE, Type-2 fuzzy clustering and LEACH clustering, We have presented an energy efficient lifetime aware cluster
respectively. Figure 12 shows the results of network lifetime based routing (EELCR) technique for WSN using optimal
comparison with respect to impact of simulation rounds. clustering and compression. MGTO algorithm is introduced

123
122 N. N. Sulthana, M. Duraipandian

Fig. 11 Throughput comparison

with respect to impact of
simulation rounds

Fig. 12 Network lifetime

comparison with respect to
impact of simulation rounds

for efficient balanced clustering which minimizes energy EELCR technique is 51.88% and 52.625% effective than
consumption. An OSS algorithm is used to selects the best the existing routing techniques with respect to node density
optimal node called as CH for prolonging the lifetime in the and simulation rounds respectively. Furthermore, the com-
sensor networks by utilizing a multiple design constraints are putational complexity of our EELCR technique is notably
energy consumption, distance between sensor node to BA, low, standing at only 12.34% when compared to existing
congestion rate and received signal strength. Each CH nodes state-of-the-art methods. From the simulation results and
compress clustering data using optimal selective Huffman comparative analysis, we conclude that EELCR technique
compression which achieved the maximum compression is perform effectively with respect to different simulation
ratio which overcomes inefficiency of area overhead prob- environments and setup. In the future, we plan to expand
lem in existing Huffman compression. Then, a hybrid deep the application of this routing scheme to real-time scenarios,
learning technique i.e. DGNN is used to find optimal route such as in smart cities, to offer energy-efficient and lifetime-
for data transmission from CH to BS. Finally, we validate the aware routing solutions.
performance of proposed EELCR technique using the dif-
ferent simulation scenario and prove the effectiveness with Author contributions All authors equally contributed in the manuscript
respect to QoS parameters. The results shows that the average
Funding No financial support was provided by any funding agency.
compression rate of our optimal selective Huffman coding is
9.349% effective than the existing state-of-art techniques.
The average energy consumption of our EELCR technique Declarations
is 27.634% and 25.404% effective than the existing rout-
Conflict of interest The authors declare that they have no known com-
ing techniques with respect to node density and simulation peting financial interests or personal relationships that could have
rounds respectively. The average network lifetime of our appeared to influence the work reported in this paper.

123
EELCR: energy efficient lifetime aware cluster based routing … 123

References data compression in wireless sensor networks. Ad Hoc Networks,

83, 149–157.
1. Mao, W., Zhao, Z., Chang, Z., Min, G., & Gao, W. (2021). Energy- 19. Pacharaney, U. S., & Gupta, R. K. (2019). Clustering and compres-
efficient industrial internet of things: Overview and open issues. sive data gathering in wireless sensor network. Wireless Personal
IEEE Transactions on Industrial Informatics, 17(11), 7225–7237. Communications, 109(2), 1311–1331.
2. Li, F., Lam, K. Y., Li, X., Sheng, Z., Hua, J., & Wang, L. (2019). 20. Chen, S., Zhang, S., Zheng, X., & Ruan, X. (2019). Layered adap-
Advances and emerging challenges in cognitive internet-of-things. tive compression design for efficient data collection in industrial
IEEE Transactions on Industrial Informatics, 16(8), 5489–5496. wireless sensor networks. Journal of Network and Computer Appli-
3. Yao, Y., Cao, Q., & Vasilakos, A. V. (2014). EDAL: An energy- cations, 129, 37–45.
efficient, delay-aware, and lifetime-balancing data collection pro- 21. Sheeja, R., & Sutha, J. (2020). Soft fuzzy computing to medical
tocol for heterogeneous wireless sensor networks. IEEE/ACM image compression in wireless sensor network-based tele medicine
Transactions on Networking, 23(3), 810–823. system. Multimedia Tools and Applications, 79(15), 10215–10232.
4. Xiao, M., Wu, J., & Huang, L. (2014). Time-sensitive utility-based 22. Ghaderi, M. R., TabatabaVakili, V., & Sheikhan, M. (2020). FGAF-
single-copy routing in low-duty-cycle wireless sensor networks. CDG: Fuzzy geographic routing protocol based on compressive
IEEE Transactions on Parallel and Distributed Systems, 26(5), data gathering in wireless sensor networks. Journal of Ambient
1452–1465. Intelligence and Humanized Computing, 11(6), 2567–2589.
5. Cota-Ruiz, J., Rivas-Perea, P., Sifuentes, E., & Gonzalez-Landaeta, 23. Singh, A., & Nagaraju, A. (2020). Low latency and energy efficient
R. (2016). A recursive shortest path routing algorithm with appli- routing-aware network coding-based data transmission in multi-
cation for wireless sensor network localization. IEEE Sensors hop and multi-sink WSN. Ad Hoc Networks, 107, 102182.
Journal, 16(11), 4631–4637. 24. Rani, M. J., & Vasanthanayaki, C. (2020). Network condition based
6. Brar, G. S., Rani, S., Chopra, V., Malhotra, R., Song, H., & Ahmed, multi-level image compression and transmission in WSN. Com-
S. H. (2016). Energy efficient direction-based PDORP routing pro- puter Communications, 150, 317–324.
tocol for WSN. IEEE Access, 4, 3182–3194. 25. Aziz, A., Osamy, W., Khedr, A. M., El-Sawy, A. A., & Singh,
7. Huynh, T. T., Dinh-Duc, A. V., & Tran, C. H. (2016). Delay- K. (2020). Grey Wolf based compressive sensing scheme for data
constrained energy-efficient cluster-based multi-hop routing in gathering in IoT based heterogeneous WSNs. Wireless Networks,
wireless sensor networks. Journal of Communications and Net- 26(5), 3395–3418.
works, 18(4), 580–588. 26. Reddy, V., & Gayathri, P. (2020). Energy efficient data transmission
8. Sasirekha, S., & Swamynathan, S. (2017). Cluster-chain mobile in WSN thru compressive slender penetrative etiquette. Jour-
agent routing algorithm for efficient data aggregation in wireless nal of Ambient Intelligence and Humanized Computing, 11(11),
sensor network. Journal of Communications and Networks, 19(4), 4681–4693.
392–401. 27. Aziz, A., Singh, K., Osamy, W., & Khedr, A. M. (2020). An effi-
9. Bhavathankar, P., Chatterjee, S., & Misra, S. (2017). Link-quality cient compressive sensing routing scheme for internet of things
aware path selection in the presence of proactive jamming in fallible based wireless sensor networks. Wireless Personal Communica-
wireless sensor networks. IEEE Transactions on Communications, tions, 114(3), 1905–1925.
66(4), 1689–1704. 28. Ghaderi, M. R., TabatabaVakili, V., & Sheikhan, M. (2021).
10. Saleem, F., Majeed, M. N., Iqbal, J., Waheed, A., Rauf, A., Zareei, Compressive sensing-based energy consumption model for data
M., & Mohamed, E. M. (2021). Ant lion optimizer based clustering gathering techniques in wireless sensor networks. Telecommuni-
algorithm for wireless body area networks in livestock industry. cation Systems, 77(1), 83–108.
IEEE Access, 9, 114495–114513. 29. Jari, A., & Avokh, A. (2021). PSO-based sink placement and
11. Yang, L., Lu, Y., Yang, S. X., Guo, T., & Liang, Z. (2020). A secure load-balanced anycast routing in multi-sink WSNs considering
clustering protocol with fuzzy trust evaluation and outlier detec- compressive sensing theory. Engineering Applications of Artificial
tion for industrial wireless sensor networks. IEEE Transactions on Intelligence, 100, 104164.
Industrial Informatics, 17(7), 4837–4847. 30. Molk, A.M.N.G., Ghoreishi, S.M., Ghasemi, F. and Elyasi, I.
12. Zheng, J., Wang, P., & Li, C. (2010). Distributed data aggrega- (2022). Improve performances of wireless sensor networks for data
tion using Slepian-Wolf coding in cluster-based wireless sensor transfer based on fuzzy clustering and huffman compression. Jour-
networks. IEEE Transactions on Vehicular Technology, 59(5), nal of Sensors.
2564–2574. 31. Mishra, M., Sen Gupta, G., & Gui, X. (2022). Investigation of
13. Paek, J., & Ko, J. (2015). $ K $-Means clustering-based data energy cost of data compression algorithms in WSN for IoT appli-
compression scheme for wireless imaging sensor networks. IEEE cations. Sensors, 22(19), 7685.
Systems Journal, 11(4), 2652–2662.
14. Arunraja, M., Malathi, V., & Sakthivel, E. (2015). Distributed sim-
ilarity based clustering and compressed forwarding for wireless Publisher’s Note Springer Nature remains neutral with regard to juris-
sensor networks. ISA Transactions, 59, 180–192. dictional claims in published maps and institutional affiliations.
15. Lan, K. C., & Wei, M. Z. (2017). A compressibility-based cluster-
ing algorithm for hierarchical compressive data gathering. IEEE Springer Nature or its licensor (e.g. a society or other partner) holds
Sensors Journal, 17(8), 2550–2562. exclusive rights to this article under a publishing agreement with the
16. Wei, Z., Lijuan, S., Jian, G., & Linfeng, L. (2016). Image com- author(s) or other rightsholder(s); author self-archiving of the accepted
pression scheme based on PCA for wireless multimedia sensor manuscript version of this article is solely governed by the terms of such
networks. The Journal of China Universities of Posts and Telecom- publishing agreement and applicable law.
munications, 23(1), 22–30.
17. Chen, S., Liu, J., Wang, K., & Wu, M. (2019). A hierarchical adap-
tive spatio-temporal data compression scheme for wireless sensor
networks. Wireless Networks, 25(1), 429–438.
18. Uthayakumar, J., Vengattaraman, T., & Dhavachelvan, P. (2019). A
new lossless neighborhood indexing sequence (NIS) algorithm for

123
124 N. N. Sulthana, M. Duraipandian

N. Nisha Sulthana is work- M. Duraipandian received B.E.,

ing as Assistant Professor M.E., Degree in the Branch of
in Vivekanandha College Computer Science and Engineer-
of Technology for Women, ing from Kumaraguru College of
Tiruchengode, Namakkal. She Technology, Coimbatore. He then
received the B.E degree in Com- completed his Ph.D from Anna
puter Science and Engineering University-Chennai in 2015. He is
from A.M.S Engineering College currently working as Professor &
affiliated to Anna University, Head in the Department of Com-
Chennai in the year 2005 and puter Science and Engineering
M.E degree in Computer Sci- in Hindusthan Institute of Tech-
ence and Engineering from nology, Coimbatore.He has 20+
M.Kumarasamy College of years of teaching and research
Engineering affiliated to Anna experience He has published
University, Coimbatore in the more than 25 research papers in
year 2009. She is a Research Scholar of Anna University, Chennai various national and international journals. He is a fellow member
in the department of Information and Communication Engineering. in Indian Society for Technical Education and Computer Society of
She is a fellow member in Indian Society for Technical Education India.
and Computer Society of India. Her area of interest includes Wireless
Networks, Internet of Things and Network Security.

123