FLOC: HESITANT FUZZY LINGUISTIC TERM SET ANALYSIS IN ENERGY HARVESTING OPPORTUNISTIC CLUSTERING USING RELATIVE THERMAL ENTROPY AND RF ENERGY TRANSFER

International Journal of Computer Networks & Communications (IJCNC) Vol.14, No.2, March 2022
DOI: 10.5121/ijcnc.2022.14202 19
FLOC: HESITANT FUZZY LINGUISTIC TERM SET
ANALYSIS IN ENERGY HARVESTING
OPPORTUNISTIC CLUSTERING USING RELATIVE
THERMAL ENTROPY AND RF ENERGY TRANSFER
Junaid Anees and Hao-Chun Zhang
School of Energy Science & Engineering,
Harbin Institute of Technology, Harbin, China
ABSTRACT
Limited energy resources and sensor nodes’ adaptability with the surrounding environment play a
significant role in the sustainable Wireless Sensor Networks. This paper proposes a novel, dynamic, self-
organizing opportunistic clustering using Hesitant Fuzzy Linguistic Term Analysis- based Multi-Criteria
Decision Modeling methodology in order to overcome the CH decision-making problems and network
lifetime bottlenecks. The asynchronous sleep/awake cycle strategy could be exploited to make an
opportunistic connection between sensor nodes using opportunistic connection random graph. Every node
in the network observe the node gain degree, energy welfare, relative thermal entropy, link connectivity,
expected optimal hop, link quality factor etc. to form the criteria for Hesitant Fuzzy Linguistic Term Set. It
makes the node to evaluate its current state and make the decision about the required action (‘CH’, ‘CM’
or ‘relay’). Our proposed scheme leads to an improvement in network lifetime, packet delivery ratio and
overall energy consumption against existing benchmarks.
KEYWORDS
Graph Theory, Wireless Sensor Networks, Hesitant Fuzzy Linguistic Term Set, Opportunistic Routing and
RF Energy Transfer.
1. INTRODUCTION
Power-constrained WSNs have to adjust their sleep/wake cycle according to the application
requirements in order to maximize the network lifetime and overall energy consumption [1].
Sectional failure and thermal exposure can cause significant damage to sensor nodes. Moreover,
different units of a sensor node behave differently when exposed in sunlight for long period of
time for example; the performance of a typical transceiver is degraded with the increase in
temperature. The purpose of deploying WSNs is to achieve a shared goal through sensor
collaboration and data aggregation. In order to allocate the resources to sensor nodes effectively,
topology architecture is needed in which sensors are organized in clusters [1]. The multi-hop
routing in this clustering topology can result in the decrease of overall energy consumption and
interference among sensor nodes due to specific timeslots allocation [2]. In addition to it, it could
also effectively optimize the data redundancy by significantly reducing the collected data size
using data aggregation techniques at Cluster Head (CH) level [1-2].
Researchers have proposed different node scheduling techniques to save battery power of sensor
nodes i.e. synchronous and asynchronous sleep/awake scheduling. Asynchronous sleep/awake

20
scheduling is designed to prolong the network lifetime and improve energy utilization by creating
an opportunistic node connection between sensor nodes in the network [3-6]. Opportunistic
Routing (OR) is a popular technique to ensure sustainable operation of sensor nodes in which
WSNs can benefit from wireless medium broadcasting characteristics by selecting appropriate
candidate forwarders. The performance of OR significantly depends on several key factors, such
as OR metric, candidate selection algorithm, and candidate coordination method. Based on the
asynchronous sleep/awake scheduling in OR, a node can sense, process, and transmit/receive
during its active state [3-4]. Researchers have also worked on temperature adaptive sleep/awake
scheduling techniques [7-8].
The Hesitant Fuzzy Linguistic Term Set (HFLTS) is the best way to deal with this uncertainty.
The fuzzy multi-criteria analysis presents an effective framework where the alternative actions
are ranked according to the nodes’ assessments concerning each criterion. This motivated us to
work on this problem [9]. Keeping in view OR and temperature adaptive sleep/awake scheduling,
we have selected multiple parameters including time-frequency parameter, node’s gain energy,
relative thermal entropy, expected optimal hops, link quality factor in terms of signal-to-noise
ratio, as our attributes of hesitant fuzzy linguistic term set. These attributes are used to assess the
role of nodes and self adaptively make the appropriate decision in a round of operation.
Furthermore, our proposed scheme FLOC uses this information in a Multi-Attribute Decision
Modelling (MADM) framework to efficiently utilize our hesitant fuzzy linguistic term set to
incorporate a qualitative assessment of the parameters by a node and help the node observe a
situation adaptive role transition.
The rest of the paper is organized as follows: Section 2 contains the discussion on some related
works. System modelling is presented in Section 3. Our proposed scheme FLOC is presented in
Section 4. HFLTS analysis is provided in Section 5. Section 6 presents the simulation framework
and performance evaluation of the proposed technique. Finally, section 7 concludes the paper
with some targeted future works.
2. RELATED WORK
Various researchers have focused on proposing different routing protocols for WSNs based on
different parameters such as end–end delay, packet delivery ratio, network lifetime, overall
energy consumption, control packet overhead etc. Ogundile et al. [1] presented a detailed survey
for energy-efficient and energy balanced routing protocols for WSNs including the taxonomy of
cluster-based routing protocols for WSNs. Routing protocols in WSNs can be segmented into two
main categories, i.e., hierarchical and non-hierarchical routing protocols. Non-hierarchical
routing protocols are designed in accordance with overhearing, flooding, and sink node position
advertisement, whereas hierarchical routing protocols are designed on the basis of grid, tree,
cluster and area [1-3][10]. The multi-dynamic (higher-tier and lower-tier) roles of sensor nodes in
hierarchical routing can be useful in reducing energy consumption and traffic overhead. Higher-
tier nodes are responsible to store the information related to current position of mobile sink
whereas lower-tier nodes acquire this information by soliciting the higher-tier nodes [11].
Different hierarchical routing protocols have their own merits and demerits, but as far as cluster-
based hierarchical routing protocols are concerned, researchers have been challenged with a task
of achieving an optimal balance between end–end delay and energy consumption [10-11].
Yang et al. [5] introduced the utilization of sleep/awake cycle of sensor nodes to prolong the
network lifetime. The sleep/awake cycle can be segmented into two categories—synchronous and
asynchronous sleep/awake cycle. Depending on the network connectivity requirements in terms
of traffic coverage, Mukherjee et al. [12] proposed an asynchronous sleep/awake scheduling

21
technique with a minimum number of sensor nodes to achieve the required network coverage. As
a result of asynchronous sleep/awake scheduling, opportunistic node connections are established
between sensor nodes and their neighbours, which brings the need of Opportunistic Connection
Random Graph (OCRG) theory to properly model the opportunistic node connections by forming
a spanning tree. Anees et al. [6] proposed an energy-efficient multi-disjoint path opportunistic
node connection routing protocol for smart grids (SGs) neighbourhood area networks (NAN).
Anees et al. [10] also proposed a delay-aware energy-efficient opportunistic node selection in
restricted routing for delay-sensitive applications. In this protocol, the information related to
updated position of sink is advertised by multiple ring nodes and data is forwarded to mobile sink
using ring nodes having maximum residual energy.
In a few asynchronous sleep/awake scheduling techniques, the sensor nodes are found to remain
inactive listening mode for a long amount of time, resulting in unnecessary consumption of
energy. A popular WSN MAC protocol, Sensor Medium Access Control (SMAC), has been
proposed by Ye et al. [13]. SMAC protocol lets the node listen for a fixed interval of time and
turn their radio off (sleep state) for a fixed duration. Barkley-MAC (BMAC) [14] provides an
adaptive preamble sampling technique to effectively reduce the duty cycle and idle listening by
the sensor nodes. Shah et al. [22] devised a guaranteed lifetime protocol in which the sink node
assigns sleep/awake periods for other nodes depending on residual energy, sleep duration, and
coverage by the nodes. A mathematical model for temperature adaptive sleep/awake strategy is
developed by Bachir et al. [8] with three proposed algorithms i.e. Stop Operate (SO), a Power
control (PC), and Stop-Operate-Power-Control (SOPC). The sensor nodes running any of the
algorithms are supposed to observe the contemporary state based on a pre-calculated relationship
between node-density and temperature. Thermal entropy of the sensor nodes has been explored in
the intelligent sleep-scheduling technique iSleep [15]. Reinforcement Learning based sleep-
scheduling algorithm RL-Sleep has been proposed in [7] in which the authors have used a
temperature model and Q-learning technique to switch the sleep/awake states adaptively,
depending on the environmental situation.
It has been revealed through a detailed literature review that most of the clustering schemes
consider energy efficiency, traffic distribution, or coverage efficiency as the prime criteria for
state-scheduling and decision modelling of sensor nodes instead of relative thermal entropy,
temperature adaptability or hesitant fuzziness used for nodes’ role transition etc. A few entropy-
based clustering schemes have been proposed in which entropy weight coefficient method is
adopted for decision making in cluster-based hierarchical routing protocol [16-18]. Multi-Criteria
Decision Analysis (MCDA) and Multi-Attribute Sensors Decision Modelling (MADM) using
entropy weight coefficients are also types of entropy weight-based multi-criteria decision routing
[16]. Anees et al. [19] proposed hesitant fuzzy entropy based opportunistic clustering and data
fusion algorithm for heterogeneous WSNs. In this algorithm, the local sensory data is gathered
from sensor nodes by utilizing hesitant fuzzy entropy based multi-attribute decision modelling for
cluster head election procedure.
Afsar et al. [20] proposed an unequal size clustering method known as CREST in which the
probability of becoming CH in a cluster is based on a function of the distance between BS and
the node by employing track-based algorithms. Varshney et al. [21] proposed an emerging
concept of simultaneous wireless information and power transfer (SWIPT) in which both energy
and data are transferred over RF links simultaneously. Guo et al. [22] utilized the concept of the
SWIPT to extend the network lifetime of a clustered WSN by wirelessly charging the relay nodes
which are responsible to share data with BS. Zhou et al. [23] proposed dynamic power splitting
(DPS) to adjust the power ratio of information encoding and energy harvesting in EHWSNs.
Anees et al. [24] proposed harvested energy scavenging and transfer capabilities in opportunistic
ring routing in which a distinguishing approach of hybrid (ring + cluster) topology is used in a

22
virtual ring structure and then a two-tier routing topology is used in the virtual ring as an overlay
by grouping nodes into clusters.
Overall, to the best of our knowledge, there is no published literature which focuses on thermal
entropy-based HFLTS analysis for energy-efficient opportunistic clustering. In this paper, we
have considered a set of attributes that regulate the nodes’ decisions about its role transition
conducive to the current situation in a cluster and provided a detailed solution for optimally
handling problems in energy efficient opportunistic clustering using relative thermal entropy
based HFLTS analysis. The comparison between FLOC and various scheduling algorithms is
given in Table 1.
Table 1. Comparison between FLOC and various scheduling algorithms
Attribut
es
Articles
Networ
k
Lifetim
e
Hetero
g--
eneity
Conect
ivity
Mobilit
y
pattern
Energ
y
Efficie
ncy
Therma
l
Entropy
Traffic
Distrib
ution
Energy
Harves
ting
Cove
rage
[6] √ √ √ √ √ x x x x
[7] √ √ x x √ x √ x x
[8] √ √ √ x √ √ x x x
[34] √ x √ x √ √ x x x
[10] √ √ √ √ √ x x x x
[12] √ √ √ √ √ x x x x
[15] √ x √ x √ x x x x
[13] √ x x x √ x x x x
[14] √ x x x √ x √ x x
[35] √ x √ x √ x √ x √
[15] √ √ √ x √ x x x x
[36] √ x x x √ x x x √
[37] √ x √ x √ x x x √
[38] √ x x x √ x x x √
[24] √ √ √ √ √ x x √ x
[29] √ x √ x √ x x x x
[33] √ x √ x √ x √ x x
FLOC √ √ √ √ √ √ √ √ √
3. SYSTEM MODELLING
3.1. Network model
A 𝑀𝑥𝑀 network area denoted as 𝐴 is considered for FLOC in which 𝑁 sensor nodes are
deployed randomly and independently. We have assumed that sensor nodes follow a uniform
distribution. The node-density of the network is denoted as 𝜆0 =
𝑁
𝐴
. All sensor nodes use short
radio range (RS) for sensing and transmission purposes whereas sink nodes can use RS for
transmission & reception and long radio range (RL) for data collection tasks using a tag message.
However, all sensor nodes can exploit the power control function and communicate with different
neighbouring nodes within various power levels. A probe message is shared by each sensor node
to acquire the neighbour information as discussed in [6]. Each sensor node is equipped with a
power splitting radio, which is composed of a signal processing unit to transfer energy to or from
neighbours using RF link. Moreover, it is also assumed that every sensor node is aware of its
position using the energy-efficient localization method [25-28]. Each sensor node is

23
characterized by a set of k attributes named as 𝐶 = {𝑐1, 𝑐2, … . 𝑐𝑘} and a set of weights 𝑤𝑡 =
{𝑤𝑡1, 𝑤𝑡2, … . 𝑤𝑡𝑝} is assigned by sensor node to the 𝑝 criteria of 𝐶. Furthermore, the sensor node
undergoes 𝑦 states i.e. 𝑆𝑇 = {𝑆𝑇1, 𝑆𝑇2, … . 𝑆𝑇𝑦}, where 𝑆𝑇1 represents the favorable state
(attribute values are above threshold) and 𝑆𝑇𝑦 represents the stressed state. Depending on
multiple parameters, the sensor node decides about the most suitable action against the
contemporary state i.e. 𝐴𝐶 = {𝐶𝐻, 𝐶𝑀, 𝑅𝑒𝑙𝑎𝑦}.
3.2.Energy Model
The energy consumption model [10] for radio energy dissipation during transmission and
reception is considered in which the energy required to transmit l bits of data over distance d can
be given in (1) as:
𝐸𝑇𝑥(𝑉𝑖,𝑉
𝑗) = {
𝐸𝑒𝑙𝑒𝑐𝑙 + 𝜀𝑓𝑠𝑙𝑑𝑉𝑖𝑉𝑗
2
𝑑 < 𝑑0
𝐸𝑒𝑙𝑒𝑐𝑙 + 𝜀𝑚𝑝𝑙𝑑𝑉𝑖𝑉𝑗
4
𝑑 ≥ 𝑑0
(1)
where Eelec is the energy spent by transmitter on running the radio electronics, εfs is the free
space energy dissipated by power amplifier depending on the Euclidean distance dViVj
between
the transmitter and receiver, εmp is the muti-path fading factor for energy dissipated by power
amplifier depending on Euclidean distance dViVj
between transmitter and receiver. The threshold
distance dois given as do = √
εfs
εmp
⁄ . Likewise, the energy required to receive l bits of data over
distance d is given in (2) as:
𝐸𝑅𝑥 = 𝐸𝑒𝑙𝑒𝑐𝑙 (2)
The energy used for sensing l bits of data in the virtual ring at the beginning of each round can be
given as 𝐸𝑠𝑒𝑛𝑠𝑒 = 𝐸𝑒𝑙𝑒𝑐𝑙. Accordingly, the total energy consumed by cluster member (CM) can
be computed in (3) as:
𝐸𝐶𝑀 = 𝐸𝑠𝑒𝑛𝑠𝑒 + 𝐸𝑇𝑥 = 𝐸𝑒𝑙𝑒𝑐𝑙 + 𝐸𝑒𝑙𝑒𝑐𝑙 + 𝜀𝑓𝑠𝑙𝑑𝑉𝑖𝑉𝑗
2
(3)
Each CH is responsible for data gathering, aggregating the received data and then relaying that
data towards sink, so the total energy consumed by a CH can be computed in (4) - (5) as
𝐸𝐶𝐻 = 𝐸𝑠𝑒𝑛𝑠𝑒 + (
𝑁
𝑁𝐶
− 1) 𝐸𝑅𝑥 + (
𝑁
𝑁𝐶
)𝑙𝐸𝐴 + (
𝑁
𝑟
)𝐸𝑇𝑥 (4)
𝐸𝐶𝐻 = 𝐸𝑒𝑙𝑒𝑐𝑙 + (
𝑁
𝑁𝐶
− 1) 𝐸𝑒𝑙𝑒𝑐𝑙 + (
𝑁
𝑁𝐶
)𝑙
𝐸𝑒𝑙𝑒𝑐
𝑅𝐶𝐶
+ (
𝑁
𝑟
) 𝐸𝑒𝑙𝑒𝑐𝑙 + (
𝑁
𝑟
) 𝜀𝑚𝑝𝑙𝑑𝑉𝑖𝑉𝑗
4
(5)
where NC represents the number of clusters in the network,
N
NC
is the number of working sensor
nodes per cluster in which we have 1 CH and
N
NC
− 1 CMs. EA signifies the data aggregating
energy at CH level, r represents the compression ratio and RCC symbolizes the communication to
computation ratio.

24
4. PROPOSED SCHEME FLOC
In this section, the proposed scheme FLOC is discussed in detail. The sink node launches data
collection by broadcasting a tag message containing the mobile sink address and data collection
duration. Sensor nodes calculate their working-sleeping cycle keeping in view the data collection
duration of the mobile sink. Subsequently, each sensor node broadcasts a probe message
containing source address, broadcast address, working-sleeping schedule, neighbour address,
total energy, thermal entropy and Expected Optimal Hop (EOH) [6].
4.1. Ambient Temperature and Relative Thermal Entropy
Keeping in view the diurnal temperature variation caused by solar radiation, the sensor nodes
placed under direct sunlight absorb higher heat energy than the sensor nodes in shadow.
According to temperature model in [7], the temperature of a sensor node 𝑖 after solar heat
absorption for amount of time ∆𝑡 can be represented in (6) as,
𝑇𝑡+∆𝑡
𝑖
= 𝑚𝑎𝑥 {𝑇𝑡
𝑖
+
(𝑆𝑆𝑈𝑁(𝑡)𝛼(𝑡)−𝜂𝑇4)
𝑐𝑝𝜃
𝐴𝑟𝑒𝑎𝑠𝑒𝑛𝛥𝑡, 𝑇𝑡
𝑖} (6)
where 𝑇𝑡
𝑖
is the temperature of a node 𝑖 at time 𝑡, 𝑆𝑆𝑈𝑁(𝑡) denotes the amount of radiation by the
sun at that time, 𝛼(𝑡) is the temporal variation of sun exposure, 𝐴𝑟𝑒𝑎𝑠𝑒𝑛 is the exposed area
through which the sensor node absorbs solar heat, 𝜂 is Boltzman constant, 𝜃 represents the mass
of the sensor node, 𝑐𝑝 represents the specific heat and 𝑇𝑡+∆𝑡
𝑖
symbolizes the ambient temperature.
The change in temperature of a sensor node can be extracted from equation (7) i.e.
𝛥𝑇𝑖 =
(𝑆𝑆𝑈𝑁(𝑡)𝛼(𝑡)−𝜂𝑇4)
𝑐𝑝𝜃
𝐴𝑟𝑒𝑎𝑠𝑒𝑛𝛥𝑡 (7)
The resultant temperature 𝑇𝑖 of a sensor node i is given as 𝑇𝑖 = 𝑇𝑡+∆𝑡
𝑖
= 𝑇 + 𝛥𝑇𝑖. Foregoing in
view, the solar radiation pattern for a day can be represented as 𝑆𝑆𝑈𝑁(𝑡) = 𝑆𝑆𝑈𝑁
𝑚𝑎𝑥
𝑒𝑥𝑝
−(𝑡−𝜌)2
2𝜎2
, 0 ≤
t ≤ 2ρ where SSUN
max
is the peak value of the solar radiation during the day [7]. Let Ti be the
temperature of the 𝑖th node and 𝑇𝐻 be the highest temperature for which the 𝑖th node becomes
non-operational. The probability of failure of a sensor node due to temperature increase can be
represented as 𝑝𝑖 =
𝑇𝑖
𝑇𝐻
, where 𝑇𝑖 can be acquired from equation (6) & (7) and 𝑇𝐻 symbolizes the
highest temperature the sensor node can withstand. Here we have assumed that 𝑇𝐻 is the same
for all sensor nodes in the network. The cumulative effect of failure likelihood leads to network
instability; therefore we need a probability distribution function (PDF) to measure the degree of
uncertainty in the sensor network. This hesitancy or irresolution resembles entropy.
The entropy H(X) of a random variable 𝑋 = {𝑥1, 𝑥2, . . . 𝑥𝑛} having probability distribution as
p(X) can be given as 𝐻(𝑋) = − ∑ 𝑝(𝑥)𝑙𝑜𝑔2
𝑥∈𝑋 𝑝(𝑥) for 0 ≤ H(X) ≤ 1 [10]. Similarly, the
Shannon’s entropy at 𝑖th node can be defined as 𝐻(𝑝𝑖) = −𝑝𝑖𝑙𝑜𝑔2𝑝𝑖. The relative contribution of
a sensor node towards the probable instability of the network can be estimated using relative
thermal entropy by calculating the entropy in neighborhood i.e.
𝐻𝑟𝑒𝑙
𝑡ℎ𝑒𝑟𝑚
=
𝐻(𝑝𝑖)
∑ 𝐻(𝑝𝑗)
𝑗∈𝑛𝑏𝑟𝑖
(8)
where 𝐻𝑟𝑒𝑙
𝑡ℎ𝑒𝑟𝑚
indicates the relative thermal entropy and 𝑛𝑏𝑟𝑖 represents the neighbourhood
dataset in equation (8).

25
4.2. Energy Transfer and Asynchronous Sleep/Awake Cycle
We have assumed that the sensor nodes in the network are able to control their power levels in
order to communicate with the neighbours. In this perspective, the amount of energy a node 𝑖
could acquire from its neighbouring sensor node 𝑗 through RF transfer can be defined in equation
(9) and (10) as:
𝐸𝑡𝑟𝑎𝑛𝑠(𝑉𝑗,𝑉𝑖) = 𝜂1𝜇𝑃𝑗|ℎ𝑉𝑖,𝑉𝑗
|2
= 𝜂1𝜇𝑃𝑗|𝛽1𝑑(𝑉𝑖,𝑉𝑗)
−𝛼1
|2
(9)
𝛤
𝑉𝑖
= ∑ 𝐸𝑡𝑟𝑎𝑛𝑠(𝑉𝑗,𝑉𝑖)
𝑘
𝑗=1 = ∑ 𝜂1𝜇𝑃𝑗|𝛽1𝑑(𝑉𝑖,𝑉𝑗)
−𝛼1
|2
𝑘
𝑗=1 (10)
where 𝐸𝑡𝑟𝑎𝑛𝑠(𝑉𝑗,𝑉𝑖) is the amount of energy node 𝑗 can transfer to its neighbor 𝑖, 𝜂1 is the energy
conversion efficiency 0 <𝜂1< 1, 𝜇 is the energy and data splitting ratio 0 <𝜇< 1, 𝑃𝑗 is the signal
power received from node j, ℎ𝑉𝑖,𝑉𝑗
is the channel gain, 𝛽1 is a constant which depends on the
radio propagation properties of the environment, 𝛼1 is the path loss exponent, and 𝛤
𝑉𝑖
is the node
𝑉𝑖’s gain degree. Furthermore, the total available energy at node 𝑖 can be computed as:
𝐸𝑇(𝑉𝑖) = 𝛤
𝑉𝑖
+ 𝐸𝐵𝑎𝑡(𝑉𝑖) (11)
where 𝐸𝐵𝑎𝑡(𝑉𝑖) is the remaining battery energy of node i in (11). In contrast to conventional
routing algorithms in WSNs, our proposed scheme can serve both data and energy in its routing
topology. We used opportunistic connection random graph (OCRG) in FLOC to model the
opportunistic node connections between sensor nodes. Let 𝐺(𝑆𝑆𝑁, 𝑂𝐶, 𝐿) be the graph
representing OCRG in which 𝑆𝑆𝑁 represents the set of nodes in the network, 𝑂𝐶 represents set of
opportunistic connections existing between any two adjacent neighbours and 𝐿 represents the link
connectivity of any two adjacent nodes in 𝑆𝑆𝑁. The link connectivity also depends on the data
routing cost 𝐷𝑅𝐶: 𝑂𝐶 → 𝑅 such that 𝐷𝑅𝐶(𝑖, 𝑗) is the cost associated with link (𝑖, 𝑗). Our routing
metric can be defined as: 𝑀𝑖𝑛 ∑ 𝐷𝑅𝐶(𝑉𝑖,𝑉𝑖+1)
𝑛−1
𝑖=1 , 𝐷𝑅𝐶(𝑉𝑖,𝑉𝑖+1) ≥ 0. The data routing cost can be
computed using 𝐸𝐶(𝑉𝑖,𝑉𝑗), 𝐸𝑇(𝑉𝑖) and 𝐸𝑇(𝑉𝑗) and is given in (12).
𝐷𝑅𝐶(𝑉𝑖,𝑉𝑗) =
𝐸𝐶(𝑉𝑖,𝑉𝑗)
(𝐸𝑇(𝑉𝑖)+ 𝐸𝑇(𝑉𝑗))
(12)
where 𝐸𝐶(𝑉𝑖,𝑉𝑗) is the transmission energy consumed over link (𝑖, 𝑗), 𝐸𝑇(𝑉𝑖) is the available total
energy (including battery and gained energy through RF transfer) of node 𝑖, 𝐸𝑇(𝑉𝑗) is the
available energy (including battery and gained energy through RF transfer) of node 𝑗. It is
pertinent to mention that energy is also transferred along with the data in the routing process to
compensate for the transmission energy consumed over each link and more energy is conserved
than consumed as the sensor nodes are using strong signals for transmission purposes.
As far as asynchronous sleep/awake cycle is concerned, we have proposed the concept of
sleep/awake cycle schedule (𝑊
𝑣/𝑆𝑣) and status transition frequencies (𝐹𝑆𝑇) to investigate the
opportunistic node connection between sensor nodes in each data collection period. We
calculated the time-frequency parameter 𝑇𝐹𝑉𝑖𝑉𝑗
based on working time 𝑊𝑉𝑖
and 𝑊𝑉𝑗
of sensor
nodes 𝑉𝑖 and 𝑉
𝑗, data collection duration 𝑇𝐶𝑃, status transition frequencies 𝐹𝑆𝑇𝑖 and 𝐹𝑆𝑇𝑗 in
equation (13) and (14) as,
𝑇𝐹𝑉𝑖𝑉𝑗
= (
𝐹𝑆𝑇𝑉𝑖
𝐹𝑆𝑇𝑚𝑎𝑥
×
𝑊𝑉𝑖
𝑇𝐶𝑃
) (
𝐹𝑆𝑇𝑉𝑗
×
𝑊𝑉𝑗
𝑇𝐶𝑃
) (13)

26
𝑇𝐹𝑉𝑖𝑉𝑆𝐼𝑁𝐾
= (
×
𝑊𝑉𝑖
𝑇𝐶𝑃
) (𝑊𝑉𝑆𝐼𝑁𝐾
) (14)
Using the time-frequency parameter 𝑇𝐹𝑉𝑖𝑉𝑗
and data routing cost 𝐷𝑅𝐶, our link connectivity
𝐿𝑉𝑖𝑉𝑗
can be computed in (15) as,
= 𝛼2𝐷𝑅𝐶(𝑖, 𝑗) + (1 − 𝛼2)𝑇𝐹𝑉𝑖𝑉𝑗
(15)
= 𝛼2 (
𝐸
𝐶(𝑉𝑖,𝑉𝑗)
(𝐸𝑇(𝑉𝑖)+ 𝐸
𝑇(𝑉𝑗)
) + (1 − 𝛼2) (
×
𝑊𝑉𝑖
𝑇𝐶𝑃
) (
𝐹𝑆𝑇𝑉𝑗
×
𝑊𝑉𝑗
𝑇𝐶𝑃
) (16)
where α2 is the appropriate weight assigned to data routing cost and time-frequency parameter in
(16).
4.3. Expected Optimal Hop (𝐄𝐎𝐇):
In order to minimize the energy consumption, we employed the expected optimal hops (EOH)
approach. EOH can be defined as the number of hops needed to forward a probe message from
any node to the mobile sink node with minimum energy consumption [6]. Assuming that there is
B-bit data from node 𝑉𝑖 to be forwarded to the mobile sink 𝑉𝑆𝐼𝑁𝐾 along the path (𝑉𝑖,𝑉𝑆𝐼𝑁𝐾, 𝑘),
where 𝑘 is the hops on the path and satisfies 1 ≤ 𝑘 ≤ 𝑁 − 1. To unify the path representation,
VSINK is represented by 𝑉𝑖+𝑘 which means that it needs k hops to forward a probe message from
𝑉𝑖 to 𝑉𝑆𝐼𝑁𝐾 . Then, the energy consumption function with respect to 𝑘 for this forwarding can be
calculated in equation (17) as follows:
𝐶𝑉𝑖
(𝑘) = ∑ (𝐸𝑇 + 𝐸𝑅)
𝑘−1
𝑗=0 = ∑ {[𝐸𝑒𝑙𝑒𝑐 + 𝜀𝑎𝑚𝑝𝑑(𝑉𝑖+𝑗,𝑉𝑖+(𝑗+1))
2
] 𝐵 + 𝐸𝑒𝑙𝑒𝑐𝐵}
𝑘−1
𝑗=0
= 2𝑘𝐸𝑒𝑙𝑒𝑐𝐵 + 𝜀𝑎𝑚𝑝 ∑ [𝑑(𝑉𝑖+𝑗,𝑉𝑖+(𝑗+1))
2
]
𝑘−1
𝑗=0 𝐵 (17)
To get the minimum value of 𝐶𝑉𝑖
(𝑘), we can use the average value inequality to derive inequality
of 𝐶𝑉𝑖
(𝑘) which can be described in equation (18) as:
𝐶𝑉𝑖
(𝑘) ≥ 2𝑘𝐸𝑒𝑙𝑒𝑐𝐵 +
𝜀𝑎𝑚𝑝[∑ 𝑑(𝑉𝑖+𝑗,𝑉𝑖+(𝑗+1))
𝑘−1
𝑗=0 ]
2
𝐵
𝑘
(18)
Since the sum of the distances for any two adjacent nodes on the path (𝑉𝑖, 𝑉𝑖+𝑘, 𝑘) satisfies the
following inequality in equation (19):
∑ 𝑑(𝑉𝑖+𝑗,𝑉𝑖+(𝑗+1))
𝑘−1
𝑗=0 ≥ 𝑑(𝑉𝑖,𝑉𝑖+𝑘) (19)
Therefore, the minimum energy consumption function 𝐶𝑉𝑖
𝑚𝑖𝑛
(𝑘) with respect to 𝑘 satisfies
equation (20):
𝐶𝑉𝑖
𝑚𝑖𝑛(𝑘) = 2𝑘𝐸𝑒𝑙𝑒𝑐𝐵 +
𝜀𝑎𝑚𝑝[𝑑(𝑉𝑖,𝑉𝑖+𝑘)]
2
𝐵
𝑘
(20)
To get the value of k that can minimize 𝐶𝑉𝑖
𝑚𝑖𝑛(𝑘), we can take the first derivative of 𝐶𝑉𝑖
𝑚𝑖𝑛(𝑘)
with respect to 𝑘 in equation (21) as follows:

27
𝜕 𝐶𝑉𝑖
𝑚𝑖𝑛
𝜕𝑘
= 2𝐸𝑒𝑙𝑒𝑐𝐵 −
𝜀𝑎𝑚𝑝[𝑑(𝑉𝑖,𝑉𝑖+𝑘)]
2
𝐵
𝑘2 (21)
Let
𝜕 𝐶𝑉𝑖
𝑚𝑖𝑛
𝜕𝑘
equal to 0, so that we can get a value of 𝑘𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 as √
𝜀𝑎𝑚𝑝
2𝐸𝑒𝑙𝑒𝑐
𝑑(𝑉𝑖,𝑉𝑖+𝑘). Therefore, the
𝐶𝑉𝑖
𝑚𝑖𝑛(𝑘) has a minimum value at 𝑘 = 𝑘𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑, and since 𝑉𝑖+𝑘 represents 𝑉𝑆𝐼𝑁𝐾, so the value of
𝐸𝑂𝐻 for node 𝑉𝑖 in equation (22) will be:
𝐸𝑂𝐻𝑉𝑖
= √
𝜀𝑎𝑚𝑝
2𝐸𝑒𝑙𝑒𝑐
𝑑(𝑉𝑖,𝑉𝑆𝐼𝑁𝐾) (22)
where E is the basic energy consumed during transmission and reception per bit, 𝜀𝑎𝑚𝑝 is the
energy consumed by the transmission amplifier, 𝑑(𝑉𝑖,𝑉𝑆𝑁𝐾)is the distance between node 𝑉𝑖 and
𝑉𝑆𝐼𝑁𝐾 and 𝐸𝑂𝐻𝑉𝑖
is the expected optimal hops of node 𝑉𝑖 needed to select proper forwarders to
the mobile sink with minimum energy consumed.
4.4. Energy-welfare (EW)
Energy balance is required for a properly active WSN. The measure of energy distribution pattern
among sensor nodes can be represented by a parameter known as Energy welfare. We use 𝐸𝑊 to
provide us an estimate of energy balance within sensor node’s neighborhood. Equation (23) can
be used to formulate the 𝐸𝑊.
𝐸𝑊 = (
1
𝑛𝑏𝑟𝑖
∑ (𝐸𝑛𝑒𝑟𝑔𝑦𝑗)
(1−𝜀)
𝑗∈𝑁𝐵𝑖
)
1
(1−𝜀)
(23)
where 𝑛𝑏𝑟𝑖 is the number of neighbors of node 𝑖, 𝑁𝐵𝑖 is the set of actual neighbors of 𝑖,
𝐸𝑛𝑒𝑟𝑔𝑦𝑗is the residual energy of actual neighbors. 𝜀 is the inequality aversion factor and its
value will be 1.5, 2.0 or 2.5. In our case, we have chosen 𝜀 = 2.0. We have also normalized the
value of 𝐸𝑊 as its one of the attributes involved in decision modeling and also a member of
hesitant fuzzy set. The normalized value of EW can be represented as,
𝐸𝑊
𝑛𝑜𝑟𝑚 =
𝐸𝑊𝑟𝑒𝑐𝑒𝑛𝑡
𝐸𝑊𝑚𝑎𝑥
(24)
where 𝐸𝑊𝑟𝑒𝑐𝑒𝑛𝑡 is the recent value of 𝐸𝑊 obtained during the current iteration in equation (24).
4.5. Link Quality Factor (𝐋𝐐𝐅)
We measure the link quality factor in terms of the signal-to-noise ratio between sensor nodes.
The acceptable dynamic range of sensor nodes for signal-to-noise ratio is 10-15 dB [6]. The
ambient temperature increase due to solar radiation could affect the As we are dealing with
opportunistic clustering environment where the mobility of sink node could result in the
intermittency of communication links between sensor nodes and sink, LQR should be selected as
one of the decision attributes in HFLTS analysis.

28
5. HESITANT FUZZY LINGUISTIC TERM SET (HFLTS) ANALYSIS
A generalization of the basic fuzzy set which deals with the uncertainty starting from the
hesitation in the assignment of membership degrees of an element is known as Hesitant Fuzzy
Set (HFS) [9-10,16-18]. We start the HFLTS analysis with a set of inputs containing a total
number of nodes, sink, neighbour information, context-free grammar, transformation function,
set of alternatives, set of criteria and weight assignment. In FLOC, a node can attain two states
based on the node’s gain energy and energy welfare. The state evaluation of a node will be
‘optimistic’ if its total energy is greater than the threshold and the normalized energy welfare is
greater than half of the maximum value of energy welfare. Likewise, the state evaluation of a
node will be ‘pessimistic’ if its total energy is less than the threshold and the normalized value of
energy welfare is less than the half of the maximum value of energy welfare. After evaluating the
state and acquiring the neighbourhood information, the node calculates the relative thermal
entropy with reference to neighbourhood. The next step is to store different attributes in an array
and perform the data standardization by normalizing different attributes to obtain the fractional
representation of attributes within [0 1] before defining the criteria [29-33]. The pseudo code for
FLOC is given in Algorithm 1.
𝑨𝒍𝒈𝒐𝒓𝒊𝒕𝒉𝒎 𝟏: 𝑭𝑳𝑶𝑪
𝑰𝑵𝑷𝑼𝑻:
1. 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑛𝑜𝑑𝑒𝑠 𝑵, 𝑆𝑖𝑛𝑘 𝑛𝑜𝑑𝑒𝑽𝑺𝑰𝑵𝑲
2. 𝐶𝑜𝑛𝑡𝑒𝑥𝑡 𝑓𝑟𝑒𝑒 grammar 𝐺𝐶𝐹 (𝑉𝑁, V𝑇, 𝐼, 𝑃), 𝑡𝑟𝑎𝑛𝑠𝑓𝑜𝑟𝑚𝑎𝑡𝑖𝑜𝑛 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛 𝑬𝑮𝑪𝑭
3. 𝑨𝒍𝒕𝒆𝒓𝒏𝒂𝒕𝒊𝒗𝒆𝒔 = {𝐶𝐻, 𝐶𝑀, 𝑅𝑒𝑙𝑎𝑦}
4. 𝑪𝒓𝒊𝒕𝒆𝒓𝒊𝒂, 𝒘𝒕 = {𝑤1, 𝑤2,…𝑤|𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎|}
5. 𝐿𝑖𝑛𝑔𝑢𝑖𝑠𝑡𝑖𝑐 𝑡𝑒𝑟𝑚 𝑠𝑒𝑡 𝑺 = { 𝒔𝟏 ∶ 𝒏𝒆𝒈𝒍𝒊𝒈𝒊𝒃𝒍𝒆(𝒏), 𝒔𝟐 ∶ 𝒗𝒆𝒓𝒚 𝒍𝒐𝒘(𝒗𝒍), 𝒔𝟑 ∶ 𝒍𝒐𝒘(𝒍), 𝒔𝟒 ∶
𝒎𝒆𝒅𝒊𝒖𝒎(𝒎), 𝒔𝟓 ∶ 𝒉𝒊𝒈𝒉(𝒉), 𝒔𝟔 ∶ 𝒗𝒆𝒓𝒚 𝒉𝒊𝒈𝒉(𝒗𝒉), 𝒔𝟕 ∶ 𝒑𝒆𝒓𝒇𝒆𝒄𝒕(𝒑) }
6. 𝑀𝑜𝑏𝑖𝑙𝑒 𝑠𝑖𝑛𝑘 𝑏𝑟𝑜𝑎𝑑𝑐𝑎𝑠𝑡𝑖𝑛𝑔 𝑆𝐼𝐷 𝑎𝑛𝑑 𝑑𝑎𝑡𝑎 𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛 𝑡𝑜 𝑎𝑙𝑙 𝑛𝑜𝑑𝑒𝑠 𝑤𝑖𝑡ℎ𝑖𝑛 𝑅𝑙
7. 𝑆𝑒𝑛𝑠𝑜𝑟 𝑛𝑜𝑑𝑒𝑠 𝑤𝑖𝑡ℎ𝑖𝑛 𝑅𝑙 𝑟𝑒𝑐𝑒𝑖𝑣𝑒 𝑡ℎ𝑒 𝑑𝑎𝑡𝑎 𝑐𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑚𝑒𝑠𝑠𝑎𝑔𝑒
𝑶𝑼𝑻𝑷𝑼𝑻:
8. State Evaluation of the node
𝑩𝑬𝑮𝑰𝑵:
9. 𝑭𝒐𝒓 𝑒𝑎𝑐ℎ 𝑛𝑜𝑑𝑒 𝑖 𝜖 (𝑁 −𝑉𝑆𝐼𝑁𝐾), 𝒅𝒐
10. 𝑰𝒇 𝑬𝒏𝒆𝒓𝒈𝒚𝒓𝒆𝒎𝒂𝒊n(𝑖) ≤ 𝐸𝑇ℎ𝑟𝑒𝑠ℎ
11. 𝑨𝒄𝒕𝒊𝒐𝒏(𝒊) = ′𝑆𝑙𝑒𝑒𝑝′
12. 𝑬𝒏𝒅𝑰𝒇
13. 𝑬𝒏𝒅 𝐹𝑜𝑟
14. 𝑪𝒂𝒍𝒄𝒖𝒍𝒂𝒕𝒆 𝑡ℎ𝑒 𝑤𝑜𝑟𝑘𝑖𝑛𝑔–𝑠𝑙𝑒𝑒𝑝𝑖𝑛𝑔 𝑐𝑦𝑐𝑙𝑒 𝑎𝑛𝑑 𝑠𝑡𝑎𝑡𝑢𝑠 𝑡𝑟𝑎𝑛𝑠𝑖𝑡𝑖𝑜𝑛 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑖𝑒𝑠
15. 𝑬𝒔𝒕𝒊𝒎𝒂𝒕𝒆 𝑡ℎ𝑒 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑠𝑒𝑛𝑠𝑜𝑟 𝑛𝑜𝑑𝑒 𝑎𝑛𝑑 𝑚𝑜𝑏𝑖𝑙𝑒 𝑠𝑖𝑛𝑘 𝑢𝑠𝑖𝑛𝑔 𝑅𝑆𝑆𝐼
16. 𝑭𝒊𝒏𝒅 𝑜𝑤𝑛 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑇𝑖 ← 𝑇 + 𝛥𝑇 using (6)
17. 𝑪𝒂𝒍𝒄𝒖𝒍𝒂𝒕𝒆 𝑝𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑓𝑎𝑖𝑙𝑢𝑟𝑒 𝑝𝑖 ←
𝑇𝑖
𝑇𝐻𝑖
18. 𝑪𝒂𝒍𝒄𝒖𝒍𝒂𝒕𝒆 𝐸𝑛𝑡𝑟𝑜𝑝𝑦 𝐻(𝑝𝑖) = −𝑝𝑖𝑙𝑜𝑔2𝑝𝑖
19. 𝑪𝒂𝒍𝒄𝒖𝒍𝒂𝒕𝒆 𝐸𝑛𝑡𝑟𝑜𝑝𝑦 𝑜𝑓 𝑛𝑒𝑖𝑔ℎ𝑏𝑜𝑟ℎ𝑜𝑜𝑑 𝑠𝑢𝑏𝑔𝑟𝑎𝑝ℎ 𝑎𝑠 𝐻(𝑆𝐺𝑖 ) ← ∑ 𝐻(𝑆𝐺𝑖 )
𝑘 , 𝑘 ∈ 𝐶𝑖
20. 𝑭𝒊𝒏𝒅 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 thermal entropy using (8)
21. 𝑭𝒊𝒏𝒅 𝑁𝑜𝑑𝑒 𝑔𝑎𝑖𝑛 𝑑𝑒𝑔𝑟𝑒𝑒 𝑎𝑛𝑑 𝑡𝑜𝑡𝑎𝑙 𝑒𝑛𝑒𝑟𝑔𝑦 𝑢𝑠𝑖𝑛𝑔 (10)𝑎𝑛𝑑 (11)
22. 𝑭𝒊𝒏𝒅 𝐸𝑛𝑒𝑟𝑔𝑦 𝑊𝑒𝑙𝑓𝑎𝑟𝑒 (𝐸𝑊) 𝑢𝑠𝑖𝑛𝑔 (23)
23. 𝑪𝒂𝒍𝒄𝒖𝒍𝒂𝒕𝒆 𝐿𝑖𝑛𝑘 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑣𝑖𝑡𝑦 𝑓𝑜𝑟 𝑒𝑎𝑐ℎ 𝑜𝑝𝑝𝑜𝑟𝑡𝑢𝑛𝑖𝑠𝑡𝑖𝑐 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛 𝑢𝑠𝑖𝑛𝑔 (16)
24. 𝑭𝒊𝒏𝒅 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑂𝑝𝑡𝑖𝑚𝑎𝑙 𝐻𝑜𝑝𝑠 (𝐸𝑂𝐻) 𝑢𝑠𝑖𝑛𝑔 (22)
25. 𝑫𝒆𝒕𝒆𝒓𝒎𝒊𝒏𝒆 𝑡ℎ𝑒 𝐿𝑖𝑛𝑘 𝑄𝑢𝑎𝑙𝑖𝑡𝑦 𝐹𝑎𝑐𝑡𝑜𝑟 (𝐿𝑄𝐹)
26. 𝑭𝒐𝒓𝒎 𝑪𝒓𝒊𝒕𝒆𝒓𝒊𝒂 = {𝐸𝑇(𝑉𝑖), 𝐸𝑊, 𝐻𝑟𝑒𝑙
𝑡ℎ𝑒𝑟𝑚
, 𝐿𝑉𝑖𝑉𝑗
, 𝐸𝑂𝐻, 𝐿𝑄𝑅 }
27. 𝑵𝒐𝒓𝒎𝒂𝒍𝒊𝒛𝒆 𝑡ℎ𝑒 𝑣𝑎𝑙𝑢𝑒𝑠 𝑖𝑛 𝑪𝒓𝒊𝒕𝒆𝒓𝒊𝒂
28. 𝑼𝒔𝒆 𝑡𝑟𝑖𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑓𝑢𝑧𝑧𝑦 𝑚𝑒𝑚𝑏𝑒𝑟𝑠ℎ𝑖𝑝 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛 𝑎𝑛𝑑 𝐺𝐶𝐹 𝑡𝑜 𝑔𝑒𝑛𝑒𝑟a𝑡𝑒 𝑒𝑞𝑢𝑎𝑡𝑖𝑜𝑛 (27)
29. 𝑼𝒔𝒆 𝑡𝑟𝑎𝑛𝑠𝑓𝑜𝑟𝑚𝑎𝑡𝑖𝑜𝑛 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛 𝐸𝐺𝐶𝐹 𝑡𝑜 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒 𝑡ℎ𝑒 𝐻𝐹𝐿𝑇𝑆 i.e. 𝐻𝑆
30. 𝑼𝒔𝒆 1 − 𝑐𝑢𝑡 𝐻𝐹𝐿𝑇𝑆 𝑡𝑜 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑒 𝑡ℎ𝑒 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 𝑚𝑎𝑡𝑟𝑖𝑥 𝑌
31. 𝑺𝒕𝒂𝒕𝒆 = 𝑆𝑡𝑎𝑡𝑒_𝐸𝑣𝑎𝑙𝑢𝑎𝑡𝑖𝑜(𝐸𝑇𝑖𝑛𝑖𝑡𝑖𝑎𝑙
, 𝐸𝑇𝑟𝑒𝑚𝑎𝑖𝑛
, 𝐸𝑊𝑛𝑜𝑟𝑚𝑎𝑙𝑖𝑧𝑒𝑑,𝐸𝑊𝑚𝑎𝑥)
32. 𝑰𝒇 𝑠𝑡𝑎𝑡𝑒 = 𝑶𝒑𝒕𝒊𝒎𝒊𝒔𝒕𝒊𝒄
33. 𝐴𝒄𝒕𝒊𝒐𝒏 = 𝑹𝒆𝒕𝒂𝒊𝒏𝑭𝒖𝒏𝒄(𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒𝑠, 𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎,𝑌)

29
35. 𝑰𝒇 𝑠𝑡𝑎𝑡𝑒 = 𝑷𝒆𝒔𝒔𝒊𝒎𝒊𝒔𝒕𝒊𝒄
36. 𝐴𝒄𝒕𝒊𝒐𝒏 = 𝑪𝒉𝒂𝒏𝒈𝒆𝑭𝒖𝒏𝒄(𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒𝑠, 𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎,𝑌)
The set of required actions of a sensor node is known as alternatives which can be denoted as
{‘CH’, ’CM’, ’Relay’} and the suitable action chosen by the sensor node 𝑖 from alternatives is
based on the criteria defined in equation (25) i.e.,
𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎 =
{
𝑁𝑜𝑑𝑒 𝑔𝑎𝑖𝑛 𝑑𝑒𝑔𝑟𝑒𝑒
𝐸𝑛𝑒𝑟𝑔𝑦 𝑊𝑒𝑙𝑓𝑎𝑟𝑒
𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑡ℎ𝑒𝑟𝑚𝑎𝑙 𝑒𝑛𝑡𝑟𝑜𝑝𝑦
𝐿𝑖𝑛𝑘 𝐶𝑜𝑛𝑛
𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑂𝑝𝑡𝑖𝑚𝑎𝑙 𝐻𝑜𝑝
𝐿𝑖𝑛𝑘 𝑞𝑢𝑎𝑙𝑖𝑡𝑦 𝑓𝑎𝑐𝑡𝑜𝑟 }
= {𝐸𝑇(𝑉𝑖), 𝐸𝑊, 𝐻𝑟𝑒𝑙
𝑡ℎ𝑒𝑟𝑚
, 𝐿𝑉𝑖𝑉𝑗
, 𝐸𝑂𝐻, 𝐿𝑄𝑅} (25)
And the corresponding weights assigned to the members defining the criteria will be 𝑤𝑇 =
{𝑤1, 𝑤2, … 𝑤|𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎|}, |𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎| is the cardinality of Criteria. Now we assume our linguistic
term set as,
𝑆 =
{
𝑠1:𝐸𝑥𝑡𝑟𝑒𝑚𝑒𝑙𝑦 𝑙𝑜𝑤(𝑒𝑙)
𝑠2:𝑣𝑒𝑟𝑦 𝑙𝑜𝑤(𝑣𝑙)
𝑠3:𝑙𝑜𝑤(𝑙)
𝑠4:𝑚𝑒𝑑𝑖𝑢𝑚(𝑚)
𝑠5:ℎ𝑖𝑔ℎ(ℎ)
𝑠6:𝑣𝑒𝑟𝑦 ℎ𝑖𝑔ℎ(𝑣ℎ)
𝑠7:𝑝𝑒𝑟𝑓𝑒𝑐𝑡(𝑝) }
(26)
The normalized attribute values after data standardization in the hesitant fuzzy set are converted
to linguistic term set S using triangular membership function as depicted from Figure 2. A
context-free grammar 𝐺𝐶𝐹 [7] has been used to produce linguistic terms for the alternatives
against different values of the criteria. These HFS membership values are then transformed into
HFLTS using a transformation function 𝐸𝐺𝐶𝐹 as shown in equation (27) and (28).
Figure 1. Linguistic term set conversion using triangular membership function
𝐻 =
𝑥1
𝑥2
𝑥3
[
𝑐1 𝑐2 𝑐3
𝑔𝑟𝑒𝑎𝑡𝑒𝑟 𝑡ℎ𝑎𝑛 ℎ 𝑔𝑟𝑒𝑎𝑡𝑒𝑟 𝑡ℎ𝑎𝑛 ℎ 𝑙𝑜𝑤𝑒𝑟 𝑡ℎ𝑎𝑛 𝑚
𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑙 𝑎𝑛𝑑 ℎ 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑣𝑙 & ℎ 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑙 𝑎𝑛𝑑 𝑣ℎ
𝑔𝑟𝑒𝑎𝑡𝑒𝑟 𝑡ℎ𝑎𝑛 𝑙 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑣𝑙 & ℎ 𝑙𝑜𝑤𝑒𝑟 𝑡ℎ𝑎𝑛 ℎ
𝑐4 𝑐5 𝑐6
𝑔𝑟𝑒𝑎𝑡𝑒𝑟 𝑡ℎ𝑎𝑛 ℎ 𝑙𝑜𝑤𝑒𝑟 𝑡ℎ𝑎𝑛 𝑚 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 ℎ 𝑎𝑛𝑑 𝑝
𝑔𝑟𝑒𝑎𝑡𝑒𝑟 𝑡ℎ𝑎𝑛 𝑚 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑙 𝑎𝑛𝑑 𝑣ℎ 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑚 𝑎𝑛𝑑 𝑣ℎ
𝑔𝑟𝑒𝑎𝑡𝑒𝑟 𝑡ℎ𝑎𝑛 𝑚 𝑙𝑜𝑤𝑒𝑟 𝑡ℎ𝑎𝑛 𝑚 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 ℎ 𝑎𝑛𝑑 𝑝
] (27)
where, 𝑐1 = 𝑁𝑜𝑑𝑒 𝑔𝑎𝑖𝑛 𝑑𝑒𝑔𝑟𝑒𝑒 , 𝑐2 = 𝐸𝑛𝑒𝑟𝑔𝑦 𝑊𝑒𝑙𝑓𝑎𝑟𝑒, 𝑐3 = 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑡ℎ𝑒𝑟𝑚𝑎𝑙 𝑒𝑛𝑡𝑟𝑜𝑝𝑦, 𝑐4 =
𝐿𝑖𝑛𝑘 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑣𝑖𝑡𝑦, 𝑐5 = 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑂𝑝𝑡𝑖𝑚𝑎𝑙 𝐻𝑜𝑝, 𝑐6 = 𝐿𝑖𝑛𝑘 𝑞𝑢𝑎𝑙𝑖𝑡𝑦 𝑓𝑎𝑐𝑡𝑜𝑟, 𝑥1 = CH state, 𝑥2 = CM

30
state, 𝑥3 = Relay state. Subsequently, the decision matrix 𝐷𝑀 is then converted to HFLTS
by using a transformation function 𝐸𝐺𝐶𝐹.
𝐻 = 𝑥1
𝑥2
𝑥3
[
𝑐1 𝑐2 𝑐3
{𝑣ℎ, 𝑝} {𝑣ℎ, 𝑝} {𝑒𝑙, 𝑣𝑙, 𝑙}
{𝑙, 𝑚, ℎ} {𝑣𝑙, 𝑙, 𝑚 ℎ} { 𝑙, 𝑚, ℎ, 𝑣ℎ}
{𝑚, ℎ, 𝑣ℎ, 𝑝} {𝑣𝑙, 𝑙, 𝑚 ℎ} {𝑒𝑙, 𝑣𝑙, 𝑙, 𝑚}
𝑐4 𝑐5 𝑐6
{𝑣ℎ, 𝑝} {𝑒𝑙, 𝑣𝑙, 𝑙} {ℎ, 𝑣ℎ, 𝑝}
{ℎ, 𝑣ℎ, 𝑝} {𝑙, 𝑚, ℎ, 𝑣ℎ} {𝑚, ℎ, 𝑣ℎ}
{ℎ, 𝑣ℎ, 𝑝} {𝑒𝑙, 𝑣𝑙, 𝑙} {ℎ, 𝑣ℎ, 𝑝}
] (28)
The decision matrix 𝐷𝑀 includes the members ℎ𝑖𝑗 where 𝑖 ∈ {𝑥1, 𝑥2, 𝑥3} and 𝑗 ∈
{𝑐1, 𝑐2, 𝑐3, 𝑐4, 𝑐5, 𝑐6}. According to the definition of hesitant fuzzy linguistic term set, we can
easily calculate the envelope of its members ℎ𝑖𝑗 using upper bound and lower bound rules.
Accordingly, the new decision matrix Y containing the envelopes of H is given in equation (29)
as,
𝑌 = 𝑥1
𝑥2
𝑥3
[
𝑐1 𝑐2 𝑐3
{𝑣ℎ, 𝑝} {𝑣ℎ, 𝑝} {𝑒𝑙, 𝑙}
{𝑙, ℎ} {𝑣𝑙, ℎ} { 𝑙,𝑣ℎ}
{𝑚, 𝑝} {𝑣𝑙, ℎ} {𝑒𝑙, 𝑚}
𝑐4 𝑐5 𝑐6
{𝑣ℎ, 𝑝} {𝑒𝑙, 𝑙} {ℎ, 𝑝}
{ℎ, 𝑝} {𝑙, 𝑣ℎ} {𝑚,𝑣ℎ}
{ℎ, 𝑝} {𝑒𝑙, 𝑙} {ℎ, 𝑝}
] (29)
Now we utilize the node gain degree and energy welfare to classify the status of a node as
‘Optimistic’ and ‘Pessimistic’. The ‘RetainFunc’ will be called if the status of a node is evaluated
as ‘Optimistic’ and ‘ChangeFunc’ will be called if the status of a node is evaluated as
‘Pessimistic’ depending upon the status evaluation criteria already discussed.
Function: RetainFunc (𝑨𝒍𝒕𝒆𝒓𝒏𝒂𝒕𝒊𝒗𝒆𝒔, 𝑪𝒓𝒊𝒕𝒆𝒓𝒊𝒂, 𝒀)
1. 𝑭𝒐𝒓 𝑖 = 1 𝑡𝑜 |𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒𝑠| of 𝑌,
2. 𝑭𝒐𝒓 𝒋 = 1 𝑡𝑜 |𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎| of 𝑌
3. Get 1-cut hesitant fuzzy set 𝐻𝑆
𝑗
(𝑥𝑖)𝛼=1 for 𝐻𝑆
𝑗
(𝑥𝑖),
where, 𝐻𝑆
𝑗
(𝑥𝑖)𝛼=1 = [{𝐻𝑆−
𝑗
(𝑥𝑖)𝛼=1, 𝐻𝑆+
𝑗
(𝑥𝑖)𝛼=1}]
4. 𝑬𝒏𝒅 𝑭𝒐𝒓
6. 𝑭𝒐r 𝒆 = 𝟏 𝒕𝒐 |𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒𝑠| of 𝑌
7. 𝑭𝒐𝒓 𝒇 = 𝟏 𝒕𝒐 |𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎| of 𝑌
8. 𝑮𝒆𝒕 𝑡ℎ𝑒 𝑖𝑛𝑡𝑒𝑟𝑣𝑎𝑙𝑠 𝐼𝑚𝑎𝑥(𝑥𝑒 )𝑓𝑜𝑟 𝑒𝑎𝑐ℎ 𝑎𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒 𝑥𝑖 𝑤𝑖𝑡ℎ 𝑟𝑒𝑠𝑝𝑒𝑐𝑡 𝑡𝑜 𝑒𝑎𝑐ℎ
𝑐𝑟𝑖𝑡𝑒𝑟𝑖𝑜𝑛 𝑓; 𝑤ℎ𝑒𝑟e,𝐼𝑚𝑎𝑥(𝑥𝑒 ) = [𝑀𝑎𝑥(𝐻𝑆−
𝑓
(𝑥𝑖)𝛼=1), 𝑀𝑎𝑥(𝐻𝑆+
𝑓
(𝑥𝑖)𝛼=1)] 𝑓∈𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎 = [𝑢𝑒1
𝑚𝑎𝑥
,
𝑢𝑒2
𝑚𝑎𝑥
]
9. 𝑅𝑎𝑛𝑘𝑒1
𝑜𝑝𝑡𝑖
= max (1 − 𝑚𝑎𝑥 (
1−𝑢𝑖1
𝑚𝑎𝑥
𝑢𝑖2
𝑚𝑎𝑥 −𝑢𝑖1
𝑚𝑎𝑥+1
, 0) , 0)
12. 𝑹𝒆𝒕𝒖𝒓𝒏 𝑚𝑎𝑥(𝑅𝑎𝑛𝑘𝑒1
𝑜𝑝𝑡𝑖
)
Function: ChangeFunc (𝑨𝒍𝒕𝒆𝒓𝒏𝒂𝒕𝒊𝒗𝒆𝒔,𝑪𝒓𝒊𝒕𝒆𝒓𝒊𝒂, 𝒀) 𝑨𝒍𝒈𝒐𝒓𝒊𝒕𝒉𝒎:
1. 𝑭𝒐𝒓 𝑖 = 1 𝑡𝑜 |𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒𝑠|of 𝑌
2. 𝑭𝒐𝒓 𝑗 = 1 𝑡𝑜|𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎| of 𝑌
3. Get 1-cut hesitant fuzzy set HS
j
(𝑥𝑖)𝛼=1 for HS
j
(𝑥𝑖),
where, HS
j
(𝑥𝑖)𝛼=1 = [{HS−
j
(𝑥𝑖)𝛼=1, HS+
j
(𝑥𝑖)𝛼=1}]
6. 𝑭𝒐𝒓 𝑒 = 1 𝑡𝑜|𝐴𝑙𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑣𝑒𝑠| of 𝑌
7. 𝑭𝒐𝒓 𝑓 = 1 𝑡𝑜 |𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎| of 𝑌
8. Get the intervals 𝐼𝑚𝑖𝑛(𝑥𝑒) for each alternative 𝑥𝑖 for each criterion f;

31
where 𝐼𝑚𝑖𝑛(𝑥𝑒 ) = [𝑀𝑎𝑥(HS−
f
(𝑥𝑖)𝛼=1), 𝑀𝑎𝑥(HS+
f
(𝑥𝑖)𝛼=1)] 𝑓∈𝐶𝑟𝑖𝑡𝑒𝑟𝑖𝑎 =
[ue1
max
, ue2
max
]
9. Ranke1
pessi
= max (1 − 𝑚𝑎𝑥 (
1−ui1
max
ui2
max −ui1
max+1
, 0) , 0)
12. 𝑹𝒆𝒕𝒖𝒓𝒏 𝑚𝑎𝑥(Ranke1
pessi
)
The ‘RetainFunc’ and ‘ChangeFunc’ functions applies the 1-cut HFLTS to fuzzy sets in 𝑌 to
generate the envelope for each criteria against every alternative and calculates the probabilistic
ranking of the alternatives based on the interval calculated from the envelopes. For instance, if
the probabilistic ranking of alternatives is [𝑥1>𝑥3 > 𝑥2], it indicates that the corresponding
sensor node in its current state will probably retain its state and perform the role of a CH instead
of CM or Relay node. But if the probabilistic ranking of alternatives is [𝑥2>𝑥3 > 𝑥1] or [𝑥3>𝑥2 >
𝑥1], it indicates that the sensor node’s preferred action will be to change its role as CM or relay
instead of CH.
6. RESULTS
6.1. Simulation Environment
We have evaluated the performance of FLOC in MATLAB 2019b and OMNET++ using cross-
platform library (MEX-API). This Application Programming Interface (API) can provide the user
an easy bidirectional connection interface between MATLAB and OMNET++. Nodes are
arranged in random topology. We have utilized low rate, low cost, short-range, flexible and low
power consumption standard IEEE 802.15.4 for our PHY and MAC layer. The performance
metrics like active node ratio, average energy consumption, and packet delivery ratio are
analyzed against parametric benchmarks viz. node density and temperature variation. The
performance of FLOC is compared with three different approaches i.e. 1) SOPC [8], 2) BMAC
[14], 3) RL-Sleep [7]. Stop-Operate Power-Control (SOPC) is a temperature-aware asynchronous
sleep-scheduling algorithm in which energy, link connectivity and network coverage are
preserved by putting a few sensor nodes in hibernation mode and controlling the rest of the
sensor nodes’ transmission power. The communication range and a number of active nodes are
adjusted to maintain the critical density for consistent connectivity in the network. Berkley-MAC
(BMAC) is a low-traffic, low-power-consuming MAC protocol based on adaptive preamble
sampling for duty cycling to preserve energy, provide effective collision avoidance and high
channel utilization. RL-Sleep is an asynchronous reinforcement learning based procedure based
on the adaptive state transition determined by sensor nodes. The state transition is based on
temperature sensing and collecting information from the neighbourhood. The effect of various
parameters on the performance of FLOC with other existing benchmarks is provided in this
section.
Table 2. Simulation parameters
Parameter Value
Deployment area 500 m X 500 m
𝑁 60-90
𝑇𝐻 80o
C
𝑆𝑆𝑈𝑁
𝑚𝑎𝑥
1

32
Maximum Temperature 80o
C
𝐴𝑟𝑒𝑎𝑠𝑒𝑛 20cm2
𝑑0 20m
𝑅 200m
𝜀𝑓𝑠 50nJ/bit/m2
𝜀𝑚𝑝 10pJ/bit/m2
𝐸𝑒𝑙𝑒𝑐 50nJ/bit
Initial Energy for nodes 5J(for neighbors of sink node)
3J (For other nodes)
𝑛𝑝 2
𝑐𝑝 0.5
𝑚𝑎𝑠𝑠 50g
𝑟 0.25
Number of packets 1024
Length of packet 8000bits
6.1.1. Active node ratio
Figure (3) depicts the active node ratio comparison of FLOC with SOPC, BMAC, and RL-Sleep.
It is evident from the figure that the ratio of an average number of active nodes to total number of
sensor nodes in the network is higher for FLOC than in any other benchmark. Furthermore, the
active node ratio for all approaches is optimum for N=80. We also evaluated the performance of
FLOC against SOPC, BMAC and RL-Sleep for varying diurnal temperatures. Figure (4) shows
the comparison of active node ratio of FLOC and other benchmarks for diurnal temperature
variations. It has been observed that FLOC outperforms all three approaches in terms of active
node ratio. The number of active sensor nodes in the network varies inversely with the diurnal
temperature.
Figure 2. Active node ratio against node density

33
Figure 3. Active node ratio for diurnal temperature variations
6.1.2. Average energy consumption
Figure (5) shows the performance comparison of FLOC with SOPC, BMAC, and RL-Sleep in
terms of average energy consumption. BMAC outperforms all other algorithms due to its
adaptive preamble strategy and short duty cycle which play a significant role in preserving
energy. The adaptive adjustment of temperature with respect to communication range leverages
higher energy consumption for SOPC. FLOC performs better than SOPC and RL-Sleep but
exhibits a higher amount of energy consumption against BMAC due to packet broadcasting in the
neighbourhood. Figure (6) depicts the average energy consumption of FLOC against other
approaches for diurnal temperature variation. FLOC and BMAC exhibit almost similar profile for
average energy consumed whereas SOPC and RL-Sleep consumed higher amount of energy for
N=80.
Figure 4. Average energy consumption against node density

34
Figure 5. Average energy consumption for diurnal temperature variations
6.1.3. Packet Delivery Ratio (PDR)
Figure (7) depicts the comparison of FLOC with existing benchmarks in terms of PDR. FLOC
outperforms other approaches in the case of PD. Due to its opportunistic and environment
adaptive sleep scheduling strategy, the additional power loss in FLOC can be compensated due to
control packet overhead. BMAC shows the worst performance against existing benchmarks.
Figure (8) shows the PDR of FLOC with other approaches for diurnal temperature variations.
FLOC leverages a better packet delivery ratio in comparison to other techniques. It is pertinent to
mention that the PDR of FLOC decreases with the increase in diurnal temperature.
Figure 6. Packet delivery ratio against node density
6.1.4. Network Lifetime
Network lifetime can be evaluated in terms of the dead node ratio in the network. The time when
25% or 50% of the nodes present in the network have no residual energy left to continue their
data delivery tasks can be treated as the network lifetime [10]. Here, we considered two such time
intervals, i.e., (i) when the first sensor node dies (FND) and (ii) when half the nodes in the
network die, i.e., 50% of the sensor nodes have no residual energy left to continue their data

35
sensing tasks (HND). The stacked bar chart in Figure (9) demonstrates the network lifetime of
FLOC with SOPC, BMAC, and RL-Sleep against the node density in terms of FND and HND. In
every stack, we have four bars representing four different schemes, i.e., SOPC, BMAC, RL-Sleep
and FLOC. In each bar, we have two groups, i.e., FND and HND. It can be observed from the
figure (9) that our proposed scheme FLOC outperforms other schemes for both FND and HND
scenarios. The best performance is achieved when number of nodes in the network are 80.
Figure 7. Packet delivery ratio for diurnal temperature variations
Figure 8. Network lifetime against node density

36
Figure 9. Network lifetime in terms of remaining energy against number of rounds
Similarly, we can also observe the network lifetime against a number of communication rounds
in terms of total remaining energy of sensor nodes in Figure (10). Dense deployment of the nodes
provides a healthy neighborhood which drives the proposed technique to perform better than
others. SOPC outperforms RL-Sleep as it completely depends on its perception of the
environment. FLOC, on the other hand, performs better than others as it adapts itself to the status
of the neighborhood.
7. CONCLUSIONS
In this paper, a novel, distributed, FLOC algorithm is proposed based on the hesitant fuzzy
linguistic term set (HFLTS) analysis in order to resolve the CH decision-making problems and
network lifetime bottlenecks using a dynamic network architecture involving opportunistic
clustering. The attributes such as energy transfer-based opportunistic routing, energy welfare,
relative thermal entropy; expected optimal hops and link quality factor are utilized to form the
criteria for Hesitant Fuzzy Linguistic Term Set and make a decision about the contemporary role
of the node based on its current state. The effectiveness of FLOC is confirmed after carefully
analyzing and evaluating its performance against several existing benchmarks. The simulation
results have clearly shown that employing FLOC algorithm results in the improvement of the
active node ratio, average energy consumption and packet delivery ratio. The possible future
work would be to perform the hesitant fuzzy linguistic term set analysis for harvested energy
scavenging and transfer capabilities in opportunistic clustering.
ACKNOWLEDGMENTS
The authors would like to thank all the reviewers for their insightful comments and deliberate
suggestions for improving the quality of this paper. This work was supported in part by the
National Key Research and Development Program of China under Grant 2020YFB1901900, and
in part by the National Natural Science Foundation of China (NSFC) under Grant 51776050 and
Grant 51536001.
CONFLICTS OF INTEREST
The author declares no conflict of interest.

37
REFERENCES
[1] O. Ogundile, A. Alfa, “A Survey on an Energy-Efficient and Energy-Balanced Routing Protocol for
Wireless Sensor Networks,” Sensors, vol. 17, no. 1084, 2017.
[2] M. S. Manshahia, “Wireless Sensor Networks: A Survey,” IJSER, vol. 74, p. 710–716, 2016.
[3] A. Boukerche and A. Darehshoorzadeh, ‘‘Opportunistic routing in wireless networks: Models,
algorithms, and classifications,’’ ACM Comput. Surv., vol. 47, no. 2, pp. 1–36, Jan. 2015, doi:
10.1145/2635675.
[4] J. Luo, J. Hu, D. Wu, and R. Li, ‘‘Opportunistic routing algorithm for relay node selection in wireless
sensor networks,’’ IEEE Trans. Ind. Informat., vol. 11, no. 1, pp. 112–121, Feb. 2015, doi:
10.1109/TII.2014.2374071.
[5] G. Yang, Z. Peng, and X. He, ‘‘Data collection based on opportunistic node connections in wireless
sensor networks,’’ Sensors, vol. 18, no. 11, p. 3697, Oct. 2018, doi: 10.3390/s18113697.
[6] Anees, Zhang, Baig, and Lougou, ‘‘Energy-efficient multi-disjoint path opportunistic node
connection routing protocol in wireless sensor networks for smart grids,’’ Sensors, vol. 19, no. 17, p.
3789, Sep. 2019, doi: 10.3390/s19173789.
[7] Banerjee, Partha Sarathi, Satyendra Nath Mandal, Debashis De, and Biswajit Maiti. "RL-sleep:
Temperature adaptive sleep scheduling using reinforcement learning for sustainable connectivity in
wireless sensor networks." Sustainable Computing: Informatics and Systems, vol. 26, no. 100380,
2020.
[8] A. Bachir, W. Bechkit, Y. Challal and A. Bouabdallah, "Joint Connectivity-Coverage Temperature-
Aware Algorithms for Wireless Sensor Networks," in IEEE Transactions on Parallel and Distributed
Systems, vol. 26, no. 7, pp. 1923-1936, 1 July 2015, doi: 10.1109/TPDS.2014.2331063.
[9] R. M. Rodriguez, L. Martinez and F. Herrera, "Hesitant Fuzzy Linguistic Term Sets for Decision
Making," in IEEE Transactions on Fuzzy Systems, vol. 20, no. 1, pp. 109-119, Feb. 2012, doi:
10.1109/TFUZZ.2011.2170076.
[10] J. Anees, H.-C. Zhang, B. G. Lougou, S. Baig, and Y. G. Dessie, ‘‘Delay aware energy-efficient
opportunistic node selection in restricted routing,’’ Comput. Netw., vol. 181, Nov. 2020, Art. no.
107536, doi: 10.1016/j.comnet.2020.107536.
[11] C. Tunca, S. Isik, M. Y. Donmez and C. Ersoy, "Ring Routing: An Energy-Efficient Routing Protocol
for Wireless Sensor Networks with a Mobile Sink," in IEEE Transactions on Mobile Computing, vol.
14, no. 9, pp. 1947-1960, 1 Sept. 2015, doi: 10.1109/TMC.2014.2366776.
[12] M. Mukherjee, L. Shu, L. Hu, G. P. Hancke, C. Zhu, “Sleep Scheduling in Industrial Wireless Sensor
Networks for Toxic Gas Monitoring,” IEEE Wirel. Commun. vol. 99, p. 2–8, 2017.
[13] Ye, Wei, John Heidemann, and Deborah Estrin, "Medium access control with coordinated adaptive
sleeping for wireless sensor networks," IEEE/ACM Transactions on Networking, vol. 12, no.3, p.
493- 506, 2004.
[14] Polastre, Joseph, Jason Hill, and David Culler, "Versatile low power media access for wireless sensor
networks," Proceedings of the 2nd international conference on Embedded networked sensor systems,
2004.
[15] P. S. Banerjee, S. N. Mandal, D. De et al, “iSleep: thermal entropy aware intelligent sleep scheduling
algorithm for wireless sensor network,” MicrosystTechnol, vol. 26, p. 2305–2323, 2020. DOI:
https://doi.org/10.1007/s00542-019-04706-7.
[16] Mo, Xiaoyi, Hua Zhao, and Zeshui Xu, "Feature-based hesitant fuzzy aggregation method for
satisfaction with life scale," in Applied Soft Computing, vol. 94, no. 106493,2020.
[17] P. Musílek, P. Krömer and T. Barton, "E-BACH: Entropy-Based Clustering Hierarchy for Wireless
Sensor Networks," 2015 IEEE/WIC/ACM International Conference on Web Intelligence and
Intelligent Agent Technology (WI-IAT), 2015, pp. 231-232, doi: 10.1109/WI-IAT.2015.88.
[18] W. Liang, M. Goh, Y. M. Wang, “Multi-attribute group decision making method based on prospect
theory under hesitant probabilistic fuzzy environment,” in Computers & Industrial Engineering, vol.
149, no. 106804, 2020.
[19] J. Anees, H.-C. Zhang, S. Baig, B. Guene Lougou, and T. G. Robert Bona, ‘‘Hesitant fuzzy entropy-
based opportunistic clustering and data fusion algorithm for heterogeneous wireless sensor
networks,’’ Sensors, vol. 20, no. 3, p. 913, Feb. 2020, doi: 10.3390/s20030913.
[20] M. M. Afsar, M. Younis, “A load-balanced cross-layer design for energy-harvesting sensor
networks,” J. Netw. Comput. Appl., vol. 145, pp. 102390, 2019. doi: 10.1016/j.jnca.2019.06.010.

38
[21] L. R. Varshney, ‘‘Transporting information and energy simultaneously,’’ in Proc. IEEE Int. Symp.
Inf. Theory, Toronto, ON, Canada, Jul. 2008, pp. 1612–1616, doi: 10.1109/ISIT.2008.4595260.
[22] S. Guo, F. Wang, Y. Yang and B. Xiao, "Energy-Efficient Cooperative Tfor Simultaneous Wireless
Information and Power Transfer in Clustered Wireless Sensor Networks," in IEEE Transactions on
Communications, vol. 63, no. 11, pp. 4405-4417, Nov. 2015, doi: 10.1109/TCOMM.2015.2478782.
[23] X. Zhou, R. Zhang, and C. K. Ho, ‘‘Wireless information and power transfer: Architecture design and
rate-energy tradeoff,’’ IEEE Trans. Commun., vol. 61, no. 11, pp. 4754–4767, Nov. 2013, doi:
10.1109/TCOMM.2013.13.120855.
[24] J. Anees, H. -C. Zhang, B. G. Lougou, S. Baig, Y. G. Dessie and Y. Li, "Harvested Energy
Scavenging and Transfer capabilities in Opportunistic Ring Routing," in IEEE Access, vol. 9, pp.
75801-75825, 2021, doi: 10.1109/ACCESS.2021.3082296.
[25] G. Han, J. Jiang, C. Zhang, T. Q. Duong, M. Guizani, and G. K. Karagiannidis, ‘‘A survey on mobile
anchor node assisted localization in wireless sensor networks,’’ IEEE Commun. Surveys Tuts., vol.
18, no. 3, pp. 2220–2243, 3rd Quart., 2016, doi: 10.1109/comst.2016.2544751.
[26] L. Karim, N. Nasser, and T. El Salti, ‘‘RELMA: A range free localization approach using mobile
anchor node for wireless sensor networks,’’ in Proc. IEEE Global Telecommun. Conf. GLOBECOM,
Miami, FL, USA, Dec. 2010, pp. 1–5, doi: 10.1109/glocom.2010.5683802.
[27] A. Gopakumar and L. Jacob, ‘‘Localization in wireless sensor networks using particle swarm
optimization,’’ in Proc. IET Conf. Wireless, Mobile Multimedia Netw., Beijing, China, 2008, pp.
227–230, doi: 10.1049/cp:20080185.
[28] X. Li, J. Yang, A. Nayak, and I. Stojmenovic, ‘‘Localized geographic routing to a mobile sink with
guaranteed delivery in sensor networks,’’ IEEE J. Sel. Areas Commun., vol. 30, no. 9, pp. 1719–
1729, Oct. 2012, doi: 10.1109/jsac.2012.121016.
[29] RodríGuez, Rosa M., Luis MartıNez, and Francisco Herrera, "A group decision making model
dealing with comparative linguistic expressions based on hesitant fuzzy linguistic term sets,"
Information Sciences, vol. 241, pp. 28-42, 2013.
[30] Lee, Li-Wei, and Shyi-Ming Chen, "Fuzzy decision making based on likelihood-based comparison
relations of hesitant fuzzy linguistic term sets and hesitant fuzzy linguistic operators." Information
Sciences, vol. 294, pp. 513-529, 2015.
[31] Lotfi A Zadeh, "Fuzzy logic—a personal perspective." Fuzzy sets and systems, vol. 281, pp. 4-20,
2015.
[32] Torra, Vicenç, "Hesitant fuzzy sets." International Journal of Intelligent Systems, vol. 25, no. 6 pp.
529- 539, 2010.
[33] Chen Shyi-Ming, and Jia-An Hong, "Multicriteria linguistic decision making based on hesitant fuzzy
linguistic term sets and the aggregation of fuzzy sets," Information Sciences, vol. 286, pp. 63- 74,
2014.
[34] C. A. Boano, M. Zúñiga, J. Brown, U. Roedig, C. Keppitiyagama and K. Römer, "TempLab: A
testbed infrastructure to study the impact of temperature on wireless sensor networks," IPSN-14
Proceedings of the 13th International Symposium on Information Processing in Sensor Networks,
2014, pp. 95-106, doi: 10.1109/IPSN.2014.6846744.
[35] Shah, Babar, et al. "Guaranteed Lifetime Protocol for IoT based Wireless Sensor Networks with
Multiple Constraints." Ad Hoc Networks, no. 102158, 2020.
[36] K. Xiao, R. Wang, H. Deng, L. Zhang, C. Yang, “Energy-aware Scheduling for Information Fusion
in Wireless Sensor Network Surveillance,” Information Fusion, 2018. DOI:
https://doi.org/10.1016/j.inffus.2018.08.005
[37] H. Mostafaei, A. Montieri, V. Persico, A. Pescapé, “A sleep scheduling approach based on learning
automata for WSN partial coverage,” Journal of Network and Computer Applications, vol. 80, pp.
67-78, 2017.
[38] H. S. AbdelSalam and S. Olariu, "Toward Adaptive Sleep Schedules for Balancing Energy
Consumption in Wireless Sensor Networks," in IEEE Transactions on Computers, vol. 61, no. 10, pp.
1443-1458, Oct. 2012, doi: 10.1109/TC.2011.157.

39
AUTHORS
Junaid Anees received a B.S. degree from the Institute of Space Technology, Islamabad,
Pakistan in 2010 and an M.S. degree in Electrical Engineering from COMSATS
University Islamabad, Pakistan in 2015. Currently, he is a Ph.D. scholar in the School of
Energy Science & Engineering at Harbin Institute of Technology, China. He holds a
Senior Manager Position in Ground Segment Network Operations in Public Sector
Organization in Pakistan. His research interest includes Energy harvesting Wireless
Sensor Networks, Opportunistic Routing, Smart Grids, and Distributed Computing.
Prof. & Dr. Hao-Chun Zhang is currently the Head of the Department of Nuclear
Science and Engineering and executive professor of HIT-CORYS Nuclear System
Simulation International Joint Research Center (Sino-France). With BE in 1999, ME in
2001, and Ph.D. in 2007 from Harbin Institute of Technology (HIT), Dr. Zhang joined
HIT in September 2004. Dr. Zhang has about 200 research publications in peer-reviewed
journals and conferences, 5 books, and 2 translations of foreign books.

FLOC: HESITANT FUZZY LINGUISTIC TERM SET ANALYSIS IN ENERGY HARVESTING OPPORTUNISTIC CLUSTERING USING RELATIVE THERMAL ENTROPY AND RF ENERGY TRANSFER

More Related Content

Similar to FLOC: HESITANT FUZZY LINGUISTIC TERM SET ANALYSIS IN ENERGY HARVESTING OPPORTUNISTIC CLUSTERING USING RELATIVE THERMAL ENTROPY AND RF ENERGY TRANSFER (20)

More from IJCNCJournal (20)

Recently uploaded (20)

FLOC: HESITANT FUZZY LINGUISTIC TERM SET ANALYSIS IN ENERGY HARVESTING OPPORTUNISTIC CLUSTERING USING RELATIVE THERMAL ENTROPY AND RF ENERGY TRANSFER