Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497

Contents lists available at ScienceDirect

Journal of King Saud University –

Computer and Information Sciences
journal homepage: www.sciencedirect.com

Fuzzy logic based adaptive duty cycling for sustainability in energy

harvesting sensor actor networks
Sai Krishna Mothku ⇑, Rashmi Ranjan Rout
Computer Science and Engineering, National Institute of Technology Warangal, 506004, India

a r t i c l e i n f o a b s t r a c t

Article history: In an energy harvesting sensor actor network, a node recharges its battery from harvestable sources, such
Received 29 June 2018 as solar, wind and vibrations. Sustainability of the network till next recharge time is one of the most
Revised 12 September 2018 important challenges in harvesting sensor networks. In this paper, a fuzzy based adaptive duty cycling
Accepted 27 September 2018
algorithm has been proposed to achieve the network sustainability in harvesting sensor actor networks.
Available online 3 October 2018
In this work, current residual energy, predicted harvesting energy (for a futuristic time slot) and predicted
residual energy parameters are considered as fuzzy input variables to estimate duty cycle for a sensor
Keywords:
node. In this work, a harvesting model has been adopted to predict the harvesting energy. Further, resid-
Wireless sensor and actor networks
Energy harvesting nodes
ual energy has been estimated for future time slot using predicted harvesting energy, energy consump-
Network sustainability tion model and current residual energy. Simulation results are presented to show the efficacy of the
Duty cycle proposed mechanism by considering network sustainability metrics, such as number of rounds in which
Fuzzy logic network is connected, the round at which first node dies, maximum number of dead nodes and average
number of received packets at the actor node.
Ó 2018 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction
Wireless Sensor Actor Networks (WSANs) remain operational
for limited amount of time due to limited battery capacity of a sen-
sor node (Akyildiz and Kasimoglu, 2004). A long operational life-
time of WSAN is required for applications, such as monitoring
forest fires, nuclear plants, object tracking and military surveil-
lance (Akyildiz and Kasimoglu, 2004; Rout and Ghosh, 2013;
Imon et al., 2015; Rout et al., 2016; Guo et al., 2014; Elshrkawey
et al., 2018). Energy harvesting technology provides a long lasting
lifetime in WSAN. A node recharges battery from the charging
vehicles or from the natural resources, such as solar light, thermal,
wind and vibration (Guo et al., 2014; Wang et al., 2016; Tan and
Tang, 2017; Ye and Liang, 2017; Wu and Liu, 2013). A node is cap-
able of harvesting energy from sunlight. The volume of harvested
energy varies dynamically with the weather conditions over time
⇑ Corresponding author.
E-mail addresses: saikrishna@student.nitw.ac.in (S.K. Mothku), rashrr@nitw.ac.
in (R.R. Rout).
Peer review under responsibility of King Saud University.
Production and hosting by Elsevier
(Tan and Tang, 2017). The residual energy may reduce with usages
and time. This results in temporary disconnection of the nodes
from the network. The temporal-disconnected nodes may join into
the network in the next energy available time slot. Hence, the node
should sustain (survive) till next available period of the energy
source for improving the lifetime of the network. Node sustains till
the next recharge round by using the energy conservation tech-
niques, such as duty cycling, load balancing and data aggregation
(Imon et al., 2015).
The existing models (such as Guo et al., 2014; Wang et al., 2016)
are based on the wireless charging by a mobile charger. Wireless
charging technology may not be suitable for applications (like deep
forest) where a mobile charger cannot travel to replenish the
energy of sensor nodes. In this type of applications, solar energy
harvesting technology provides a long lasting lifetime to the net-
work. Further, the energy resources are typically dynamic and
uncontrolled (Aoudia et al., 2017). Achieving network sustainabil-
ity with solar based harvesting technology is one of the challenging
issues which has not been addressed adequately in the literature.
Duty cycle and load balancing techniques collectively improve
the network sustainability and throughput. Duty cycled wireless
sensor network is categorized into two types: random duty cycling
and coordinated duty cycling (fan Hsin et al., 2006). In the random
duty cycled wireless sensor network, the sensor nodes are turned
on and off randomly on their own. In a coordinated duty cycled

https://doi.org/10.1016/j.jksuci.2018.09.023
1319-1578/Ó 2018 The Authors. Production and hosting by Elsevier B.V. on behalf of King Saud University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1490 S.K. Mothku, R.R. Rout / Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497
wireless sensor network, the sensor nodes coordinate with each
other (via exchanging the information) to follow on and off (i.e.
active and sleep) schedule. The coordinated duty cycled WSN is
potentially efficient for communication in terms of synchroniza-
tion among nodes and network connectivity (fan Hsin et al.,
2006). In our work, a fuzzy based adaptive duty cycling algorithm
has been proposed to achieve the network sustainability in a tree-
based energy harvesting sensor actor network. Moreover, a predic-
tion model has been proposed to estimate energy consumption (i.e.
residual energy) for a future interval of time. A node estimates duty
cycle value using fuzzy logic system for a future time period by
considering residual energy, predicted harvesting energy (for a
future time slot) and predicted residual energy (for a future time
slot). Further, a switching decision has been taken by a node based
on the predicted duty cycle value of a node for survival. Coordi-
nated duty cycling is considered in this work. To achieve efficient
communication, active and sleep schedules are forwarded to the
children nodes.
The major contributions of the work are as follows:

Design of a fuzzy based adaptive duty cycling algorithm to
achieve sustainability in an energy harvesting sensor actor
network.

Prediction of duty cycle using fuzzy logic by considering current
residual energy, futuristic harvesting energy and residual
energy.

Simulation results are presented to show the efficacy of the pro-
posed mechanism to provide improved network sustainability
and successful data reception.
The rest of the work is organized as follows. Section 2 presents
the literature reviews of existing works, motivation of our work
and problem formulation. In Section 3.1, a network model is intro-
duced. In Section 3.2, the energy consumption has been computed
for future time period. In Section 3.3, energy consumption of the
network has been derived for the proposed network model. Fur-
ther, a fuzzy based adaptive duty cycle algorithm has been pro-
posed in Section 3.4. Section 3.5 presents the fuzzy logic system
procedure. Section 4 discusses the performance evaluations of
the proposed mechanism and finally, Section 5 concludes the
paper.
2. Related work
In this section, existing works on energy analysis models,
energy replenishment mechanisms and fuzzy logic system for
wireless sensor networks (WSN) are discussed.
In Bhardwaj et al. (2001), upper bounds on lifetime of wireless
sensor network has been derived. Rout et al. (2016) have analyzed
the energy consumptions for tree-based sensor network. Lee and
Lee (2010) have derived the network lifetime bound for a
cluster-based sensor network. In Rout and Ghosh (2013), lifetime
upper bounds are analyzed by considering the duty cycle and net-
work coding in WSN. Mothku and Rout (2017) have analyzed the
energy consumptions for a tree-based sensor network while con-
sidering the real-time network traffic. In Lee and Cheng (2012),
Lee et al. have estimated the expected residual energy for selection
of cluster head in a WSN. However, in our work, we have estimated
the expected residual energy of a node for future time period by
considering the energy consumption model and a predicted har-
vesting model. Further, an energy consumption model has been
designed for a proposed duty cycle based sensor actor network.
Tan and Tang (2017) have proposed a markov model to trace
the energy harvesting process and discussed the performance
aspects of a rechargeable node. A joint wireless energy replenish-
ment and mobile data gathering framework has been proposed
by Guo et al. (2014) for a rechargeable sensor network. Rout
et al. (2016) have proposed a markov decision process based
switching algorithm to achieve sustainable data collection in a
rechargeable sensor network. Shu et al. (2017) have proposed a
joint energy replenishment and scheduling mechanism to maxi-
mize the lifetime for rechargeable sensor networks. Imon et al.
(2015) have proposed a randomized switching algorithm for load
balancing in a tree-based WSN.
In Kang et al. (2017), a distributed delay efficient data aggrega-
tion scheduling mechanism has been proposed for duty cycled
WSN. Yoo et al. (2012) have proposed dynamic duty cycle schedul-
ing schemes to reduce sleep latency and achieve the balanced
energy consumptions for an energy harvesting WSN. In
Neamatollahi et al. (2017), a fuzzy-based clustering task schedul-
ing mechanism has been proposed to mitigate the reclustering
overhead in a WSN. In Arjunan and Sujatha (2017), a fuzzy-based
mechanism is proposed for cluster head selection and inter-
cluster routing in WSN. In Lee and Cheng (2012), a fuzzy based
clustering algorithm has been proposed using energy prediction
for a cluster-based sensor network. However, in our work, we have
proposed an adaptive duty cycling mechanism using fuzzy logic
system to achieve sustainability in an energy harvesting sensor
actor network.
2.1. Motivation and problem formulation
In an energy harvesting sensor networks, a node harvests
energy from ambient sources, such as solar light, thermal, wind
and vibration. The amount of harvested energy may vary with
weather conditions or seasonal changes. Therefore, the volume of
harvested energy may be less than the energy consumption of a
node in a time slot. This results in temporary disconnection of
the nodes. A node may join into the network in the next (solar)
energy available period. The temporary disconnected nodes affect
the quality of service of the network. To address the temporary dis-
connection of the nodes, we have proposed a fuzzy based adaptive
duty cycle mechanism for sustainability of the nodes till next
recharge period. Fuzzy logic system (FLS) is used to estimate the
expected duty cycle of the node for future time slot. Current resid-
ual energy, predicted harvesting energy (for a future time slot) and
expected residual energy (for a future time slot) are considered as
input parameters and expected duty cycle is considered as an out-
put parameter for fuzzy logic system. FLS generates an optimal
output value (expected duty cycle) by integrating the input vari-
ables and fuzzy rules set (Lee and Cheng, 2012; Collotta et al.,
2017; Yousaf et al., 2017).
In a tree-based sensor network, load balancing is one of the
techniques to improve the lifetime of the network. A randomized
switching algorithm has been proposed in Imon et al. (2015) to bal-
ance the load (in terms of data forwarding) in a sensor tree, so that
all nodes have uniform load. In a network, load balancing is
achieved by switching operations (switch the children from its par-
ent node to other potential parent nodes) (Imon et al., 2015). How-
ever, rechargeable sensor nodes are not considered in Imon et al.
(2015). In a rechargeable tree-based sensor network, every node
has different energy levels due to the unequal amount of harvested
energy and energy consumption. Therefore, load balancing tech-
nique may not achieve sustainability. Duty cycle is a suitable
mechanism for a rechargeable sensor network to achieve the sus-
tainability (by conserving the energy). In our work, duty cycle is
estimated using a fuzzy logic system. Based on the expected duty
cycle value, a node forwards the active and sleep schedule to the
children nodes or switches its children to potential parent nodes
for reliable data delivery.
 S.K. Mothku, R.R. Rout / Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497 1491

3. Adaptive duty cycling mechanism using fuzzy logic Table 1

Summary of Notations.

In this section, the proposed network model is introduced. An Symbol Description

energy consumption model and a harvesting model are presented n Number of sensor nodes in the network
to estimate the residual energy of a node for future time slots. Fur- T Sensor tree
ther, a fuzzy based adaptive duty cycling mechanism has been pro- Ei ðt Þ Residual energy of a node i at time t
Exi ðtÞ Expected residual energy of a node i at time t
posed for energy harvesting sensor networks.
H i ðt Þ Harvesting energy of a node i at time t
Eci ðt Þ Energy consumption of a node i at time t
Ci Number of children of node v i
3.1. Network model Br Radius of bottleneck zone
k Maximum number of hops in bottleneck zone
h Height of sensor tree
A tree-based wireless sensor and actor network is considered

n Maximum number of children in sensor tree
(shown in Fig. 1), where the root node is a local actor which col-
lects sensing data from a sensor tree. Let, T i ¼ ðV; EÞ be a sensor
tree, where V ¼ fv 0 ; v 1 ; . . . ; v n g represents a set of n sensor nodes
children (C i is the total number of children of a node v i ) and trans-
and v 0 is a local actor (i.e. Fig. 1 shows the sub-tree from a large
mitting the total data to its parent node and is given by
scale network tree) and E represents a set of communication links


between the sensors. Here, coordinated duty cycle enabled sensor ^n^
Eci ¼ a3 ltb þ a12 C i ltb þ ðC i ltb þ ltbÞ a11 þ a2 d ð1Þ
nodes are considered and a node harvests energy from the sunlight
(Piorno et al., 2009). Further, a node switches to a potential parent In our work, we have adopted an energy prediction model
(a node which has more duty cycle). A data collection round is called Weather-Conditioned Moving Average Model (Piorno et al.,
defined as the duration that an actor collects the data from a sensor 2009) to predict the harvested energy (using solar sensors) for a
tree (Imon et al., 2015). The notations used in this paper are men- future time slot. In our work, a day is divided into r time slots
tioned in Table 1. ðt 0 ; t 1 ; . . . ; tr Þ. According to the model (Piorno et al., 2009), the pre-
dicted harvesting energy is computed using the harvesting energy
value at previous time slot of the same day and the mean value of
3.2. Energy consumption and harvested model the past days of same time slot and is given by

In this paper, energy consumptions are estimated by adopting Hðd; t þ 1Þ ¼ a:Hðd; t Þ þ GAPk :ð1  aÞ:M D ðd; t þ 1Þ ð2Þ
energy parameters as proposed in Rout and Ghosh (2013), Rout where Hðd; t þ 1Þ represents the predicted harvesting energy for a
et al. (2016). The parameters of energy consumptions for sensing, day d at time slot t; a is a weighting factor, MD ðd; t þ 1Þ is the mean
^ are Es ; Er and Et ,
receiving and transmitting data (over distance d) of the harvesting energies at ðt þ 1Þ time slots for the previous D
respectively and are given by Es ¼ a3 , Er ¼ a12 , Et ¼ a11 þ a2 d ^n^ , days and GAPk measures the solar conditions in the present day rel-
where n ^ : path loss exponent, a11 : energy consumption per bit by ative to the previous days (Piorno et al., 2009).
the transmitter electronics, a2 : energy dissipation in the transmit The energy consumption (Eq. (1)), residual energy ðEi Þ and the
op-amp, a12 : energy consumption per bit by the receiver electron- harvesting energy (Eq. (2)) are used in the estimation of expected
ics and a3 : energy consumption for sensing a bit (Rout et al., 2016). residual energy ðExi Þ of a node for the future time slot (refer line
Further, an event centric application has been considered in our 4, Algorithm 1).
work. Energy estimation is done by considering occurrence of
events. A sensor node generates l bits of sensing data per event. Algorithm 1 Fuzzy Based Adaptive Duty Cycling Algorithm
b is an average rate of events that occur per unit time (Rout Input: Tree-based sensor actor network (T i ). Output:
et al., 2016). Therefore, the energy consumption for sensing in time Sustainable sensor tree
slot t is given by a3 ltb. The energy consumption (Eci ) of a node ðv i Þ 1: for i ¼ 1 to n do
is the sum of the consumptions for sensing, receiving data from its 2: Eci ðt þ 1Þ ¼ a3 lðt þ 1Þb þ a12 C i lðt þ 1Þb þ ðC i lðt þ 1Þb


þ lðt þ 1ÞbÞ a þ a d ^n^
11 2

3: Predict the value of Hi ðt þ 1Þ using the Eq. (2).

4: Exi ðt þ 1Þ=Ei ðt Þ þ Hi ðt þ 1Þ  Eci ðt þ 1Þ
5: Ed½i = Fuzzy_Duty_Cycle(Ei ðtÞ; Hi ðt þ 1Þ; Exi ðt þ 1Þ)
0 t
Active and Sleep Schedule

6: if Ed½i < T h then

7: Switch_Children(i)
8: else
9: Forward the active and sleep schedule to the children
10: end if
11: end for

3.3. Energy consumption model

Switching to potential parent
Local actor In this section, energy consumption of bottleneck zone in a sen-
Sensor node sor tree is estimated by considering duty cycle enabled nodes. In a
sensor tree (shown in Fig. 5), nodes around the root node (called
Fig. 1. The illustration of duty cycle based switching process in a tree-based sensor bottleneck zone) consume more energy than other nodes in the
and actor network. tree (Wang and Zhang, 2009). Br is the radius of the bottleneck
1492 S.K. Mothku, R.R. Rout / Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497

zone. The symbol k is the maximum number of hops in the bottle-

neck zone and h is the height of the sensor tree. Height of a sensor Fuzzy rule set
Predicted harvested
tree is the longest path (number of hops) from the root node to a

Expected duty cycle

Residual energy
energy
leaf node in the tree. Each sensor node in the tree has maximum
of n children. Fuzzy inference
Fuzzifier Defuzzifier
system
The total energy consumption in the bottleneck zone is the
summation of consumptions due to (i) sensing operation of the
nodes in the bottleneck zone (Esb ), (ii) relaying its own sensed data Expected residual
(generated by bottleneck zone nodes) to the root node (Erb ), and energy
(iii) relay the data (which is generated by the outside of the bottle-
neck zone nodes) to the root node (Erob ). Therefore, the total energy Fig. 4. Fuzzy Logic System.
consumption of the bottleneck zone (Ebc ) is given by
Ebc ¼ Esb þ Erb þ Erob ð3Þ
The energy consumption of bottleneck zone nodes for sensing oper-
 kþ1 a3 ltb, where pn
ation (Esb ) is pn  kþ1 is the average number of bot-
tleneck zone nodes which are in the active state, probability p

Bottle
denotes the average proportion of time (during the time slot t) in
B
which a node will be in the active state. The energy consumption r

neck z
(Erb ) for relying sensed data (generated by bottleneck zone nodes)
is given by

one
Z Br Z 2p
Erb ¼ kðxÞpltbqt rdhdr ð4Þ
0 0 h

h - height of sensor tree

B r - bottleneck zone radius

Fig. 5. Bottleneck zone in a duty cycle enabled sensor tree.

dr
r
d

Fig. 2. Membership functions for input variables (Residual energy, Predicted Br

harvesting energy, Expected residual energy).

Root node of a sensor tree

B r -Bottleneck zone radius

Fig. 6. A bottleneck zone of radius Br at root node of sensor tree.

where kðxÞ is the energy consumption for transmitting a bit by a

bottleneck zone node which is x distance away from the root (actor)
node (Bhardwaj et al., 2001; Mothku and Rout, 2017; Rout and
Ghosh, 2014) and is given by

^
n x
kðxÞ P a1  a12
^  1 dm
n

where dm is the length of one hop (Bhardwaj et al., 2001) and

qffiffiffiffiffiffiffiffiffiffiffiffi
dm ¼ n^ a2 ðan^11Þ , a1 ¼ a11 þ a12 . Density of a sensor tree is qt ¼ n A ,
hþ1

Fig. 3. Membership function for output variable (Duty cycle).

 S.K. Mothku, R.R. Rout / Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497 1493

where A is the area of a sensor tree ðT i Þ. The term (pqt rdhdr) is the the node will be switched to potential parent (refer Algorithm 3).
number of active nodes in the differential area (as shown in Fig. 6) Otherwise, the node forwards its active and sleep schedule to the
(Dabirmoghaddam et al., 2010). children so that the children send their data at active time periods
The energy consumption (Erob ) for relaying sensed data (gener- (for reliable data delivery) (lines 6 to 10).
ated at out side of the bottleneck zone) is given by
Algorithm 2 Fuzzy_Duty_Cycle(Ei ðtÞ; Hi ðt þ 1Þ; Exi ðt þ 1Þ)
pðn  kþ1 Þltb
 hþ1 n
X
Erob ¼ Ei ð5Þ Output: Expected duty cycle Edi
i¼1 1: Empty the list lhv alue; membershiplev eli
2: Find membership values (lðEi ðt ÞÞ; lðHi ðt þ 1ÞÞ and
where Ei is the energy consumption of a sensor node (which is
th
lðExi ðt þ 1Þ) and linguistic levels using Triangular
inside the bottleneck zone) to relay i bit from outside of the bot- membership function.
tleneck zone to the root (actor) node and is given by 3: DR ¼{A rule set with all possible combinations of
^ Br
n determined linguistic levels (from line  2Þ}
Ei P a1 ð6Þ 4: for each rule in DR do
^  1 dm
n
5: if lðEi ðtÞÞ; lðHi ðt þ 1ÞÞ; lðExi ðt þ 1ÞÞ then fit the
Therefore, the total energy consumption in the bottleneck zone membership levels of this rule
with duty cycle (i.e. p) is given by 6: Add an entry to the list l with
Z Br Z 2p 7: value = MaxðlðEi ðtÞÞ; lðHi ðt þ 1ÞÞ; lðExi ðt þ 1ÞÞÞ
 kþ1 a3 ltb þ
Ebc ¼ pn kðxÞpltbqt rdhdr 8: membershiplev el = output membership level of this
0 0 rule
Xpðnhþ1 nkþ1 Þltb
þ Ei ð7Þ 9: end if
i¼1
10: end for
11: Edi = Defuzzify(l)
3.3.1. Duty cycle and its impact on sustainability 12: return Edi
Duty cycle mechanism is one of the energy conservation tech-
niques in wireless sensor networks. Most of the existing works
have not addressed the estimation of duty cycle value in the liter-
ature. In a harvesting sensor network, a node does not sustain till
next recharge time if it does not have sufficient residual energy. Algorithm 3 Switch_Children(i)
The node conserves some amount of energy to sustain till next
1: Enqueue(children(i))
rechargeable time. Estimating the appropriate amount of conserva-
2: while IsEmptyQueue==false do
tion energy (when a node sleeps) to achieve the sustainability and
3: j = Dequeue()
reliable data delivery is a challenging problem in energy harvesting
4: P p = Find_Potential_Parents(j)
sensor networks. In our work, duty cycle (for a future time slot) is
estimated by considering the current residual energy, harvesting 5: if Pp ¼¼ £ then
energy (for a future time slot) and expected residual energy (for 6: Enqueue(children(j))
a future time slot) to achieve the sustainability in the harvesting 7: end if
8: 8p  with
 2 Pp , Select a potential parent node p
sensor network. If the duty cycle value is very low, then the node
does not perform its operations for a long period of time (because maximum duty cycle value ðEd½p Þ.
node spends long period of time in sleep state). This leads to unre- 9:  is the new parent of node j
p
liable data delivery. Therefore, children (of those nodes) will be 10: Update the tree
switched to other potential parent nodes to achieve reliable data 11: end while
delivery. Further, active and sleep schedule (coordinated duty cycle
schedule) improves the reliable data delivery. The proposed fuzzy
based adaptive duty cycling mechanism has been presented in the In Algorithm 3, switching process is presented. Children of a
next section. node (which has the less expected duty cycle value than the
threshold value) will be inserted into a queue (called Enqueue) (line
3.4. Fuzzy based adaptive duty cycling algorithm 1). Further, the potential parents will be found for every removal
node from the queue (called Dequeue) (lines 3 to 4). Potential par-
In this section, fuzzy based adaptive duty cycle algorithm (Algo- ents of a node is a set of nodes which are in the communication
rithm 1) is proposed for an energy harvesting sensor actor network range of that node (other than the current parent). A node with
to achieve sustainability. At the end of time slot t, a node estimates highest expected duty cycle value will be chosen from the poten-
the energy consumption ðEcðt þ 1ÞÞ for next time slot ðt þ 1Þ using tial parents as a new parent. If a child node doesn’t has a potential
the Eq. (1) and predicts the value of harvested energy ðHi ðt þ 1ÞÞ for parent, then its subsequent children will be inserted into the queue
time slot ðt þ 1Þ using the Eq. (2) (lines 2 to 3). Further, the node (lines 5 to 10).
determines expected residual energy ðExi ðt þ 1ÞÞ for time slot
ðt þ 1Þ (line 4). A Fuzzy Logic System (FLS) determines the expected 3.5. Fuzzy logic system
duty cycle ðEdÞ adaptively by considering the residual energy ðEi Þ,
predicted harvested energy ðHi Þ and expected residual energy In this section, a fuzzy logic system (FLS) is presented and the
ðExi Þ (line 5) as input parameters. The explanation of the FLS pro- duty cycle is estimated for a future time slot. The process of fuzzy
cess (Algorithm 2) is discussed in the next section. Further, a node logic system has been presented in Algorithm 2. Fig. 4 shows the
takes the switching decision based on the determined expected fuzzy logic system with three input parameters, such as current
duty cycle value to sustain till the next rechargeable time. If the residual energy, predicted harvesting energy (for a future time slot)
expected duty cycle value of a node is less than a threshold value and predicted residual energy (for a future time slot) and (one out-
ðT h Þ (which varies application to application), then children of put parameter) expected duty cycle. FLS process has four steps: (i)
1494 S.K. Mothku, R.R. Rout / Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497

Fuzzifier, (ii) Fuzzy rule set, (iii) Fuzzy inference system and (iv) Table 3
Defuzzifier. The fuzzifier converts the crisp values of input param- Fuzzy input/output variables and their linguistic variables.

eters to the membership values and linguistic levels using the tri- Input/Output variable Linguistic values
angular membership functions (line 2, Algorithm 2). The Residual energy Low, Medium, High
membership functions of the input and output variables are shown Predicted harvesting energy Low, Medium, High
in Figs. 2 and 3, respectively. The linguistic variables of input vari- Expected residual energy Low, Medium, High
ables are considered as low, medium and high and the linguistic Duty cycle Very Low, Low, Medium,
High, Very High
variables of output variable (expected duty cycle) are considered
as very low, low, medium, high and very high (as shown in Table 3).
Fuzzy rule set is a collection of fuzzy decision making rules
which are generated from three input variables as shown in Table 2.
START
Fuzzy rules are generated based on the following principle. A node
with high current residual energy, high predicted harvesting
energy (for a future time slot), high expected residual energy (for
a future time slot) acquires highest expected duty cycle value. At the end of time slot (t), a node estimates the
Fuzzy inference system finds the fuzzy output by using the fuzzy future residual energy and predicted harvested energy for
time slot (t+1)
rule set and membership functions. In Algorithm 2, line 3 to line
9 explains the fuzzy inference process. Further, the defuzzifier con-
verts the fuzzy output to a crisp value (which indicates the
The node estimates duty cycle value for time slot (t+1) using fuzzy
expected duty cycle of a node) (line 11). In our work, center of area logic system by considering the current residual energy,
(COA) method (is also called center of gravity in the literature harvested energy and predicted residual energy

(Ross, 2010; Kim et al., 2000) is used in the defuzzification process

(Lee and Cheng, 2012).
The complete working operation of the proposed fuzzy based
adaptive duty cycling algorithm is given in the flowchart as shown Is duty cycle
No Yes
in Fig. 7. At the end of time slot t, a node estimates the future resid- more than a
threshold
ual energy and predicted harvested energy for time slot ðt þ 1Þ. value?
Further, the node determines the expected duty cycle (for time slot
ðt þ 1Þ) using fuzzy logic system. The parameters current residual Switch the children to other Forward active and sleep
energy, harvesting energy (for a future time slot) and expected potential parent schedule to the children nodes
residual energy (for a future time slot) are considered as fuzzy
inputs and duty cycle is the fuzzy output. The node switches its
children to another potential parent node if duty cycle value is less
than the threshold value, otherwise, the node forwards the active
and sleep schedule to its children for reliable data collection.
STOP

Fig. 7. Overview of the proposed mechanism operation.

Table 2
Fuzzy decision making rules.

Inputs Output 3.6. Analysis of fuzzy based adaptive duty cycling algorithm
Residual Predicted harvesting Expected residual Duty
energy energy energy cycle In this section, the proposed mechanism has been analyzed in
terms of time complexity and overhead. In Algorithm 3, line 1 takes
Low Low Low Very Low

Low Low Medium Low O C b max time where C b max is the maximum number of children of a
Low Low High Medium

Low Medium Low Low node in the tree. The loop on line 2 runs for O Cb max time. Line 3
Low Medium Medium Medium
Low Medium High High takes Oð1Þ time. Find_Potential_Parents() function on line 4 takes


Low High Low Low O P b p max time where P
b p max is maximum number of neighbors of
Low High Medium Medium

Low High High Very High a node in the tree. Line 5 runs in Oð1Þ time. Line 6 takes O Cb max
Medium Low Low Very Low
Medium Low Medium Low time. Lines 8 and 9 run in Oð1Þ time. Line 10 takes OðnÞ time to
Medium Low High Medium update the tree (n is the total number of nodes in the tree). Since
Medium Medium Low Low b max 6 P
b p max < n, the complexity of Switch_Children() function
C
Medium Medium Medium Medium

Medium Medium High High (Algorithm 3) is obtained as O n C b max . In Algorithm 1, the loop
Medium High Low Low

Medium High Medium Medium on line 1 runs for all nodes in the network. Line 2 takes O C b max
Medium High High High
High Low Low Very Low time. Lines 3 and 4 runs in Oð1Þ time. In line 5, Fuzzy_Duty_Cycle
High Low Medium Low () function (Algorithm 2) runs in Oð1Þ time. The Switch function


High Low High Medium
(Algorithm 3) on line 7 takes O n Cb max time. Therefore, The total
High Medium Low Low
High Medium Medium Medium time complexity of the proposed mechanism is obtained as
High Medium High Medium

b max .
O n2 C
High High Low Low
High High Medium Medium Every node in the tree forwards active and sleep schedule only to
High High High Very High
its children in each time slot (every round) (line 9 in Algorithm 1).
 S.K. Mothku, R.R. Rout / Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497 1495

Table 4
Simulation Parameters.

Parameter Value
Number of sensor nodes 100–500 nodes
Transmission range 30 m
Initial energy of a node 2500 kJ
Es ¼ a 3 a3 ¼ 50  109 J=bit
Er ¼ a12 a12 ¼ 0:787  106 J=bit
^n^
Et ¼ a11 þ a2 d a11 ¼ 0:937  106 J=bit,
a2 ¼ 10  1012 J=bit=m2
^ ¼ 85 m
d
Path lose exponent (n^) 2
Data generated per event
by a node (l) 960 bits
Average rate of events
occur per unit time (b) 100

A data packet traverses multiple hops to reach the actor node

whereas duty cycle scheduling packets are forwarded only to the
children nodes (i.e. with one hop). Therefore, in comparison to
energy consumption for data packet transmission, the energy con-
sumptions for control packet transmissions are minimal for large-
scale sensor networks. Assume that, in a network, maximum 1000
number of control packets are generated for 100 number of nodes
in 100 data collection rounds. The size of the control packet is
considered as 10 bytes (Ye et al., 2004) which is significantly
smaller than the size of a data packet (refer Table 4). The energy
consumption for transmitting the control packets is 1000 
10  8  Et ¼ 0:080 J (refer Table 4). Moreover, the maximum
number of n control packets (duty cycle scheduling packets) are
transmitted in the tree for each time slot where n is the number
of nodes in the tree.
4. Performance evaluation
In this section, the performance of the proposed mechanism is
compared with a randomized switching algorithm (Imon et al.,
2015) by considering network sustainability metrics, such as num-
ber of rounds network is connected, first node disconnected round,
number of packets received at actor node and maximum number of
dead nodes.
4.1. Simulation environment
Fig. 8. Total number of rounds for which network is connected.
the randomized switching algorithm (Imon et al., 2015). It can be
observed from Fig. 8 that, using the proposed approach the net-
work remains connected for (on an average) 60:44% of time dura-
tion (in rounds) in comparison to the randomized switching
algorithm. The proposed mechanism estimates the expected duty
cycle value of a node for future time slots. According to the duty
cycle value, the node will perform the switching operation or for-
wards the active and sleep schedule to the children for sustainability
(i.e. till next rechargeable time slot). On the other hand, the ran-
domized switching algorithm performs the switching operation
to balance the current load factor. The existing approach does
not consider the sustainability condition. It can be seen from
Fig. 8 that, as number of nodes increases, the network connectivity
rounds decreases because the nodes near to the actor are over-
loaded with more incoming data traffic. Therefore, the nodes near
to the root node deplete their energy quickly.
Fig. 9 shows that the first node die round (first round in which
first node dies) and it occurs first with the randomized algorithm
(Imon et al., 2015) in comparison to the proposed mechanism.
The proposed mechanism takes switching decision based on the
expected residual energy. This gives the survivability to a node till

The proposed fuzzy based adaptive duty cycle algorithm is sim-

ulated using the Network Simulator-3 (Network Simulator 3, 2018-
04-2). A sensor tree is constructed using the breadth-first-search
algorithm. In a data collection round, a node relays the data to a
parent node. Assume that, every node generates 1 unit of data
packet (i.e. 960 bits each) in a round (each time slot) (Imon et al.,
2015). The initial energy of a node is 2500 J and a node harvests
maximum of 200 J per round. The following simulation parameters
are considered: a3 ¼ 50  109 J=bit; a12 ¼ 0:787  106 J=bit;
a11 ¼ 0:937  10 6
J=bit; a2 ¼ 10  1012 J=bit=m2 ; d ¼ 85 m and
^ = 2 (Rout et al., 2016; Rout and Ghosh, 2014). The number of sen-
n
sor nodes are varied from 100 to 500 and the simulation results are
presented for 100 rounds. Table 4 shows the parameters which are
considered in the simulation.

4.2. Simulation results and analysis

Fig. 8 shows the comparison of the number of rounds (a param-

eter for time duration) in which all nodes are connected between
the proposed fuzzy based adaptive duty cycle mechanism and Fig. 9. First node disconnected round.
1496 S.K. Mothku, R.R. Rout / Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497

next rechargeable time. It can be seen from Fig. 9 that, a node sur-
vives on an average of 47.13% rounds with the proposed mecha-
nism. Further, Fig. 10 shows the maximum number of dead
nodes for 100 rounds. The proposed mechanism reduces the max-
imum number of dead nodes significantly as compared with the
randomized switching algorithm (Imon et al., 2015).
Figs. 11 and 12 show the average number of packets received at
the root (local actor) node and number of lost packets in a sensor
tree, respectively. In a duty cycle based network, packets may be
dropped if the receiver node is in inactive state. In the proposed
mechanism, a node forwards the active and sleep schedule to the
children so that children relay the data at active time periods. This
results in reduction of dropping packets (refer Fig. 12) with the
proposed mechanism as compared to the randomized algorithm
(Imon et al., 2015). Table 5 shows that the total (average) number
of control packets (active and sleep schedule packets) generated in a
sensor tree per data collection round. It can be seen from Table 5
that as number of nodes increases the traffic generated by the con-
trol packets increases. It has been observed from Figs. 8–10 that
Fig. 12. Total number of lost packets for 100 data collection rounds.
the sustainability of the network is improved and this provides
better network stability. Further, successful reception of packets
is also improved as shown in Fig. 11.
Table 5
Total (average) number of control packets generated per data collection round.

Number of nodes Number of control packets

100 99
150 148
200 195
250 239
300 278
350 311
400 339
450 359
500 374

5. Conclusion

In this work, a fuzzy based adaptive duty cycle algorithm has

been proposed to achieve the sustainability for an energy harvest-
ing sensor network. Duty cycle of a sensor node is adaptively cho-
sen by the fuzzy logic controller. Fuzzy logic system determines
the expected duty cycle value by considering the residual energy,
predicted harvesting energy and expected residual energy. Further,
the proposed mechanism takes a switching decision for survivabil-
Fig. 10. Maximum number of dead nodes.
ity of the network. Coordinated duty cycle schedule has been con-
sidered to improve the reliable data delivery. From the simulation
results, it has been observed that there is an improvement in net-
work connected rounds on an average of 60.44 % with the proposed
mechanism. Survivability of a node is increased on an average of
47.13% with the proposed mechanism. Further, there is a reduction
in the maximum number of dead nodes and a significant improve-
ment in the total number of received packets at actor node with
the proposed mechanism in comparison to a randomized switch-
ing algorithm. An interesting topic for future work is to apply the
inventory models to achieve the sustainability for a rechargeable
sensor network.

References

Akyildiz, I.F., Kasimoglu, I.H., 2004. Wireless sensor and actor networks: research
challenges. Ad Hoc Netw. 2 (4), 351–367.
Aoudia, F.A., Gautier, M., Berder, O., 2017. Distributed computation of fair packet
rates in energy harvestingwireless sensor networks. IEEE Wireless Commun.
Lett. 6 (5), 626–629.
Arjunan, S., Sujatha, P., 2017. Lifetime maximization of wireless sensor network
using fuzzy based unequal clustering and ACO based routing hybrid protocol.
Appl. Intell., 1–18 https://doi.org/10.1007/s10489-017-1077-y.
Bhardwaj, M., Garnett, T., Chandrakasan, A.P., 2001. Upper bounds on the lifetime of
Fig. 11. Total number of packets received for 100 data collection rounds. wireless sensor networks. IEEE Int. Conf. Commun. (ICC), 785–790.
 S.K. Mothku, R.R. Rout / Journal of King Saud University – Computer and Information Sciences 34 (2022) 1489–1497
Collotta, M., Pau, G., Maniscalco, V., 2017. A fuzzy logic approach by using particle
swarm optimization for effective energy management in IWSNs. IEEE Trans. Ind.
Electron. 64 (12), 9496–9506.
Dabirmoghaddam, A, Ghaderi, M., Williamson C., Energy-Efficient Clustering in
Wireless Sensor Networks with Spatially Correlated Data, IEEE INFOCOM 2010.
Elshrkawey, M., Elsherif, S.M., Wahed, M.E., 2018. An enhancement approach for
reducing the energy consumption in wireless sensor networks. J. King Saud
Univ. Comput. Inf. Sci. 30 (2), 259–267.
fan Hsin, C., Liu, M., 2006. Randomly duty-cycled wireless sensor networks:
dynamics of coverage. IEEE Trans. Wireless Commun. 5 (11), 3182–3192.
Guo, S., Wang, C., Yang, Y., 2014. Joint mobile data gathering and energy
provisioning in wireless rechargeable sensor networks. IEEE Trans. Mob.
Comput. 13 (12), 2836–2852.
Imon, S.K.A., Khan, A., Francesco, M.D., Das, S.K., 2015. Energy-efficient randomized
switching for maximizing lifetime in tree-based wireless sensor networks. IEEE/
ACM Trans. Networking 23 (5), 1401–1415.
Kang, B., Nguyen, P.K.H., Zalyubovskiy, V., Choo, H., 2017. A distributed delay-
efficient data aggregation scheduling for duty-cycled WSNs. IEEE Sens. J. 17
(11), 3422–3437.
Kim, Y.H., Ahn, S.C., Kwon, W.H., 2000. Computational complexity of general fuzzy
logic control and its simplification for a loop controller. Fuzzy Sets Syst. 111 (2),
215–224.
Lee, J.-S., Cheng, W.-L., 2012. Fuzzy-logic-based clustering approach for wireless
sensor networks using energy predication. IEEE Sens. J. 12 (9), 2891–2897.
Lee, S., Lee, H.S., 2010. Analysis of network lifetime in clustere-based sensor
networks. IEEE Commun. Lett. 14 (10), 900–902.
Mothku, S.K., Rout, R.R., 2017. Adaptive buffering using markov decision process in
tree-based wireless sensor and actor networks. Comput. Electr. Eng.
Neamatollahi, P., Naghibzadeh, M., Abrishami, S., 2017. Fuzzy-based clustering-task
scheduling for lifetime enhancement in wireless sensor networks. IEEE Sens. J.
17 (20), 6837–6844.
Network Simulator 3. https://www.nsnam.org, Accessed: 2018-04-25.
Piorno, J.R., Bergonzini, C., Atienza, D., Rosing, T.S., 2009. Prediction and
management in energy harvested wireless sensor nodes. Wireless VITAE’09,
6–10.
1497
Ross, T.J., 2010. Fuzzy Logic with Engineering Applications. Wiley.
Rout, R.R., Ghosh, S.K., 2013. Enhancement of lifetime using duty cycle and network
coding in wireless sensor networks. IEEE Trans. Wireless Commun. 12 (2), 656–
667.
Rout, R.R., Ghosh, S.K., 2014. Adaptive data aggregation and energy efficiency using
network coding in a clustered wireless sensor network: an analytical approach.
Comput. Commun. 40, 65–75.
Rout, R.R., Krishna, M.S., Gupta, S., 2016. Markov decision process-based switching
algorithm for sustainable rechargeable wireless sensor networks. IEEE Sens. J.
16 (8), 2788–2797.
Shu, Y., Shin, K.G., Chen, J., Sun, Y., 2017. Joint energy replenishment and operation
scheduling in wireless rechargeable sensor networks. IEEE Trans. Ind. Inf. 13 (1),
125–134.
Tan, L., Tang, S., 2017. Energy harvesting wireless sensor node with temporal death:
novel models and analyses. IEEE/ACM Trans. Networking 25 (2), 896–909.
Wang, Q., Zhang, T., 2009. Bottleneck zone analysis in energy-constrained wireless
sensor networks. IEEE Commun. Lett. 13 (6), 423–425.
Wang, C., Li, J., Ye, F., Yang, Y., 2016. A mobile data gathering framework for wireless
rechargeable sensor networks with vehicle movement costs and capacity
constraints. IEEE Trans. Comput. 65 (8), 2411–2427.
Wu, Y., Liu, W., 2013. Routing protocol based on genetic algorithm for energy
harvesting-wireless sensor networks. IET Wireless Sensor Syst. 3 (2), 112–118.
Ye, X., Liang, W., 2017. Charging utility maximization in wireless rechargeable
sensor networks. Wireless Netw. 23 (7), 2069–2081.
Ye, W., Heidemann, J., Estrinsss, D., 2004. Medium access control with coordinated
adaptive sleeping for wireless sensor networks. IEEE/ACM Trans. Networking 12
(3), 493–506.
Yoo, H., Shim, M., Kim, D., 2012. Dynamic duty-cycle scheduling schemes for
energy-harvesting wireless sensor networks. IEEE Commun. Lett. 16 (2), 202–
204.
Yousaf, R., Ahmad, R., Ahmed, W., Haseeb, A., 2017. Fuzzy power allocation for
opportunistic relay in energy harvesting wireless sensor networks. IEEE Access
5, 17165–17176.