future internet

Article
Article
SCMC: An Efficient Scheme for Minimizing Energy in
SCMC: An Efficient Scheme for Minimizing Energy
WSNs
in Using
WSNsaUsing
Set Cover
a SetApproach
Cover Approach
Ahmed Redha Mahlous
Ahmed Redha Mahlous
Computer Science Department, Prince Sultan University, Riyadh 11586, Saudi Arabia,
Computer Science Department, Prince Sultan University, Riyadh 11586, Saudi Arabia, [email protected];
[email protected]; Tel.: +966-50-845-0765
Tel.: +966-50-845-0765
Received: 1 November 2017; Accepted: 8 December 2017; Published: 13 December 2017
Received: 1 November 2017; Accepted: 8 December 2017; Published: 13 December 2017

Abstract: Energy-efficient clustering and routing are well known optimization problems in the
Abstract: Energy-efficient clustering and routing are well known optimization problems in the study
study of Wireless Sensor Network (WSN) lifetime extension. In this paper, we propose an intelligent
of Wireless Sensor Network (WSN) lifetime extension. In this paper, we propose an intelligent hybrid
hybrid optimization algorithm based on a Set Cover approach to create clusters, and min-cost max-
optimization algorithm based on a Set Cover approach to create clusters, and min-cost max-flow
flow for routing (SCMC) to increase the lifetime of WSNs. In our method we used linear
for routing (SCMC) to increase the lifetime of WSNs. In our method we used linear programming
programming (LP) to model the WSN optimization problem. This model considers minimizing the
(LP) to model the WSN optimization problem. This model considers minimizing the energy for all
energy for all nodes in each set cover (cluster), and then minimizing the routing energy between the
nodes in each set cover (cluster), and then minimizing the routing energy between the nodes and the
nodes and the base station through intermediate nodes, namely cluster heads. To evaluate the
base station through intermediate nodes, namely cluster heads. To evaluate the performance of our
performance of our scheme, extensive simulations were conducted with different scenarios. The
scheme, extensive simulations were conducted with different scenarios. The results show that the set
results show that the set cover approach combined with the min-cost max-flow algorithm reduces
cover approach combined with the min-cost max-flow algorithm reduces energy consumption and
energy consumption and increases the network’s lifetime and throughput.
increases the network’s lifetime and throughput.

Keywords: set cover; min cost flow; cluster; optimization; wireless sensor network; routing
Keywords: set cover; min cost flow; cluster; optimization; wireless sensor network; routing

1. Introduction
1. Introduction
The termThe“Wireless Sensor Network”
term “Wireless (WSN) refers
Sensor Network” (WSN) to refers
a largetonumber of spatially
a large number distributed
of spatially distributed
autonomous nodes that organize themselves into a multi-hop wireless network for the monitoring
autonomous nodes that organize themselves into a multi-hop wireless network for the monitoring and
and recording of of
recording physical
physicaloror environmental conditions.Their
environmental conditions. Theirapplications
applications include
include battlefield
battlefield surveillance,
surveillance, target tracking, security, environmental control, habitat monitoring, source localization,
target tracking, security, environmental control, habitat monitoring, source localization, fire detection,
fire detection, oilgas
oil and andpumping,
gas pumping, and many
and many more.more.
A WSN system
A WSNconsists
system of distributed
consists nodes and
of distributed a gateway
nodes that provides
and a gateway wireless connectivity
that provides wireless connectivity
back to the
back to the wired world (see Figure 1). They can be organized into threeof
wired world (see Figure 1). They can be organized into three types network
types topology
of network topology as
as shownshown
in Figure 2.
in Figure 2.

Figure 1. Wireless
Figure 1.Sensor Network
Wireless Sensor (WSN) Components,
Network Gateway, Gateway,
(WSN) Components, and Distributed Nodes. Nodes.
and Distributed

Future Internet 2017, 9, 95; doi:10.3390/fi9040095 www.mdpi.com/journal/futureinternet

Future Internet 2017, 9, 95; doi:10.3390/fi9040095 www.mdpi.com/journal/futureinternet
 Future Internet 2017, 9, 95 2 of 18
Future Internet 2017, 9, 95 2 of 18

Figure 2. WSN Topologies.

Figure 2. WSN Topologies.

A star network topology consists of a single base station and multiple remote nodes. The nodes
A star network
cannot communicate directly topology
with each consists
other,of a single
while base station
the base station canandcommunicate
multiple remote with nodes.
all of The
the nodes
cannot communicate directly with each other, while the base
nodes. In a mesh network topology, a node can directly communicate with any other node within its station can communicate with all of the
nodes.range.
transmission In a mesh
This network
solves the topology,
single point a node can directly
of failure problem communicate
of star networkswith any other node
by allowing whatwithin its
is called multi-hop communication. In a cluster network topology, nodes are grouped to form a what
transmission range. This solves the single point of failure problem of star networks by allowing
cluster is called
with onemulti-hop
of the nodes communication. In a cluster
elected as a “cluster head” network topology,
(CH). Cluster nodes arewhen
topologies grouped to formina cluster
employed
with one of the nodes elected as a “cluster head” (CH). Cluster
WSNs generally result in lower energy consumption and an increased overall lifetime of the network. topologies when employed in WSNs
generally result in lower energy consumption and an increased
Because they are usually deployed in harsh environmental scenarios, WSN nodes rely solely on overall lifetime of the network.
their batteries Because
most of theytheare usually
time. deployed
As batteries havein harsh
a finite environmental
capacity, minimizingscenarios, WSN consumption
energy nodes rely solely on
their batteries most of the time. As batteries have a finite
in order to extend the lifetime of sensor nodes without compromising their functionality capacity, minimizing energy consumption
is an in
order to extend the lifetime
important area of study in WSN research [1–5]. of sensor nodes without compromising their functionality is an important
In area
WSNs,of study in WSN
all nodes research
usually share[1–5].
common sensing tasks, each covering a different target area.
In WSNs, all nodes usually
This implies that not all sensors that are deployed share common withinsensing
the same tasks,
target each
areacovering
are required a different
to perform target area.
This
their task implies that throughout
continuously not all sensors thethat are deployed
system’s within the
entire lifetime. same
Sensor targetconsume
nodes area are more
required to perform
energy
when they are in transmission mode (active) than when they are in sleeping mode (inactive). If all energy
their task continuously throughout the system’s entire lifetime. Sensor nodes consume more
when they
sensor nodes within area coverage
in transmission
target areamode are(active)
active, than
the datawhen they are
collected inbe
will sleeping
redundant mode and (inactive).
a large If all
amountsensor
of energynodes within
will a coverage
be consumed target
and area To
wasted. are save
active,thethe data collected
overall energy ofwill WSNs,be redundant
our approach and a large
amount of energy will be consumed and wasted. To save the overall
was to activate some sensor nodes while turning off others within each target area (set cover/cluster) energy of WSNs, our approach
was to activate some sensor
without losing or disrupting system connectivity. nodes while turning off others within each target area (set cover/cluster)
without losing
Specifically, or disrupting
the proposed model system
aims to connectivity.
determine a minimum set cover within a WSN, and to
Specifically, the proposed
elect a relay node (cluster head) that will forward model aimssensor
to determine a minimum
node messages to theset
base cover within
station. a WSN, and to
Moreover,
the modelelectaddresses
a relay node the (cluster
problemhead) that will
of finding theforward
minimum sensor
cost node
routesmessages
from sensor to the base to
nodes station.
the base Moreover,
station through the elected cluster heads using a min-cost flow algorithm. To achieve the best solutionthe base
the model addresses the problem of finding the minimum cost routes from sensor nodes to
station through
for expanding a WSN’sthe electedwe
lifetime, cluster headsa using
propose modified a min-cost
versionflowof thealgorithm.
min-costToflow achieve the bestbysolution
algorithm
for expanding
taking residual energyaas WSN’s lifetime,
a constraint we propose
value to judge adifferent
modified version
paths and of the min-cost
choose the best flow
one amongalgorithm by
taking residual energy as a constraint
them, i.e., the one with the highest residual energy. value to judge different paths and choose the best one among
them, i.e., the one with the highest residual energy.
While most of the approaches found in literature employ a Particle Swarm Optimization (PSO)
algorithm toWhile most of the
find clusters approaches
[6–10], in this paperfoundwe in literature
present aemploy a Particlebased
novel approach Swarm onOptimization
a set cover (PSO)
algorithm to find clusters [6–10], in this paper we present
algorithm for finding clusters and a min-cost max-flow algorithm to find paths with minimum a novel approach based on cost
a set cover
(energyalgorithm
consumption) for finding clusters and
and maximum a min-cost
residual energy. max-flow
There are algorithm
two main to find paths with
motivations for minimum
finding cost
(energy consumption) and maximum residual energy. There
minimum set covers and using min-cost max-flow routing. The first is the need for an efficient energy are two main motivations for finding
scheme in WSNs, where sensor nodes tend to use small batteries for energy supply, and which are in energy
minimum set covers and using min-cost max-flow routing. The first is the need for an efficient
scheme
many cases in WSNs, where Therefore
non-replenishable. sensor nodes tend to use
an efficient small batteries
activation management for energy
scheme supply,
is neededand which
to are
in many cases non-replenishable. Therefore an efficient activation
conserve their energy, and thus that of the whole network. The second is the need for reliability in management scheme is needed to
conserve their energy, and thus that of the whole network.
wireless networks. This need stems from the unpredictable nature of the wireless environment, whichThe second is the need for reliability in
unlike wireless
its wirednetworks.
counterpart Thisisneed
more stems
prone from tothe
link unpredictable
failures duenature to dead of the wireless
sensor nodes.environment,
A typical which
unlike its wired counterpart is more prone to link failures due
scenario where our approach can be of particular interest is if we consider a monitoring applicationto dead sensor nodes. A typical scenario
targeted at the structural integrity of buildings, with WSN nodes deployed for data collection
Future Internet 2017, 9, 95 3 of 18

where our approach can be of particular interest is if we consider a monitoring application targeted
at the structural integrity of buildings, with WSN nodes deployed for data collection purposes over
key city constructions such as roads, buildings, and bridges. When one critical event is detected
(e.g., a dangerous flexure of a column), the monitoring application should trigger an alert to be
delivered immediately to the WSN’s data collection points (base station). In this case, it is particularly
important to increase resiliency against link failures and ensure a long lifetime, which is what the
SCMC technique presented in this paper achieves.
The remainder of the paper is organized as follows: Section 2 presents related work, Section 3
presents the approach, Section 4 describes simulation results, and Section 5 concludes the paper.

2. Related Works
Three key considerations in WSNs are coverage [11], connectivity [12], and battery lifetime [13].
Coverage refers to how well an area is monitored, while connectivity means that each node or device
should be connected such that the data is sent back to the base station (BS). Any applied heuristic
needs to fulfill the requirement of Q-coverage and P-connectivity, where Q is the minimum number
of sensors monitoring any one point in the area, and P is the number of disjoint paths between any
two sensors.
In this section, we shall take a brief tour of the past approaches at solving these issues, starting
with [14] in which the authors proposed a heuristic method called QC-MCSC to maximize the network
lifetime while satisfying both the coverage and connectivity requirements. This method consists of
three phases or layers where the output of one phase is the input of the next one. The first phase
is the Coverage phase, which outputs the set of sensors that cover all the target areas. Next is the
Connectivity and Redundancy Reduction phase, which takes the set of the first phase and applies the
Breadth-First Search (BFS) algorithm to find the shortest path for each sensor. In this phase, unnecessary
nodes are removed and any extra nodes are only added if the network is not fully connected, and the
output is a minimal connected set cover of nodes. Finally, the Energy and Priority Updating phase
determines the lifetime constant of each set cover from the second phase. The lifetime constant is the
minimum lifetime available from sensors in a set cover. In this phase, if the battery life ‘B’ of one node
is less than the energy needed for communicating and sensing ‘E’, the node is removed from the set
cover. Through simulation, the authors compared the Q-Coverage Maximum Connected Set Cover
(QC-MCSC) algorithm with the Three Phase Iteration Connected Set Cover (TPICSC) algorithm [15]
and the results showed that the former achieves a higher lifetime. Furthermore, the results of the
simulations reveal that the proposed heuristic (QC-MCSC) is very close to the actual optimal result.
A difficulty that researchers face when they want to expand the coverage time span is the question
of how to distribute the sensors into separate independent subdivisions, known as Disjoint Set Covers
(DSC). This problem is identified as having an NP-hard complexity, and many researches have used
the Boolean sensing model to solve it.
To facilitate topic lookup for the reader, we shall broadly divide the remaining studies by the type
of approach and/or the specific WSN issue being addressed.

2.1. Target Coverage

A number of studies have focused on addressing the target coverage problem. The objective of
the authors in [16] for example, was to have the maximum number of targets monitored by sensor
nodes before they run out of energy. To extend the network lifetime, they proposed an approach to
calculate the maximum number of disjoint sensor covers with a heuristic greedy algorithm using three
parameters: (1) the set covers; (2) the number of targets; (3) the number of sensors. The proposed
approach finds the set covers that can be used to maximize the network lifetime of a WSN in polynomial
time. They showed that it gives a local maximum solution due to its heuristic search.
Maintaining full connectivity of the WSN was the primary focus of [17], in which the authors
proposed an algorithm for ensuring that both conditions (i.e., coverage and connectivity) are met for
Future Internet 2017, 9, 95 4 of 18

conserving energy and extending the lifetime of the WSN. It relies on a distributed response procedure
to locate the routes between the sensors and the sinks. The procedure allows for several routes leading
from a sink to individual sensors to be discovered. Depending on the set of routes, a Minimum
Distance Path (MDP) greedy algorithm is derived from resolving the Maximum Set Cover (MSC)
issue with the goal of locating the highest possible number of set covers. The MDP-MSC algorithm
works with the concept of choosing the sensors possessing the shortest distance to the set cover under
construction, in order to include them into the contemporary set cover after each process. In this
algorithm, the discovered routes, which are determined by the maximum number of routes per cover,
influence the choice of sensors that are added to the set cover. The authors stated that the optimum
value of the maximum number of routes per cover, as well as the level of density of the sensors within
the target area, all impact the performance and control of transmission overhead in the WSN.
Based on simulation and experimental analysis using MDP-MSC, a number of set covers
received a noticeable growth by 13% and 21% compared to the Greedy-MSC [18] and Heuristic
Approach-Multidimensional Scaling (HA-MDS) algorithms [19]. MDP-MSC did not just outperform
Greedy-MSC and HA-MDS; it also had the lowest time complexity between them. MDP-MSC achieved
a balanced consumption of energy and less computational time.
On the other hand, the authors in [20] proposed a distributed lifetime coverage optimization
(DiLCO) protocol to preserve coverage as well as boost the WSN’s lifetime. The protocol begins by
applying a divide and conquer algorithm to split the target location into smaller regions known as
sub-locations, after which it is concurrently carried out in every single sub-location. Assumptions are
made that the partitioning of the sub-locations is carried out in a regular manner and that the sensors
are set uniformly in each sub-location. The periodic DiLCO protocol possesses four stages in each
period: information exchange, leader election, decision, and sensing. Every period possesses a single
set cover handling the sensing operation. This type of cyclic scheduling boosts the reliability of the
WSN against sensor failures. Sensors without sufficient power to successfully carry out a period or
those that fail before the decision phase is completed are not included in the scheduling procedure.
If a sensor fails after being chosen, the quality is only reduced for the duration of the period as another
functioning sensor is selected in the next period.
Two types of packets were used when carrying out the DiLCO protocol: info and Active/Sleep.
Info packets are transmitted from a sensor to all others within the same sub-location with the purpose of
providing information. Active/Sleep packets are transmitted by the leader to the other sensors within
the same sub-location to notify them to set their status to either Active or Sleep in the sensing stage.
The challenges faced in implementing the DiLCO protocol lie in selecting the most appropriate
leader in every single sub-location and the most suitable collection of active sensors to ascertain a high
standard in coverage.
In [21], the aim was to find the maximum number of sensor coverage sets. The authors used two
centralized algorithms and one localized algorithm. The two centralized algorithms deal with how
the sensors are added to make a connected cover. The localized algorithm is used for communication
between the local nodes. All three algorithms were only run once during the initialization phase. They
concluded that keeping one coverage set active in a round-robin fashion would extensively reduce the
energy consumption.

2.2. Node Activation

With regards to node activation, many existing studies propose alternating nodes between Sleep
and Active modes, or regulating the sensing range of the nodes. The authors in [22] however, proposed
to use both approaches together by setting up the minimum sensing radius for each sensor in the
Active state. This reduces the active node’s density, thereby reducing useless Media Access Control
(MAC) layer interference, which would otherwise consume lots of power.
Future Internet 2017, 9, 95 5 of 18

2.3. Node Distribution

The authors in [23] proposed to choose the smallest possible number of nodes that will cover the
critical targets by using a cost function called the Critical Control Factor. This function is mapped into
3 loops: (1) sensor availability check (adding the cover sets); (2) uncovered target check (populating
the added cover sets with sensors); and (3) sensor applicability check (sorting the nodes and choosing
the top-scoring node). The main benefits of this proposed solution are that it is flexible, has a low
complexity, and has a short execution time. In the end, the authors concluded that there are two
solutions for node deployment: static and dynamic. In the static method, the proposed algorithm
performed better than the other approaches in terms of execution time, and exhibited near-optimal
results in the generated coverage sets. The dynamic method on the other hand yielded better results
but with a longer execution time.

2.4. Redundancy Management

Redundancy plays an important role in maintaining the reliability of a WSN by handling
unforeseen failures and boosting the accuracy of calculated computations, due to the infeasibility
of replacing the failed sensors. Due to this infeasibility, prolonging the network lifetime to ensure
that measurements are taken for as long as possible is another desired trait. The Location-Unaware
Coverage (LUC) algorithm [24] addresses the issue of finding a balance between efficient redundancy
and network lifetime. The LUC algorithm determines whether a sensory node will be actively receiving
measurements or sleeping to preserve its energy. To carry this out, each sensor node determines its
neighbors and their relative distance in a two-hop neighborhood. LUC carries out a series of geometric
and heuristic tests to determine the ‘strength’ of the nodes, with the weaker ones put to sleep initially.
The stronger nodes possess a higher reliability, and hence are kept active for a longer duration. The LUC
algorithm has two protocols: Iterative LUC (LUC-I) and Probabilistic LUC (LUC-P), which determine
whether nodes will be set in Active or Sleep mode through a series of hops. The protocols bring about
low operating costs and may greatly minimize the number of nodes that are active.
Upon completion of the protocols, all nodes are set to either the Sleep or the Active mode. Active
nodes are responsible for covering the entire target area, and the WSN’s lifetime and reliability are
both improved.

2.5. Clustering
The authors in [25] proposed to use the Low-Energy Adaptive Clustering Hierarchy (LEACH)
protocol [26] with the Fuzzy Inference System (FIS) based on Mamdani’s method [27] to select
cluster heads and to extend the lifetime of WSNs. FIS consists of a fuzzifier, an inference engine,
and a defuzzifier. The proposed model uses three parameters: distance to coordinator, energy,
and density. IF-THEN rules are used to calculate the chance for a sensor node to become a cluster
head. The chance value is obtained by defuzzification, which aggregates the results of each rule.
Then, the node chooses the maximum of the chance values and compares it to a pre-defined threshold
value. If it is found to be less than the threshold then it becomes the cluster head. Experimental
studies conducted with LEACH and LEACH-Fuzzy Logic (LEACH-FL) showed that using LEACH-FL
increased the network’s lifetime twice more than when using LEACH.

2.6. Set Cover

A greedy set cover algorithm was proposed by the authors in [28]. It selects the minimum number
of sensors that are disseminated into disjoint and non-disjoint sets, with the necessity that every
set cover completely fulfills the scope of all its targets. They stated that the proposed algorithm is
an improvement over the classical greedy set cover algorithm, and its approximation ratio is verified
to be not worse than log(m). Through simulations, they demonstrated that their algorithm performed
better than others from previous research efforts.
Future Internet 2017, 9, 95 6 of 18

2.7. Optimization
Finally, many authors have considered the distribution of sensors and energy optimization as
a multi-objective optimization problem. The authors in [29] for instance proposed an Efficient Cover
Set Selection (ECSS) approach, which is an improved version of the Non-Dominated Sorting Genetic
Algorithm (NSGA-II) [30]. The goal of the proposed scheme is to maintain full coverage by using
fewer sensors. They highlighted that having more active sensors results in more coverage but with
a lower overall lifetime, and vice versa.
The authors concluded that the ECSS algorithm had advantages such as the ability to reset all
the sensors’ statuses with a one-time approach. It could also be made flexible by providing it with
parameters such as the requested field coverage and the model.
In another approach named the Multi-Objective Set Cover Problem (MO-SCP) [31], the authors
considered reliable coverage as an additional conflicting objective to be optimized. They tried to
solve it using a multi-objective evolutionary algorithm to select the set with the minimum number
of sensor nodes, such that it will provide coverage of all the targets. To meet the requirement of
staying active for long periods, the authors proposed to alternate nodes between idle and active states
through something called ‘wake-up scheduling’ of the sensors. Their objective was to achieve a WSN
system which can actively sense its predefined targets almost all the time. With the implementation of
this method, only a single disjoint set is active while the others are kept in sleep mode (in a specific
interval), resulting in an energy conservation scheme. When the active set goes out of power, another
set is scheduled to be in the active state. Therefore with this approach, the more DSCs that there are
the better it is for the entire system to stay on power. Moreover, they adopted a heuristic crossover
operator designed specifically to improve the performance of the algorithm, and conducted simulation
scenarios to test the effectiveness of the proposed algorithm.
In this paper, we propose an approach that combines a set cover method with a min-cost max-flow
algorithm using a linear programming technique.

3. Approach
The proposed algorithm passes through three main phases: (1) information gathering; (2) set
cover formation; and (3) path determination.
During the information-gathering phase, all of the sensor nodes including the base station are
assigned a unique ID. They then broadcast their IDs using the Carrier-Sense Multiple Access with
Collision Avoidance (CSMA/CA) MAC layer protocol, which allows the base station to discover sensor
node locations. Now, with the help of the received information about the network, the base station
starts the creation of set covers, electing cluster heads and determining optimal routing paths using
network layer information. It is to note that once the set cover creation and path determination phases
are over, all cluster heads are aware of the next hop to the base station and sensor nodes are also
informed about the cluster head ID that they belong to. Then, each cluster head provides a TDMA
schedule to its member sensor nodes for intra-cluster communication.
We modeled the energy conservation problem as an optimization problem where a “single
objective function” is used for the minimum set covers, and for the min-cost max-flow model too. Our
key contribution is that the proposed model guarantees a longer network lifetime.

3.1. Network Model and Assumptions

A wireless sensor network is composed of a set of battery-powered sensors and a sink. The latter is
connected to a power source and so we disregard its energy consumption in this paper. Communication
between the sensors is achieved through wireless links if they are within range of each other, or through
other intermediate sensors if not.
Future Internet 2017, 9, 95 7 of 18

Thus, a WSN can be modeled as a unit disk graph [32,33] G = (V, E) where V is the set of all nodes
including the sink and E ∈ V × V is a set of possible communication links. We assume our graph
is connected.
S = {s1 , s2 , ..., sN } represents a collection of N sensor nodes that are distributed randomly in
a two-dimensional plane to monitor a set of target nodes T = {t1 , t2 , ...., tM }.
Given two nodes s1 and s2 (one of them should be a sensor), they are connected by an edge
(s1 , s2 ) ∈ E, if the first sensor is within the sensing coverage range of the second sensor. Thus, two
nodes s1 and s2 can communicate with each other (there is an edge between them) if ||d(s1 , s2 )|| ≤ 1,
where ||d(s1 , s2 )|| is the Euclidean distance between s1 and s2 .
Each sensor node si ∈ S, can operate into a number of sensing ranges r1 , r2 , ..., rp where each rk
consumes energy ek , 1 ≤ k ≤ P. The initial energy of each sensor node is E. A sensing node si ∈ S can
cover a target ti ∈ T using sensing range rk if the Euclidean distance d(si , tj ) is less than or equal to rk ,
denoted as tj → si . Then, we define the target subset Ti = {tj | tj → si , 1 ≤ j ≤ m} as the target set of the
sensor si .
We assume that at regular time intervals, the information sensed by a network node has a fixed
unit size of b bits that can be stored at the sensor in a buffer of infinite capacity, modeled as a First In
First Out (FIFO) queue. We also assume that the channel bandwidth for all links is W bps and that all
transmitted packets have a unit size of b bits. Only one packet can be transmitted in each time slot t,
which is defined as t = b/W seconds. Time-division multiple access (TDMA) is used at the MAC layer.

3.2. Cluster Creation

Our main purpose is to extend the network lifetime as much as possible, and to minimize the
energy consumption by keeping the amount of traffic to a minimum and by scheduling nodes to
monitor a set of targets continuously in turn. To facilitate the achievement of our objective, a set of
clusters is constructed. Many studies have adopted the creation of clusters in WSNs [34–36], however
only some of them considered the use of a set cover algorithm. In this study, we consider the
construction of a minimum set cover C = {Ci}, 1 ≤ i ≤ K to create clusters. The problem can be
described formally as follows.

Set Cover Definition and Creation

The set cover problem is one of Karp’s 21 NP-complete problems shown to be NP-complete in
1972 [37].
The set cover optimization has many applications in different areas [38] including manufacturing,
service planning, location problems, network security, and multiple sequence alignments for computational
biochemistry [39–42].
Given a set of elements (called the universe) and a collection S of m sets whose union equals
the universe, the set cover problem is to identify the smallest sub-collection of S whose union equals
the universe.
In our paper we adopt the following approach: using a greedy set cover algorithm [19] to choose
which cover a sensor node will belong to. In this approach, we define a finite set S = {S1 , S2 , S3 , . . . , Sm },
corresponding to the areas to be monitored, a collection {Sj }nj = 1 of subsets of S, where each Sj
represents a sensor and contains the areas that the sensor monitors from S, and a positive integer k ≥ 2.
The goal is to find a partition G of the subsets into k covers c1 , ..., ck where each cover is a set of subsets,
such that N(G) = ki=1 |∪ Sj ∈ ci Sj | is minimized.
The heuristic used in our paper consists of greedily picking sets from S. The set Sj ∈ S that contains
the most number of uncovered vertices is picked at each step. We repeat the process until all vertices
are covered. The following is the pseudo-code of the greedy set cover algorithm.
 FutureInternet
Future Internet2017,
2017,9,9,9595 8 8ofof1818

Greedy
Greedy
Future InternetSet
Set Cover
Cover
2017, Algorithm
9, 95 Algorithm 8 of 18

1.1. U←
U← ////UUisisthe
theset
setofofuncovered
uncoveredelements
elements
S’←∅∅////S’S’isisthe
2.2. S’← thecurrent
currentset
setcover
cover
Greedy Set Cover Algorithm
3.3. While U≠∅∅Do
WhileU≠ Do
1 U ← X // U is the set of uncovered elements
j j∈∈ such |Sj∩∩ ||isismaximized
24.4. S’ ← ∅ //Pick
Pick isS
S’ S suchthat
the current that
set |Sj
cover maximized
35.5. While US’←S’ ∪
6=S’←S’ ∪
∅ Do SSj j
46.6. Pick Sj ∈ Sj jsuch that |Sj ∩ U| is maximized
U←U−S
U←U−S
end S’←S’ ∪ Sj
57.7. end
6 U ← U − Sj
Once
7Once theset
the
end setcovers
covers(clusters)
(clusters)areareconstructed,
constructed,some somesensor
sensornodes
nodesbecome
becomecluster
clusterheads
heads(CHs)
(CHs)
andcollect
and collectall alltraffic
trafficfrom
fromtheir
theirrespective
respectivecluster.
cluster.The
Thecluster
clusterhead
headaggregates
aggregatesthe thecollected
collecteddata
dataandand
then
then sends
sends
Once ititto
the tothe
set the basestation
base
covers stationasasshown
(clusters) shown
are ininFigure
Figure3.
constructed, 3.The
someThesensor
CHisischosen
CH chosenbased
nodes basedon
become on thehighest
the
cluster highest residual
headsresidual
(CHs)
energy,
energy, thus
thus allowing
allowing the
the network
network lifetime
lifetime to
to increase
increase inin proportion
proportion to
to node
node
and collect all traffic from their respective cluster. The cluster head aggregates the collected data and density
density [43].
[43].
then sendsWhen
When using
itusing
to clustering,
theclustering,
base the
stationthe workload
as workload
shown onthe
on
in Figure the cluster
3.cluster
The CH head
head isisthus
is chosen thus larger
larger
based than
on than
the fornon-cluster
for
highest non-cluster
residual
heads.
heads. To
To maximize
maximize the
the lifetime
lifetime of
of cluster
cluster heads,
heads, wewe propose
propose to
to
energy, thus allowing the network lifetime to increase in proportion to node density [43]. elect
elect two
two cluster
cluster heads
heads within
within each
each
cluster
cluster that
that have
have the
the highest
highest energy,
energy, and
and interchange
interchange their
their roles
roles inina a round-robin
round-robin
When using clustering, the workload on the cluster head is thus larger than for non-cluster heads. fashion
fashion during
during their
their
lifetime
lifetime
To maximize ininorder
order totodistribute
the lifetime distribute and
and
of cluster balance
balance
heads, we thetheextra
extra
propose workload
toworkload
elect two and andenergy
cluster energy consumption
headsconsumption evenly
evenly
within each cluster
between
between them.
them.
that have the highest energy, and interchange their roles in a round-robin fashion during their lifetime
in order to distribute and balance the extra workload and energy consumption evenly between them.

Cluster
Cluster Head
Head Sensor
Sensor Node
Node S S Base
Base
StaStaonon

Figure3.3.Cluster
Figure Clustercreation
creationwith
withcluster
clusterhead
headelected.
elected.
Figure 3. Cluster creation with cluster head elected.

3.3.Energy
3.3. EnergyConsumption
ConsumptionModel
Model
3.3. Energy Consumption Model
Theenergy
The energyconsumption
consumptionmodel
modeladopted
adoptedininthis
thispaper
paperisisbased
basedon onthe
theproposed
proposedmodel
modelinin[44]
[44]
The energy consumption model adopted in this paper is based on the proposed model in [44]
whichhas
which hasbeen
beenused
usedby
bymany
manyresearchers
researchers[45–50].
[45–50].InInthis
thismodel,
model,ititisisassumed
assumedthat
thatthe
thetransmitter,
transmitter,
which has been used by many researchers [45–50]. In this model, it is assumed that the transmitter,
poweramplifier,
power amplifier,and
andreceiver
receiverall
alldissipate
dissipateenergy
energytotorun
runthe
theradio
radioelectronics
electronicsasasshown
shownininFigure
Figure4.4.
power amplifier, and receiver all dissipate energy to run the radio electronics as shown in Figure 4.

Figure 4. Radio Energy Dissipation Model.

Figure4.4.Radio
Figure RadioEnergy
EnergyDissipation
DissipationModel.
Model.
Future Internet 2017, 9, 95 9 of 18

The power attenuation is dependent on the distance between the transmitter and the receiver.
For relatively short distances, the propagation loss can be modeled as inversely proportional to d2 ,
whereas for longer distances the propagation loss can be modeled as inversely proportional to d4 .
Power control can be used to invert this loss by setting the power amplifier to ensure a certain power
at the receiver. Thus, to calculate the transmission cost (joule/bit) and the receiving cost (joule/bit)
for a message of size k bits from a sensor node i (transmitter: Tx ) to a sensor node j (receiver: Rx ) over
a distance d(i,j) (Euclidian-Distance), where the energy consumption at node i comprises operating
radio electronics and amplifiers while at node j only radio electronics, the following equations are used:
Transmission cost:

ETX (k, d) = k × ETX-elec + ETX-amp (k, d) = k × ETX-elec + k × d2 × Eamp , (1)

Receiving cost:
ERX (k, d) = k × ERX-elec , (2)

where Eelec is the energy dissipation of the radio in order to run the transmitter and receiver circuitry,
and Eamp is the transmitter amplifier.
Typical values used in this paper during the simulation are:
Eelec = 50 nJ/bit, Eamp = 100 pJ/bit/m2 , d2 = 500, and the packet size k = 2000 bits [49].
If we consider that node i wants to transmit an amount of data fij to node j, the energy consumption
(Joules) at node i will be:
n
ETX,I = ∑


ETX,ij ∗ f ij (3)
j=1
i 6= j
On the other hand, if node j wants to send an amount of data fji to node i, the energy consumption
(Joules) at node i will be:
n
ERX,I = ERX ∑ f (4)
j = 1 ji
i 6= j
Thus, the total energy consumed at node i will be:

Econsu,i = ETX,i + ERX,I (5)

3.4. Energy Optimization

Once the clusters are created, our objective is to minimize the energy consumption when routing
data to the base station. First, we minimize energy consumption within each cluster, and then we
minimize the routing energy consumption.

3.4.1. Minimizing Energy Consumption in Clusters

The objective of using a SET K-COVER approach to construct clusters in WSNs is to increase their
energy efficiency. The reason for the common placement of multiple sensors close together to cover
a single area is due to the ad-hoc nature of sensor placement, topological constraints, or to compensate
for the short lifetime of the sensors.
Therefore, in an effort to increase the longevity of the network and to conserve battery power, it is
beneficial to activate groups of sensors in rounds, so that the battery life of a sensor is not wasted on
areas that are already monitored by other sensors. Additionally, certain batteries last up to twice as
long when used in short bursts as opposed to continuously. Therefore, activating a sensor only once
every k time units can extend the lifetime of its battery. For the active nodes inside a cluster, an energy
cost ei is associated with a sensor node xi each time the latter sends a message to the CH.
Future Internet 2017, 9, 95 10 of 18

Thus, the energy consumption for a cluster can be formulated as a linear programming
optimization problem where:
k
Minimizing ∑ i =1 x i ∗ ei , (6)

where k is the number of active nodes inside the cluster.

St:
ei ≤ Einit (7)
(
1 i f active
xi = (8)
0 i f inactive
k
∑1 ei ≤ Eclust (9)

where Einit is the initial energy of node xi , where Eclust is the initial cluster total energy of all nodes.
Output: vector X = (x1 , x2 , . . . xn ) for nodes active and inactive.

3.4.2. Routing Energy Optimization

It is to note that in order to achieve the effective collection of data, WSNs must meet two
requirements. One is full coverage of the targets, and the other is complete connectivity of the
network. Full coverage means that the sensors should be able to monitor all targets, and complete
connectivity means that the data that originated from the monitoring targets should be able to reach
the base station through multi-hop wireless communication using a minimum energy path.
We believe that relying only on a minimum energy routing strategy has some deficiency in terms
of network lifetime. Since most of the time using a relay node is more energy-efficient than direct
communication with the base station, sensor nodes closer to the base station exhaust their batteries
long before other nodes residing on the perimeter of the network area. Hence, network lifetime
cannot be optimized with minimum energy routing alone. As a solution, we propose to take into
consideration the residual energy of the nodes in the path by assigning link costs inversely proportional
to the residual energy values. This way, energy consumption could be balanced throughout the whole
network. In fact, such a solution can achieve maximal network lifetime using a linear programming
(LP) based model.

Network Model
After clustering, the routing procedure is invoked during data transmission using a min-cost
max-flow algorithm [51].
The Wireless Sensor Network used in this paper is modeled as a directed graph G (V, E), where V
represents the set of sensor nodes and E the set of edges.
Each edge (u, v) ∈ E is associated with a cost c (u,v) which represents the energy consumption
when there is a flow fij going out from u to v, a capacity constraint uij expressed in the number of
packets per unit time, and a supplies/demands variable dij . The cost of a flow is: f (u,v). c (u,v).
The objective is to deliver all the data generated by the sensor nodes to the base station with
minimum energy consumption and without exceeding the link capacities.
The problem can be formulated as follows:
Minimizing the total cost of the flow over all edges:

∑ f (u, v).c(u, v) (10)

(u,v)∈ E

St:
f (u, v) ≤ c(u, v) ∀ (u, v) ∈ E (11)

f (u, v) = − f (v, u) (12)

Future Internet 2017, 9, 95 11 of 18

∑w∈V f (u, v) = 0 f or all u 6= s, t (13)

∑w∈V f (v, w)− ∑ f (w, v) = d (14)

w ∈V
Future
whereInternet 2017, 9, 95
(11) represents 11 of 18
the capacity constraint, (12) the skew symmetry, (13) the flow conservation,
and (14) the required flow.
The above problem can be converted into a minimum-cost circulation problem [52] and can be
The above problem can be converted into a minimum-cost circulation problem [52] and can
solved efficiently using well-known min-cost flow algorithms [53–56]. In this research we adopt a
be solved efficiently using well-known min-cost flow algorithms [53–56]. In this research we adopt
network simplex method [57] as it is considered the most practical algorithm [57]. Figure 5 shows a
a network simplex method [57] as it is considered the most practical algorithm [57]. Figure 5 shows
sample of a sensor network where s1 is the source and s6 is the destination. The flow capacities and
a sample of a sensor network where s1 is the source and s6 is the destination. The flow capacities and
demands are shown on the graph.
demands are shown on the graph.

Figure 5. WSN graph representation.

Figure 5. WSN graph representation.
The corresponding network simplex LP is as follows, where xij = Number of units shipped from
The corresponding network simplex LP is as follows, where xij = Number of units shipped from
node i to j using arc i–j:
node i to j using arc i–j:
Min z =zx=13x+13 2x
Min + 2x + +2x2x
21 21 2424++8x
8x3232 + 4x35
+ 4x 3x5454++x64x64
35 ++3x + 2x
+ 2x 65 65

Nodes1s:1:xx1313 −
Node − xx2121==11
Node s2 : x21 + x24 − x32 = 2
Node s2: x21 + x24 − x32 = 2
Node s3 : −x13 + x32 + x35 = 3
Node s3: −x13 + x32 + x35 = 3
Node s4 : −x24 − x54 − x64 = −7
Node
Node s5 : s−
4: x
−x3524+−xx54 − xx6465==−7−3
54 −

Node
Node s5: s−x
6 : 35x64
+ x+54x−65x65
= =4 −3

x13 ≤ 2x21 ≤ 2x24 ≤ 5x32 ≤ 5x35 ≤ 4x54 ≤ 3x64 ≤ 1x65 ≤ 6

Node s6: x64 + x65 = 4
4. Simulation Results
x13 ≤ 2x21 ≤ 2x24 ≤ 5x32 ≤ 5x35 ≤ 4x54 ≤ 3x64 ≤ 1x65 ≤ 6
In order to assess our model, we compared it to the MO-SCP, MDP-MSC, and LEACH models.
We performed a set of simulation scenarios in MATLAB [58] using a variable number of sensor
4. Simulation Results
nodes (100, 200, 300, 400) deployed randomly in a 1500 m2 area. This allowed us to evaluate the
In order to assess our model, we compared it to the MO-SCP, MDP-MSC, and LEACH models.
We performed a set of simulation scenarios in MATLAB [58] using a variable number of sensor
nodes (100, 200, 300, 400) deployed randomly in a 1500 m2 area. This allowed us to evaluate the
effectiveness of our proposed approach and analyze its impact on the energy of the entire network,
and thus on the lifetime of the WSN. Parameters used in the simulation are similar to those used in
Future Internet 2017, 9, 95 12 of 18

effectiveness of our proposed approach and analyze its impact on the energy of the entire network,
and thus on the lifetime of the WSN. Parameters used in the simulation are similar to those used in [59],
and are shown in Table 1.

Table 1. Simulation parameters.

Parameters Value
Initial energy Einit 0.5 J
Transmitting and receiving energy Eelec 5 nJ/bit
Amplification energy for short distance Ef 10 Pj/bit/m2
Amplification energy for long distance Eamp 0.013 pJ/bit/m4
Packet size 200 bits
Operating frequency 2.4 GHz
Initial energy of sensor 61,560 J
Transmit power 32 mW
Reference distance 3m
Path loss exponent 4
Signal to Noise Ratio 2 dB

Our experiment consists of two phases of WSN network design: cluster set cover creation and
path determination.
In both phases, we input the set of SNs, their locations, the required network lifetime, and other
parameters to the corresponding SCMC model. We then solved the WSN design model with the IBM®
ILOG® OPL-CPLEX® optimization solver.
In the first phase the set cover model is used to obtain the optimal number of set covers with their
corresponding CH, and the optimal packet transmission path from SNs to the selected CH inside each
cluster. In the second phase we use the results obtained from the first phase to derive the optimal
path between the CH and the BS using the min-cost max-flow algorithm. The metrics used to test the
effectiveness of our proposed approach are based on the percentage of Alive nodes, the round where
the First Node Dies (FND), the round where the Last Node Dies (LND), and the network throughput.
The results of the first phase are shown in Table 2. The Greedy algorithm creates a number of set
covers for each scenario where the number of sensor nodes is varied (N = 100, 200, 300 and 400)

Table 2. Set covers number created by Greedy algorithm.

Number of Nodes Number of Clusters

100 7
200 8
300 7
400 8

In the second phase, the min-cost max-flow model is solved as stated before and a comparison is
carried out between the different approaches on the basis of the metrics stated earlier. Tables 3 and 4
summarize the results of these metrics.

Table 3. Number of rounds when 10%, 20%, 30%, 40%, 50%, 60%, 70% and 80% of nodes die.

Percentage of Alive Nodes

Protocol
90 80 70 60 50 40 30 20
LEACH 1100 1642 1600 1465 2278 2500 2986 2987
MDP-MSC 1600 1855 1900 1978 2436 2875 3100 3300
MO-SCP 1756 1860 2150 2360 2780 3164 3600 3785
SCMC 1865 1890 2260 2600 2830 3700 4300 4500
Future Internet 2017, 9, 95 13 of 18

Table 4. Nodes die-off statistics of the different approaches.

Protocol FND LND

LEACH 1019 1986
MDP-MSC 1125 3125
MO-SCP 1019 4035
SCMC 975 9800
Future Internet 2017, 9, 95 13 of 18

Table 3 shows the number of rounds completed by various schemes when 10%, 10%, 20%,
20%, 30%,
30%, 40%,
40%,
50%, 60%, 70% and and 80%
80% ofof nodes
nodes diedie for
for N = = 100.
It is evident, as shown in Table 3, 3, that SCMC has a better better performance
performance than its counterpart
protocols. For
protocols. For instance,
instance, inin the
the case
case of
of SCMC
SCMC the the number
number of of rounds
rounds when
when 80%80% ofof nodes
nodes are dead is
are dead is
4500
4500 whereas
whereas in in the case of
the case of MO-SCP,
MO-SCP, MDP-MSC,
MDP-MSC, and and LEACH
LEACH the the number
number is is 3785,
3785, 3300,
3300, and
and 2987
2987
respectively.
respectively. Therefore,
Therefore, SCMCSCMCextends
extendsthe thelifetime
lifetimeofofthetheWSN
WSNand andcan
canprovide
provideinformation
information forfora
considerably
a considerablylonger
longerperiod
periodofoftime.
time.This
Thislifetime
lifetimeincreases
increaseswith
withnode
nodedensity,
density, where
where SCMC
SCMC takes
takes the
the
advantage of the set cover formation and the optimal
advantage of the set cover formation and the optimal routing energy. routing energy.
Despite
Despite the
the fact
fact that
that SCMC
SCMC has has its
its first
first node
node dead
dead inin an
an earlier
earlier round
round than
than the
the other
other approaches,
approaches,
it shows very
it shows very promising
promising results,
results, with
with the
the death
death of of all
all nodes
nodes as
as shown
shown in in Table
Table 44 occurring
occurring at at about
about
9800 rounds compared to 4035 for MO-SCP, 3125 for MDP-MSC, and 1986
9800 rounds compared to 4035 for MO-SCP, 3125 for MDP-MSC, and 1986 for LEACH. Furthermore, for LEACH. Furthermore,
we clearly
we clearly see
see that
that when
when 20%20% of
of nodes
nodes atat the
the same
same number
number of of rounds
rounds inin SCMC
SCMC are alive, all
are alive, all nodes
nodes in in
the
the other approaches are
other approaches are dead.
dead.
In
In the
the second
second experiment
experiment we we calculated
calculated the the throughput
throughput for for all
all schemes
schemes with
with aa different
different variation
variation
of
of sensor
sensor nodes
nodesfromfrom100100toto400
400with
withanan increment
increment of of
100. WeWe
100. observe as shown
observe as shownin Figures 6–9 that
in Figures 6–9
SCMC outperforms other algorithms in terms of a higher throughput and
that SCMC outperforms other algorithms in terms of a higher throughput and a larger number of a larger number of packets
receivedreceived
packets at the base
at thestation, which which
base station, is understandable
is understandable as SCMC has ahas
as SCMC longer lifetime
a longer andand
lifetime it uses an
it uses
optimized
an optimized routing
routingpath compared
path compared to other
to other algorithms.
algorithms.

Figure 6.
6. Throughput
Throughputcomparison
comparison of of
SetSet Cover
Cover Min-Cost
Min-Cost max-flow
max-flow (SCMC),
(SCMC), Multi-Objective
Multi-Objective Set
Set Cover
Cover
ProblemProblem (MO-SCP),
(MO-SCP), MinimumMinimum
DistanceDistance Path-Maximum
Path-Maximum Set (MDP-MSC),
Set Cover Cover (MDP-MSC), and Low-
and Low-Energy
Energy Adaptive
Adaptive Clustering
Clustering Hierarchy
Hierarchy (LEACH) (LEACH). (Size = 100).
(Size = 100).

Based on the previous simulation results, we have shown that SCMC provides a higher
throughput and a prolonged lifetime of the WSN compared to MO-SCP, MDP-MSC, and LEACH.
This is due to the fact that SCMC uses a set cover approach to create clusters, and min-cost max-flow
to route data between CHs and the base station. Although MDP-MSC also uses the concept of set
Future Internet 2017, 9, 95 14 of 18

Based on the previous simulation results, we have shown that SCMC provides a higher throughput
and a prolonged lifetime of the WSN compared to MO-SCP, MDP-MSC, and LEACH. This is due
to the fact that SCMC uses a set cover approach to create clusters, and min-cost max-flow to route
data between CHs and the base station. Although MDP-MSC also uses the concept of set covers
to create clusters, it remains different from MO-SCP and SCMC in three ways. Firstly, it searches
for k paths from the sink to all sensor nodes while MO-SCP and SCMC use a single path approach.
Secondly, it calculates maximum disjoint set covers while MO-SCP and SCMC calculate a minimum
set cover. Furthermore, MDP-MSC doesn’t consider the problem of extending the lifetime of WSNs
Future
FutureInternet
Internet2017,
2017,9,9,95
95 14
14of
of18
18
as an optimization problem, while MO-SCP considers it as a multi-objective optimization problem,
and SCMC as a single
problem,
problem, and
and SCMC objective
SCMC as aa problem.
as single In SCMC
single objective
objective the energy
problem.
problem. In consumption
In SCMC
SCMC the
the energy is well distributed
energy consumption
consumption is well among
is well
distributed
nodes and among
routingamong
distributed is done nodes and
in an
nodes routing
andefficient is done
routing isway. in an efficient way.
done in an efficient way.

7. Throughput
FigureFigure
Figure 7. comparison
7.Throughput
Throughput comparisonofof
comparison ofSCMC,
SCMC, MO-SCP,
SCMC,MO-SCP, MDP-MSC
MO-SCP,MDP-MSC
MDP-MSC andand
and LEACH
LEACH.
LEACH. (Size=(Size
(Size 200).= 200).
=200).

8. Throughput
FigureFigure
Figure 8. comparison
8.Throughput
Throughput comparisonofof
comparison ofSCMC,
SCMC, MO-SCP,
SCMC,MO-SCP, MDP-MSC
MO-SCP,MDP-MSC
MDP-MSC andand
and LEACH
LEACH.
LEACH. (Size=(Size
(Size 300).= 300).
=300).
Future Internet 2017, 9, 95 15 of 18
Future Internet 2017, 9, 95 15 of 18

Figure 9. Throughput comparison of SCMC, MO-SCP,

MO-SCP, MDP-MSC
MDP-MSC and
and LEACH
LEACH.(Size
(Size==400).
400).

5. Conclusions
5. Conclusions
In this
In this paper,
paper, weweextensively
extensively studied
studied the
the problem
problem ofof minimizing
minimizing thethe energy
energy consumption
consumption and and
maximizing
maximizing the lifetime of WSNs. For this purpose, an effective approach was presented
of WSNs. For this purpose, an effective approach was presented using using an
optimization
an optimization model.
model. The
Theproposed
proposed approach
approachcontrols
controlsthe
theenergy
energydepletion
depletion of of sensor nodes and and
minimizes the routing path using the residual energy of the nodes. The corresponding
minimizes the routing path using the residual energy of the nodes. The corresponding optimization optimization
objective functions
objective functions were
were solved
solved using
using linear
linear programing,
programing, and
and the
the simulation
simulation results
results showed
showed that
that the
the
proposed scheme
proposed scheme (SCMC)
(SCMC) outperforms
outperforms MO-SCP,
MO-SCP,MDP-MSC,
MDP-MSC, and and LEACH
LEACH in in terms
terms ofof WSN
WSN lifetime
lifetime
andthroughput.
and throughput.AsAs a future
a future project,
project, we intend
we intend to apply
to apply our approach
our approach to base
to various various base
station station
scenarios.
scenarios. Additionally,
Additionally, the use of athe use of a multi-objective
multi-objective optimization optimization
model would model
be ofwould be of great interest.
great interest.
Acknowledgments: Project funded by Research Translation Center (RTC), Grant No. GP-CCIS-2013-11-17.
Acknowledgments: Project funded by Research Translation Center (RTC), Grant No. GP-CCIS-2013-11-17.
Conflictsof
Conflicts ofInterest:
Interest: The
The author
author declares
declaresno
no conflict
conflictof
of interest.
interest. The
The founding
founding sponsors
sponsorshadhadno
no role
rolein
in the
the design
design
of the
of the study;
study; in
in the
the collection,
collection, analyses,
analyses, or
or interpretation
interpretation ofof data;
data; in
in the
the writing
writing of
of the
the manuscript,
manuscript, and
and in
in the
the
decision
decision toto publish
publish thetheresults.
results.

References
References
1.
1. Rohankar,R.;
Rohankar, R.; Katti,
Katti, C.P.;
C.P.;Kumar,
Kumar,S. S.Comparison
Comparisonof of Energy
Energy Efficient
Efficient Data
Data Collection
Collection Techniques
Techniques inin Wireless
Wireless
Sensor Network.
Sensor Network. Procedia
Procedia Comput.
Comput. Sci. Sci. 2015, 57, 146–151.
146–151, [CrossRef]
doi:10.1016/j.procs.2015.07.399.
2.
2. More, A.; Raisinghani,
More, Raisinghani, V. V. A survey
survey on on energy
energyefficient
efficientcoverage
coverageprotocols
protocolsininwireless
wirelesssensor
sensornetworks.
networks.J.
J.King
KingSaud
SaudUniv.
Univ.Comput.
Comput.Inf. Inf.Sci.
Sci.2017,
2017,29,29,428–448,
428–448.doi:10.1016/j.jksuci.2016.08.001.
[CrossRef]
3.
3. Touati,
Touati, Y.;
Y.;Ali-Chérif,
Ali-Chérif, A.; A.; Daachi,
Daachi, B. B.Routing
RoutingInformation
Information for
forEnergy
EnergyManagement
Management in inWSNs.
WSNs. In In Energy
Energy
Management
ManagementininWireless
WirelessSensor Networks;
Sensor Elsevier:
Networks; Amsterdam,
Elsevier: The Netherlands,
Amsterdam, 2017; pp. 23–51.
The Netherlands, [CrossRef]
2017; pp. 23–51,
4. Wang, K.; Ou, Y.; Ji, H.; Zhang,
doi:10.1016/B978-1-78548-219-9.50003-5. H.; Li, X. Energy aware hierarchical cluster-based routing protocol for WSNs.
4. J.Wang,
China K.;
Univ. Posts
Ou, Y.; Telecommun.
Ji, H.; Zhang,2016, H.; 23,
Li, 46–52. [CrossRef]
X. Energy aware hierarchical cluster-based routing protocol for
5. Elhoseny,
WSNs. J. M.; Tharwat,
China A.; Yuan,
Univ. Posts X.; Hassanien,
Telecommun. A.E.46–52,
2016, 23, Optimizing K-coverage of mobile WSNs. Expert Syst. Appl.
doi:10.1016/S1005-888560044-4.
5. 2018, 92, 142–153.
Elhoseny, [CrossRef]
M.; Tharwat, A.; Yuan, X.; Hassanien, A.E. Optimizing K-coverage of mobile WSNs. Expert Syst.
Appl. 2018, 92, 142–153, doi:10.1016/j.eswa.2017.09.008.
Future Internet 2017, 9, 95 16 of 18

6. Naeem, M.; Iqbal, M.; Anpalagan, A.; Ahmad, A.; Obaidat, M.S. Chapter 15—Optimization classification and
techniques of WSNs in smart grid. In Smart Cities and Homes; Obaidat, M., Nicopolitidis, P., Eds.; Morgan
Kaufmann: Burlington, MA, USA, 2016; pp. 323–343. [CrossRef]
7. Kuila, P.; Jana, P.K. Energy efficient clustering and routing algorithms for wireless sensor networks: Particle
swarm optimization approach. Eng. Appl. Artif. Intell. 2014, 33, 127–140. [CrossRef]
8. Singh, B.; Lobiyal, D.K. Energy-aware Cluster Head Selection Using Particle Swarm Optimization and
Analysis of Packet Retransmissions in WSN. Procedia Technol. 2012, 4, 171–176. [CrossRef]
9. Obaidy, M.A.; Ayesh, A. Energy efficient algorithm for swarmed sensors networks. Sustain. Comput. Inf. Syst.
2015, 5, 54–63. [CrossRef]
10. Azharuddin, M.; Jana, P.K. Particle swarm optimization for maximizing lifetime of wireless sensor networks.
Comput. Electr. Eng. 2016, 51, 26–42. [CrossRef]
11. Mortazavi, S.H.; Salehe, M.; MacGregor, M.H. Maximum WSN coverage in environments of heterogeneous
path loss. Int. J. Sens. Netw. 2014, 16, 185–198. [CrossRef]
12. Peng, J.; Chen, Y. A low energy consumption WSN node. Int. J. Embed. Syst. 2015, 7, 318–323. [CrossRef]
13. Zhou, Y.; Wang, X.; Wang, T.; Liu, B.; Sun, W. Fault-tolerant multi-path routing protocol for WSN based on
HEED. Int. J. Sens. Netw. 2016, 20, 37–45. [CrossRef]
14. Gupta, S.; Roy, K.C. Q-Coverage Maximum Connected Set Cover (QC-MCSC) Heuristic for Connected Target
Problem in Wireless Sensor Network. Glob. J. Comput. Sci. Technol. 2015, 15, 470–474.
15. Jain, S.; Gupta, S. Analysis and Comparison of Wireless Sensor Networks Coverage Algorithms. Int. J. Comput.
Sci. Inf. Technol. 2014, 5, 1668–1670.
16. Attea, B.A.; Khalil, E.A.; Ozdemir, S.; Yıldız, O. A multi-objective disjoint set covers for reliable lifetime
maximization of wireless sensor networks. Wirel. Pers. Commun. 2015, 81, 819–838. [CrossRef]
17. Jia, J.; Chena, J.; Chang, G.; Wena, Y.; Song, J. Multi-objective optimization for coverage control in wireless
sensor network with adjustable sensing radius. Comput. Math. Appl. 2009, 57, 1767–1775. [CrossRef]
18. Ji, S.; Pan, Y.; Li, Y. Reliable and Energy Efficient Target Coverage for Wireless Sensor Networks. Tsinghua Sci. Technol.
2011, 16, 464–474.
19. Diop, B.; Diongue, D.; Thiaré, O. Greedy Algorithms for Target Coverage Lifetime Management Problem in
Wireless Sensor Networks. Int. J. Control Autom. 2015, 8, 232–250. [CrossRef]
20. Younis, O.; Krunz, M.; Ramasubramanian, S. Location-Unaware coverage in wireless sensor networks.
Ad Hoc Netw. 2008, 6, 1078–1097. [CrossRef]
21. Yao, L.; Zeng, F. Maximum Set Covers Based Energy Conservation Scheme in Wireless Sensor Networks.
Tech. Gaz. 2014, 21, 733–739.
22. Cardei, M.; Thai, M.T.; Li, Y.; Wu, W. Energy-efficient target coverage in wireless sensor networks. In Proceedings
of the 24th Annual Joint Conference of the IEEE Computer and Communications Societies (INFOCOM 2005),
Miami, FL, USA, 13–17 March 2005; IEEE: Piscataway, NJ, USA.
23. Liu, L.; Hu, B.; Miao, H.; Li, H.; Li, L.; Zhao, Q. Achieving Energy Conservation, Coverage and Connectivity
Requirements in Wireless Sensor Networks. In Proceedings of the 29th IEEE International Conference on
Distributed Computing Systems Workshops, Montreal, QC, Canada, 22–26 June 2009.
24. Idrees, A.K.; Deschinkel, K.; Salomon, M.; Couturier, R. Distributed lifetime coverage optimization protocol
in wireless sensor networks. J. Supercomput. 2015, 71, 4578–4593. [CrossRef]
25. Belghith, O.B.; Sbita, L. Extending the network lifetime of wireless sensor networks using fuzzy logic.
In Proceedings of the IEEE 12th International Multi-Conference on Systems, Signals & Devices, Mahdia,
Tunisia, 16–19 March 2015.
26. Handy, M.J.; Haase, M.; Timmermann, D. Low energy adaptive clustering hierarchy with deterministic
cluster-head selection. In Proceedings of the 4th International Workshop on Mobile and Wireless
Communications Network, Stockholm, Sweden, 9–11 September 2002; pp. 368–372.
27. Mamdani, E.H.; Assilian, S. An experiment in linguistic synthesis with a fuzzy logic controller. Int. J. Man-Mach. Stud.
1975, 7, 1–13. [CrossRef]
28. Zorbas, D.; Glynos, D.; Kotzanikolaou, P.; Douligeris, C. Solving coverage problems in wireless sensor
networks using cover sets. Ad Hoc Netw. 2010, 8, 400–415. [CrossRef]
29. Jia, J.; Chen, J.; Chang, G.-R.; Wen, Y.-Y. Efficient cover set selection in wireless sensor networks. Acta Autom. Sin.
2008, 34, 1157–1162. [CrossRef]
Future Internet 2017, 9, 95 17 of 18

30. Deb, K.; Pratap, A.; Agarwal, S.; Meyarivan, T. A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II.
IEEE Trans. Evol. Comput. 2002, 6, 182–197. [CrossRef]
31. Attea, B.A. Multi-Objective Set Cover Problem for Reliable and Efficient Wireless Sensor Networks. Iraqi J. Sci.
2015, 56, 1147–1160.
32. Clark, B.N.; Colbourn, C.J.; Johnson, D.S. Unit disk graphs. Discret. Math. 1990, 86, 165–177. [CrossRef]
33. Purohit, G.N.; Verma, S.; Sharma, U. Powers of a Graph and Associated Graph Labeling. Int. J. Comput. Netw. Secur.
2010, 2, 45–49.
34. Jia, J.; Chen, J.; Chang, G.; Li, J. Find the maximum k-disjoint coverage sets in WSN using genetic algorithm.
Int. J. Model. Identif. Control 2010, 9, 43–52. [CrossRef]
35. Prabakaran, T.; Karthikeyan, N.K. Adaptive clustering and routing using fuzzy decision model in WSN.
Int. J. Mob. Netw. Des. Innov. 2015, 6, 51–62. [CrossRef]
36. Ali, N.A.; ElSayed, H.M.; El-Soudani, M.; Amer, H.H.; Daoud, R.M. Elongation of WSN lifetime using
a centralised clustering technique. Int. J. Syst. Control Commun. 2012, 4, 250–261. [CrossRef]
37. Karp, R.M. Reducibility among Combinatorial Problems. In Complexity of Computer Computations; Miller, R.E.,
Thatcher, J.W., Eds.; Plenum: New York, NY, USA, 1972; pp. 85–103.
38. Gomes, F.C.; Meneses, C.N.; Pardalos, P.M.; Viana, G.V.R. Experimental analysis of approximation algorithms
for the vertex cover and set covering problems. Comput. Oper. Res. 2006, 33, 3520–3534. [CrossRef]
39. Lan, G.; DePuy, G.W.; Whitehouse, G.E. An effctive and simple heuristic for the set covering problem.
Europ. J. Oper. Res. 2007, 176, 1387–1403. [CrossRef]
40. Ruan, L.; Du, H.; Jia, X.; Wu, W.; Li, Y.; Ko, K.I. A greedy approximation for minimum connected dominating
sets. Theor. Comput. Sci. 2004, 32, 325–330. [CrossRef]
41. Bilal, N.; Galinier, P.; Guibault, F. A New Formulation of the Set Covering Problem for Metaheuristic
Approaches. Int. Sch. Res. Not. 2013, 1–10. [CrossRef]
42. Akhter, F. A Heuristic Approach for Minimum Set Cover Problem. Int. J. Adv. Res. Artif. Intell. 2015, 4, 40–45.
[CrossRef]
43. Xu, Y.; Heidemann, J.; Estrin, D. Geography-informed energy conservation for Ad Hoc routing. In Proceedings
of the 7th Annual International Conference on Mobile Computing and Networking, Rome, Italy, 16–21 July 2001;
pp. 70–84.
44. Heinzelman, W. Application Specific Protocol Architectures for Wireless Networks. Ph.D. Thesis, Massachusetts
Institute of Technology (MIT), Cambridge, MA, USA, 2000.
45. Prajapat, M.; Barwar, N.C. Performance Analysis of Energy Dissipation in WSNs Using Multi-Chain PEGASIS.
Int. J. Comput. Sci. Inf. Technol. 2014, 5, 8033–8036.
46. Acharya, A.; Seetharam, A.; Bhattacharyya, A.; Naskar, M.K. Balancing energy dissipation in data gathering
wireless sensor networks using ant colony optimization. In Proceedings of the International Conference on
Distributed Computing and Networking, Hyderabad, India, 3–6 January 2009; pp. 437–443.
47. Lindsey, S.; Raghavendra, C.S. PEGASIS: Power Efficient Gathering in Sensor Information Systems. In Proceedings
of the IEEE ICC, Big Sky, MT, USA, 9–16 March 2001; pp. 1125–1130.
48. Tan, H.O.; Korpeoglu, I. Power efficient data gathering and aggregation in wireless sensor networks.
ACM SIGMOD Rec. 2003, 32, 66–71. [CrossRef]
49. How Can I Start for Calculating Energy Dissipation in WSN? ResearchGate. Available online: https:
//www.researchgate.net/post/How_can_I_start_for_calculating_energy_dissipation_in_WSN (accessed on
9 December 2017).
50. Heinzelman, W.; Chandrakasan, A.; Balakrishna, H. Energy-Efficient Communication Protocol for Wireless
Microsensor Networks. In Proceedings of the 33rd Hawaii International Conference on System Sciences,
Maui, HI, USA, 7 January 2000; pp. 1–10.
51. Ahuja, R.; Magnanti, T.; Orlin, J. Network Flows; Prentice Hall: Upper Saddle River, NJ, USA, 1993.
52. Goldberg, A.V.; Tarjan, R.E. Solving minimum-cost flow problems by successive approximation. In Proceedings
of the Nineteenth Annual ACM Symposium on Theory of Computing, New York, NY, USA, 25–27 May 1987.
53. Bixby, R.E.; Wagner, D.K. An almost linear-time algorithm for graph realization. Math. Oper. Res. 1988, 13,
99–123. [CrossRef]
54. Ahuja, R.; Goldberg, A.; Orlin, J.; Tarjan, R. Finding minimum-cost flows by double scaling. Math. Program.
1992, 53, 243–266. [CrossRef]
55. Orlin, J. A faster strongly polynomial minimum cost flow algorithm. Oper. Res. 1993, 41, 338–350. [CrossRef]
Future Internet 2017, 9, 95 18 of 18

56. Kovácsab, P. Minimum-cost flow algorithms: An experimental evaluation. Opt. Methods Softw. 2015, 30,
94–127. [CrossRef]
57. Orlin, J.B. A polynomial time primal network simplex algorithm for minimum cost flows. Math. Program.
1997, 78, 109–129. [CrossRef]
58. MATLAB: R2017b; The MathWorks, Inc.: Natick, MA, USA, 2017.
59. Tounsi, M.; Mahlous, A.R. A Multi-Objective Model for Optimizing Network lifetime in Wireless Sensor
Network. Int. J. Comput. Sci. Inf. Secur. 2016, 14, 562–569.

© 2017 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).