636
3, MARCH 2014
Sensor Deployment and Scheduling for Target
Coverage Problem in Wireless Sensor Networks
S. Mini, Siba K. Udgata, and Samrat L. Sabat
Abstract— Network lifetime plays an integral role in setting up
an efficient wireless sensor network. The objective of this paper
is twofold. The first one is to deploy sensor nodes at optimal
locations such that the theoretically computed network lifetime
is maximum. The second is to schedule these sensor nodes such
that the network attains the maximum lifetime. Thus, the overall
objective of this paper is to identify optimal deployment locations
of the given sensor nodes with a pre-specified sensing range, and
to schedule them such that the network lifetime is maximum with
the required coverage level. Since the upper bound of the network
lifetime for a given network can be computed mathematically,
we use this knowledge to compute locations of deployment such
that the network lifetime is maximum. Further, the nodes are
scheduled to achieve this upper bound. In this paper, we use
artificial bee colony algorithm and particle swarm optimization
for sensor deployment problem followed by a heuristic for
scheduling. A comparative study shows that artificial bee colony
algorithm performs better for sensor deployment problem. The
proposed heuristic was able to achieve the theoretical upper
bound in all the experimented cases.
Index Terms— Wireless sensor networks, target coverage,
sensor deployment, k-coverage, Q-coverage.
I. I NTRODUCTION
W IRELESS Sensor Networks (WSNs) are important for
many applications such as military sensing, physical
security, air traffic control, traffic surveillance, video surveil-
lance, industrial and manufacturing automation, environment
monitoring, and building and structural monitoring [1]. Net-
work lifetime (defined as the time instant from which the
network starts functioning to the time instant where the desired
coverage criterion is not satisfied) is a crucial factor that
determines the efficiency of a wireless sensor network. Energy
usage should be curbed to achieve enhanced lifetime. This is
because sensor nodes are battery powered and cannot be easily
recharged or replaced.
Coverage in a WSN needs to guarantee that the region is
monitored with the required degree of reliability. Locations
of sensor nodes constitute the basic input for the algorithms
Manuscript received July 16, 2013; revised September 23, 2013; accepted
October 9, 2013. Date of publication October 18, 2013; date of current version
January 7, 2014. The associate editor coordinating the review of this paper
and approving it for publication was Prof. Kiseon Kim.
S. Mini is with the Department of Computer Science and Engineering,
School of Engineering and Technology, Central University of Rajasthan,
Rajasthan 305801, India (e-mail: mini2min2002@yahoo.co.in).
S. K. Udgata is with the School of Computer and Information
Sciences, University of Hyderabad, Hyderabad 500046, India (e-mail:
skudgata@gmail.com).
S. L. Sabat is with the School of Physics, University of Hyderabad,
Hyderabad 500046, India (e-mail: slssp@uohyd.ernet.in).
Digital Object Identifier 10.1109/JSEN.2013.2286332
1530-437X © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
IEEE SENSORS JOURNAL, VOL. 14, NO.
that examine coverage of the network [2]. Coverage prob-
lems can be broadly classified as area coverage problem and
target coverage problem. Area coverage focuses on moni-
toring the entire region of interest, whereas target coverage
concerns monitoring only certain specific points in a given
region. Target coverage can be categorized as simple coverage,
k-coverage and Q-coverage. With simple coverage, each tar-
get should be monitored by at least one sensor node. For
k-coverage, each target has to be monitored by at least k sensor
nodes, where k is a predefined integer constant. In Q-coverage,
the target vector T = {T1 , T2 , . . . , Tn } should be monitored by
Q = {q1 , q2 , . . . , qn } number of sensor nodes such that target
T j is monitored by at least q j number of sensor nodes, where
n is the number of targets and 1 ≤ j ≤ n.
There are two types of sensor node deployments: random
deployment and deterministic deployment. Random deploy-
ment is suitable for applications where the details of the
regions are not known, or regions are inaccessible. An example
of random deployment of sensor nodes would be in battlefield
surveillance. In such a deployment, the most common way
of extending the network lifetime is by scheduling the sensor
nodes such that only a subset of sensor nodes that is enough
to satisfy coverage requirement need to be active at a time [3].
In deterministic deployment, the details of the region will
be known apriori and since a provision of deploying nodes
at specific locations prevail, there exists two ways by which
network lifetime can be maximized. One is at deployment
phase and the other is at scheduling phase. Given a region
with targets being monitored by sensor nodes, the upper bound
of network lifetime can be mathematically computed [4], [5].
This information can be used for computing locations which
would be appropriate for coverage to be satisfied as well as
network lifetime to be maximum. Once the deployment loca-
tions are computed, sensor nodes can be scheduled to achieve
the optimum lifetime. Sensor deployment and scheduling in
this way contributes equally to extend the network lifetime.
Hence the problem can be summarized as: Given some sen-
sor nodes that can be deterministically deployed, where to
deploy them and how to schedule them so as to achieve
the required target coverage level and maximize the network
lifetime?
There are several ways of computing deployment locations.
Bio-inspired algorithms prove to be effective for solving opti-
mization problems. In this paper, we use Artificial Bee Colony
(ABC) algorithm [6]–[10] to compute deployment locations.
Apart from ABC algorithm, we also use a heuristic and PSO
(Particle Swarm Optimization) to compute the deployment
MINI et al.: SENSOR DEPLOYMENT AND SCHEDULING FOR TARGET COVERAGE PROBLEM IN WSNs
locations. Though the heuristic performs better than random
deployment, it is not that good as ABC algorithm in maximiz-
ing lifetime. It is observed that ABC algorithm is robust than
PSO algorithm for this problem. After computing the optimal
locations, sensor nodes are scheduled using a heuristic so as
to achieve the theoretical upper bound of network lifetime.
Rest of the paper is organized as follows: In Section II, we
briefly review some of the related work. The problem is formu-
lated formally in Section III. Section IV presents the proposed
method to solve the problem. Section V presents the simulation
results. Concluding remarks in Section VI ends the paper.
A. Importance
Sensor coverage is important while evaluating the effective-
ness of a wireless sensor network. A lower coverage level
(simple coverage) is enough for environmental or habitat
monitoring [11] or applications like home security [12]. Higher
degree of coverage (k-coverage) will be required for some
applications like target tracking to track the targets accu-
rately [11], or if sensors work in a hostile environment such as
battle fields or chemically polluted areas [12]. More reliable
results are produced for higher degree of coverage which
requires multiple sensor nodes to monitor the region/targets.
In some cases, for the same application, the coverage
requirement may vary. For example, for forest fire detections,
the coverage level may be low in rainy seasons, but high
in dry seasons [12]. An example of Q-coverage is a video
surveillance system deployed for monitoring hostile territorial
area where some sensitive targets like a nuclear plant may
need more sensors cooperate to ensure source redundancy for
precise data [4].
Both sensor deployment and scheduling are important to
ensure prolonged network lifetime. Traditionally, the problems
of sensor placement and scheduling have been considered
separately from each other. A balanced performance is crucial
for most applications [13].
Different sensor deployment strategies can cause very dif-
ferent network topology, and thus different degrees of sensor
redundancy [14]. A good sensor deployment with sufficient
number of sensors which ensures a certain degree of redun-
dancy in coverage so that sensors can rotate between active
and sleep modes is required to balance the workload of
sensors [15].
II. R ELATED W ORK
Most of the existing works on sensor deployment problem
focus on area coverage. Environments with obstacles are
also considered for sensor deployment [16]–[18]. From the
perspective of coverage, Onur et al. [19] address quality of the
deployment and propose quality measures, which indicate if
the deployment provides sufficient coverage, or whether rede-
ployment is required or not. Bai et al. [20] use Voronoi-based
methodology to solve deployment problem. Yun et al. [21]
points out the disadvantages of [20] and propose deployment
patterns to achieve full coverage and 3-, 4-, 5-, and 6-connect
ivity for WSNs.
637
Bai et al. [22] study optimal deployment patterns for
more than two-connectivity and full coverage simultaneously.
A Diamond pattern, which could be viewed as a series of
evolving patterns, and another new deployment pattern called
the Double-strip pattern is also proposed. Tan et al. [23]
propose Connectivity-Preserved Virtual Force (CPVF) scheme,
a modified virtual-force-based method which results in poor
coverage in some cases and also describes a Floor based
scheme which overcomes the drawbacks of CPVF.
Cheng et al. [24] explore possible sensor network deploy-
ment strategies that maximize sensor network lifetime by
mitigating the problem of the hot spot around the data sink.
Ozturk et al. [25] use ABC algorithm to solve dynamic
deployment problem in WSNs within the scenario of mobile
and stationary sensors on a probabilistic detection model.
Many sensor node scheduling algorithms [26]–[30] are
proposed to solve area coverage problem. Tilak et al. [31]
investigate sensor networks with directional sensing and com-
munication capability and propose a method for deploying
sensor nodes with directional sensing range, and subsequent
connectivity checking and repairing. A deployment strategy,
with sensors having adjustable sensing ranges to cover an area,
is also proposed.
Unlike the above works on sensor deployment to achieve
area coverage, we look at target coverage with a goal to
maximize network lifetime.
Mini et al. [32] use ABC algorithm to solve simple cov-
erage problem. ABC algorithm was applied to the dynamic
deployment problem in WSNs with mobile sensors on a binary
sensing model by Ozturk et al. [33]. Udgata et al. [34] uses
ABC algorithm for sensor deployment problem in irregular
terrain. This is applicable for regions where the number of
targets is more compared to the number of sensors to be
deployed. The aim is to save energy by minimizing the
sensing range requirement for the sensors. Mini et al. [35]
further extended this work to solve k-coverage and Q-coverage
problems. All these works aim at placing the sensor nodes in
such a way that the required sensing range is minimum.
Gu et al. [36] address the problem of achieving an optimal
network lifetime in surveillance sensor networks, which is an
NP-complete problem. Mini et al. [37] propose a heuristic
to schedule the sensor nodes which maximizes the network
lifetime. The heuristic could achieve the theoretical upper
bound for all experimented cases. In the literature, deployment
and scheduling for coverage problem are addressed mostly as
independent problems and various methodologies have been
proposed for deployment and scheduling separately. In this
paper, we attempt to address deployment and scheduling as
one problem to maximize the network lifetime. We start
with deploying sensor nodes such that the upper bound of
network lifetime could be maximum for the specified coverage
requirement, and then proceed using the heuristic to schedule
the sensor nodes so that the network lifetime upper bound
could be achieved.
III. P ROBLEM D EFINITION
Given a set of n targets T = {T1 , T2 , . . . , Tn } located in
u × v region and m sensor nodes S = {S1 , S2 , . . . , Sm }, place
638
the nodes such that all targets are covered as per the coverage
requirement. The objective is
1) To deploy the sensor nodes such that the network
lifetime is maximum and
2) To schedule the sensor nodes so as to achieve the optimal
network lifetime.
A. Upper Bound of Network Lifetime
Assume m sensor nodes {S1 , S2 , . . . , Sm } randomly
deployed to cover the region R with n targets {T1 , T2 , . . . , Tn }.
Each sensor node has an initial energy E 0 and a sensing
radius, sr . A sensor node Si , 1 ≤ i ≤ m, is said to cover
a target T j , 1 ≤ j ≤ n, if the distance between Si and T j is
less than sr . The coverage matrix is defined as,
  y q j sensor nodes
1 if Si monitors T j
Mi j = (1)
0 otherwise
where i = 1, 2, . . . , m and j = 1, 2, . . . , n.
 Since the sensor nodes can be deterministically deployed,
We define bi = beii to denote the lifetime of battery (in
terms of time), where bi is the initial battery power and ei
is the energy consumption rate of Si . The upper bound is
the maximum achievable network lifetime for a particular
configuration and as stated in [4] and [5], the upper bound
is calculated as,
  pute the optimal deployment locations using ABC algorithm.

i M i j ∗ b i
U = mi n j (2)
qj
For k-coverage, q j = k, j = 1, 2, . . . , n.
B. Sensor Deployment
1) Sensor Deployment to Achieve 1-Coverage: Given a set
of n targets T = {T1 , T2 , . . . , Tn } located in u ×v region and m
sensor nodes S = {S1 , S2 , . . . , Sm }, place the nodes such that
each target is monitored by at least one sensor node and the
network lifetime is maximum. The objective is to maximize
U such that each target is monitored by at least one sensor
node.
2) Sensor Deployment to Achieve k-Coverage: Given a set
of n targets T = {T1 , T2 , . . . , Tn } located in u × v region and
m sensor nodes S = {S1 , S2 , . . . , Sm }, place the nodes such
that each target is monitored by at least k-sensor nodes and to
maximize U .
3) Sensor Deployment to Achieve Q-Coverage: Given a set
of n targets T = {T1 , T2 , . . . , Tn } located in u ×v region and m
sensor nodes S = {S1 , S2 , . . . , Sm }, place the nodes such that
each target T j , 1 ≤ j ≤ n, is covered by at least q j sensor
nodes and to maximize U .
C. Sensor Scheduling
1) 1-Coverage Scheduling: Given m sensor nodes S =
{S1 , S2 , . . . , Sm } with battery power B = {b1 , b2 , . . . , bm },
energy consumption rate ei for Si and n targets T =
{T1 , T2 , . . . , Tn }, find a schedule {C1 , . . . , C y } for time tick
{t1 , . . . , t y } such that for all ticks,
1. each target is covered  y by at least one sensor node
2. network lifetime p=1 t p is maximized.
IEEE SENSORS JOURNAL, VOL. 14, NO. 3, MARCH 2014
2) k-Coverage Scheduling: Given a set of sensor nodes S =
{S1 , S2 , . . . , Sm } with battery power B = {b1 , b2 , . . . , bm },
energy consumption rate ei for Si and a target set
T = {T1 , T2 , . . . , Tn }, generate a schedule {C1 , . . . , C y }, for
{t1 , . . . , t y }, such that for all ticks,
1. each target is covered  y by at least k sensor nodes, 1 ≤ k ≤ m
2. network lifetime p=1 t p is maximized.
3) Q-Coverage Scheduling: Given a set of sensor nodes S =
{S1 , S2 , . . . , Sm } with battery power B = {b1 , b2 , . . . , bm },
energy consumption rate ei for Si and a target set
T = {T1 , T2 , . . . , Tn }, generate a schedule {C1 , . . . , C y }, for
{t1 , . . . , t y }, such that for all ticks,
1. T = {T1 , T2 , . . . , Tn } is covered by at least Q =
{q1 , q2 , . . . , qn } sensor nodes, where each target T j , 1 ≤ j ≤
n, is covered by at least
2. network lifetime p=1 t p is maximized.
IV. P ROPOSED M ETHOD
the optimal deployment locations and the schedule are decided
at the base station, prior to actual deployment. The pro-
posed method has two phases: sensor deployment and sensor
scheduling. The nodes are initially deployed randomly. Based
on the theoretical upper bound of network lifetime, we com-
A heuristic is then used to schedule the sensor nodes such that
the network lifetime is maximum. Algorithm 1 describes the
proposed method.
Algorithm 1 Proposed Approach
1: Input: S, T
2: Output: Optimal location of S and sensor schedule
3: Deploy S randomly
4: Compute upper bound of network lifetime using (2)
5: Recompute sensor node positions using ABC algorithm such that the upper bound
of network lifetime is maximum
6: Design sensor schedule using the proposed heuristic for sensor scheduling such that
the network lifetime upper bound is achieved
A. Sensor Deployment
Since the upper bound of network lifetime can be computed,
we have to find the deployment locations such that the network
lifetime is maximum. First we use a heuristic to compute
the deployment locations and then we use ABC and PSO
algorithms to compute the locations.
1) A Heuristic for Sensor Deployment: Here we propose
a heuristic for sensor deployment (Algorithm 2). Initially,
place the sensor nodes randomly. If any sensor node is idle
(without monitoring any target), the node is moved to the least
monitored targets’ location. This is to ensure that all sensor
nodes play their part in monitoring the targets. The sensor
nodes are then sorted based on the number of targets it cover.
The sensor node is placed at the middle of all the targets
it cover. The next nearest target is identified and the sensor
node is placed at the middle of all these targets. If it can cover
this new target along with targets it was already monitoring,
allow this move, else discard the move. This is done till the
sensor node cannot cover any new target. At the end, upper
MINI et al.: SENSOR DEPLOYMENT AND SCHEDULING FOR TARGET COVERAGE PROBLEM IN WSNs 639

bound is computed. The drawback of this approach is that it probability pi defined as,
depends on the initial position of the sensor nodes. Though it F(Di )
may perform well for dense deployments, consistency cannot pi = m (3)
l=1 F(Dl )
always be guaranteed.
where m is the total number of food sources. The onlooker
finds a neighborhood food source in the vicinity of Di as,
Algorithm 2 A heuristic for Sensor Deployment
1: Place sensor nodes randomly
for i = 1 to m do
2:
3: if Si does not monitor any target then
Di (t + 1) = Di (t) + δi j × f (4)
4: Move Si to the least monitored target
5: Recompute sensor-target coverage matrix where δi j is the neighborhood patch size for j t h dimension of
6: end if i t h food source, and f is a random uniform variate ∈ [−1, 1].
7: end for
8: S = Sensor nodes sorted in ascending order of number of targets it covers It should be noted that the solutions are not allowed to move
9: for i = 1 to m do beyond the edge of the search region. The new solutions are
10: repeat
11: Place Si at the center of all targets it covers evaluated using the fitness function (2). If any new solution
12: Move Si to the center of all targets it covers and its next nearest target is better than the existing one, the old solution is replaced
13: if Si can cover a new target then
14: Recompute sensor-target matrix with new solution. Scout bees search for a random feasible
15: else solution. The solution with the least sensing range is finally
16: Discard move
17: end if chosen as the best solution.
18: until Si can cover another target
19: end for
20: Compute upper bound of network lifetime using (2) Algorithm 3 ABC Algorithm
1: Initialize the solution population B
2: Evaluate fitness
2) ABC Based Sensor Deployment: Artificial Bee Colony 3: cycle = 1
4: repeat
(ABC) Algorithm [6], [7] is an optimization algorithm based 5: Search for new solutions in the neighborhood
on the intelligent behavior of honey bee swarm. The colony 6: if new solution is better than old solution then
7: Memorize new solution and discard old solution
of bees contains three groups: employed bees, onlookers and 8: end if
scouts.The employed bee takes a load of nectar from the source 9: Replace the discarded solution with a new randomly generated solution
10: Memorize the best solution
and returns to the hive and unloads the nectar to a food store. 11: cycle = cycle + 1
After unloading the food, the bee performs a special form of 12: until cycle = maximumcycles
dance called waggle dance [38] which contains information
about the direction in which the food will be found, its
distance from the hive and its quality rating. Since information
3) PSO Based Sensor Deployment: Particle Swarm Opti-
about all the current rich sources is available to an onlooker
mization (PSO) consists of a swarm of particles moving in
on the dance floor, an onlooker bee probably could watch
a search space of possible solutions for a problem. Every
numerous dances and choose to employ itself at the most
particle has a position vector representing a candidate solution
qualitative source. There is a greater probability of onlookers
to the problem and a velocity vector. Moreover, each particle
choosing more qualitative sources since more information
contains a small memory that stores its own best position
is circulating about the more qualitative sources. Employed
seen so far and a global best position obtained through
foragers share their information with a probability, which is
communication with its neighbor particles [40], [41].
proportional to the quality of the food source. Hence, the
It consists of a swarm of w candidate solutions called
recruitment is proportional to quality of a food source [7], [39].
particles, which explore an nd-dimensional hyperspace in
Exploitation is carried out by employed bees and onlookers,
search of the global solution (n represents the number of
while exploration is carried out by scouts.
optimal parameters to be determined). A particle p occupies
Algorithm 3 describes ABC algorithm. Let the solution
position x pd and velocity v pd in the d t h dimension of the
population be B. The region is assumed to have only stationary
hyperspace, 1 ≤ p ≤ w and 1 ≤ d ≤ nd.
targets. Each solution Ba = {(x 1 , y1 ), (x 2 , y2 ), . . . , (x m , ym )}
In the global-best version of PSO, the position where the
where a = 1, 2, . . . , nb, where nb and m represents total
particle p has its best cost is stored as ( pbest pd ). Besides,
number of bees and total number of nodes respectively,
gbestd , the position of the best particle. In each iteration tr ,
corresponds to a bee. The initial solution is generated in such
velocity v and position x are updated using (5) and (6). The
a way that all the targets are covered, and each sensor node
update process is iteratively repeated until either an acceptable
covers at least one target. The network lifetime is computed
gbest is achieved or a fixed number of iterations trmax is
for each solution using (2).
reached.
This network lifetime is used as the fitness function for
evaluating the solutions. Each sensor node is associated with v pd (tr + 1) = w.v pd (tr ) + ϕ1 .r1 (tr ). pbest pd − x pd
a cluster, where a cluster corresponds to the set of targets + ϕ2 .r2 (tr ).(gbestd − x pd ) (5)
monitored by the sensor node. Let Di = (X i , Yi ) be the initial
x pd (tr + 1) = x pd (tr ) + v pd (tr + 1). (6)
position of i t h cluster. F(Di ) refers to the nectar amount at
food source located at Di . After watching the waggle dance Here, ϕ1 and ϕ2 are constants, and r1 (tr ) and r2 (tr ) are
of employed bees, an onlooker goes to the region of Di with random numbers uniformly distributed in [0,1].
640 IEEE SENSORS JOURNAL, VOL. 14, NO. 3, MARCH 2014

The basic steps of PSO [41] are given in Algorithm 4. Nodes with different coverage degree may coexist in a
network. Though the initial battery power of all the nodes
Algorithm 4 PSO Algorithm in the network might be the same, subsequently it may vary
1: Initialize particles in accordance with the cover activation.
2: repeat The weights are recalculated for all the nodes at each time
3: for each particle do
4: Calculate the fitness value instant if,
5: if fitness value is better than the best fitness value ( pbest ) in history then 1. Weight due to the remaining energy changes: It happens
6: Set current value as the new pbest
7: end if due to reduction in battery power for nodes which were in the
8: end for previous cover.
9: Choose the particle with the best fitness value of all the particles as the gbest
10: for each particle do 2. Node turns off due to no battery power: If a sensor node
11: Calculate particle velocity according to velocity update equation (5) that monitors a target turns off, it will reassign weights to all
12: Update particle position according to position update equation (6)
13: end for other sensor nodes monitoring it.
14: until maximum iterations or minimum error criteria is attained This weight recalculation might trigger a priority change
and subsequently a new cover might be generated at the next
time instant.
B. A Heuristic for Sensor Scheduling b) Weight due to covered targets: All sensor nodes are
assigned weights based on the targets it cover as:
As mentioned earlier, another objective of this paper is
to schedule the sensor nodes such that the theoretical upper 
n
Mi j
bound of network lifetime can be achieved. wi = m (7)
j =1 i=1 Mi j
To achieve this, we propose a weight-based method for
determining the cover sets. It includes the following main The proposed heuristic (Algorithm 5) initially finds the
steps: network lifetime using weight assigned by the nodes itself
1) Weight assignment (battery power). If the obtained network lifetime does not
2) Cover formation match the theoretical upper bound of network lifetime, the
3) Cover optimization weight due to covered targets is considered to compute the
4) Cover activation and Energy reduction. lifetime.
2) Cover Formation: A cover can be generated in different
ways if the network has nodes which make all the targets k/Q
Algorithm 5 Heuristic for Sensor Node Scheduling
covered. The proposed approach uses a priority based method
1: Input M, B
2: Initialize k/Q, max _run , priority calculated using battery power (Algorithm 6). In the order of priority, if any new sensor node
3: for r = 1 to max _run do  contributes to k/Q coverage requirement, it will be added to
4: for iteration = 1 to m i=1 bi do
5: if cover possibility exists then the cover set. In general, a sensor node Si can be added to a
6: Determine cover based on priority cover set Cov_S if and only if
7: Optimize cover
8: Activate optimized cover and reduce battery power 1) for simple coverage problem: Cov_S ∪ {Si } covers any
9: else
10: break new target
11: end if 2) for k-coverage problem: Cov_S ∪ {Si } contributes to
12: end for
13: Calculate network lifetime (nlife) k-coverage requirement
14: if nlife < U then 3) for Q-coverage problem: Cov_S ∪ {Si } contributes to
15: Consider weight due to covered targets to compute priority to check for better
lifetime Q-coverage requirement.
16: else
17: break
18: end if Algorithm 6 Cover Formation
19: end for
1: Input: Sorted S in descending order of assigned weight
2: Output: Cov _ S
3: Initialize Cov _ S = φ
4: for i = 1 to m do
1) Weight Assignment: Weight assignment is performed to 5: if Si contributes to coverage then
6: Cov _ S = Cov _ S ∪ {Si }
decide the priority of sensor nodes. The more the weight of 7: end if
a sensor node, the higher the priority. Cover sets are decided 8: if coverage requirement met then
9: break ;
based on this priority. 10: end if
The base station calculates weight for each sensor node by 11: end for
considering two factors. The initial weight deciding factor is
weight due to remaining energy and the next one is weight 3) Cover Optimization: Once the coverage requirement
due to targets that the node covers. is met, the obtained cover set is optimized (Algorithm 7).
a) Weight assigned by the node itself: Each sensor node By optimizing the generated cover, the proposed scheme
assigns a weight to itself which is equivalent to the remaining attempts to minimize the energy usage. It should be noted
battery power of the sensor node. For each node Si in the that this is the second phase of redundancy elimination, the
optimized cover set, the weight assigned by itself decrements first one being at the cover formation. A problem that arises
by the rate of energy consumption. with the cover formed at the cover formation phase is that it
MINI et al.: SENSOR DEPLOYMENT AND SCHEDULING FOR TARGET COVERAGE PROBLEM IN WSNs 641

might still have nodes that need not be turned on to cover all
the targets. This is possible because it is a step by step addition
till all the targets are covered. A node can thus be dropped
for not contributing to coverage at the time of cover formation
or for not contributing to coverage after cover formation. The
nodes in the cover set are subject to optimization using least
priority first approach. This method of elimination prevents
the higher priority nodes being discarded at the initial stages
of optimization itself. The least priority node in the cover set
cannot be eliminated from the cover set as it satisfies the k/Q
coverage requirement. Elimination starts from the last but one
node as per increasing priority. A node Si ∈ Cov_S, 1 ≤ i ≤
length(Cov_S), represented as Si .Cov_S will not be added to Fig. 1. Network lifetime for simple coverage problem using random
the optimized cover set Opt.Cov_S if Cov_S − {Si .Cov_S} deployment, proposed heuristic and ABC algorithm.
meets k/Q coverage requirement.

Algorithm 7 Cover Optimization

1: Input: Cov _ S
2: Output: Opt.Cov _ S
3: Initialize Opt.Cov _ S = φ
4: for i = length(Cov _ S) down to 1 do
5: if Cov _ S − {Si .Cov _ S} meets k /Q coverage requirement then
6: I gnore Si .Cov _ S
7: Cov _ S = Cov _ S − {Si .Cov _ S}
8: else
9: Opt.Cov _ S = Opt.Cov _ S ∪ {Si .Cov _ S}
10: end if
11: end for

4) Cover Activation and Energy Reduction: The sensor

nodes in the optimized cover are activated. The total energy
that each node consumes should not fall beyond the minimum
usable energy, E min . When the battery power reaches E min ,
the node becomes inactive and will not be able to monitor any
more targets further. As the battery power is drained when a
node is turned on, the weight assigned by the node to itself
reduces. The network terminates when no cover can further
be formed (Algorithm 8).
As we assume that the number of sensors deployed in the
area is greater than the optimum number required to monitor
the targets, determining the sensor covers and switching from
one cover to another in a scheduled manner such that only
minimum number of sensor nodes remain active at any time
instant is supposed to improve network lifetime.
Algorithm 8 Cover Activation and Energy Reduction
1: Input: Opt.Cov _ S
2: for i = 1 to length(Opt.Cov _ S) do
3: Si .state = true
4: decrement bi
5: if bi ≤ E min then
6: for j = 1 to n do
7: Mi j = 0
8: end for
9: end if
10: end for
V. R ESULTS AND D ISCUSSION
We consider a 500 m × 500 m region for experiments. The
number of targets is 25. The number of sensor nodes (m)
is varied from 100 to 250. Sensing range of each sensor
node is fixed as 75 m. Results are reported as an average of
Fig. 2. Network lifetime for k coverage problem using random deployment,
proposed heuristic and ABC algorithm.
5 instances in each configuration. Initially, each sensor node
has 100 units of battery power. Energy consumption rate is
1 unit. Simulations are carried out using MatLab 2007a.
A. Sensor Deployment
1) Random Deployment: In random deployment, there is
more chance of targets being not detected or targets not being
covered with the required level of coverage. However, this
may not hold true with dense deployment of nodes. But there
is another possibility of some targets being monitored by
many sensor nodes, and some by a few sensor nodes. This
difference in the number of sensor nodes monitoring each
target will affect the network lifetime. The sensor nodes may
be positioned in a better way so as to avoid this variation.
This will yield better lifetime. Though random deployment
has these drawbacks, there are applications where random
deployment is the only feasible strategy.
Fig. 1, Fig. 2 and Fig. 3 show the network lifetime for
various deployment methods. Random deployment does not
yield as much network lifetime as other methods do. Fig. 1
shows that for simple coverage problem, random deployment
survives when 100, 150, 200 and 250 nodes are deployed.
But Fig. 2 shows that when 100 sensor nodes are to be
randomly deployed, k = 2 will be satisfied, but k = 3 and
k = 4 will not be satisfied, ultimately giving zero network
lifetime. Fig. 3 shows that for 100 sensor nodes and Q ranging
from 1 to 5, random deployment does not work. This clearly
642
Fig. 3. Network lifetime for Q-coverage problem using random deployment,
proposed heuristic and ABC algorithm.
shows that when the number of nodes to be deployed is high,
random deployment can be done, but when the number of
sensor nodes is not sufficient enough, random deployment
fails.
2) Heuristic: The heuristic could consistently achieve better
results compared to random deployment. Fig. 1, Fig. 2, and
Fig. 3 clearly show it. Fig. 2 shows that when 100 sensor nodes
are randomly deployed, k = 3 condition cannot be satisfied.
But with the heuristic, the network can be active for some
time. The same is observed for k = 4. Random deployment
yields zero network lifetime, but the heuristic can make the
network survive for more time.
When the number of nodes is increased to 150, k = 3 and
k = 4 requirement is possible even with random deployment.
In these cases, the heuristic can achieve much higher network
lifetime. Fig. 3 shows that the heuristic is able to satisfy Q
ranging from 1 to 5 with 100 sensor nodes deployed, whereas
random deployment fails. The results confirm that the heuristic
performs better compared to random deployment.
3) ABC Based Deployment: Though deterministic deploy-
ment may be time consuming, it helps in improving network
lifetime. It might require only few nodes to be turned on at a
time to satisfy coverage requirement, as compared to random
deployment. In the experiments, we have taken the number
of bees as 10, number of cycles is 5000 and the number
of runs is 5. The limit for neighborhood search is set as
30. Deployment using ABC algorithm could achieve much
higher network lifetime compared to random deployment and
the proposed heuristic. Irrespective of the number of sensor
nodes, coverage requirement etc., this method consistently
achieves better network lifetime. Fig. 1, Fig. 2 and Fig. 3
depict it.
4) PSO Based Deployment: Table 1 shows a comparison of
network lifetime computed using ABC and PSO based sensor
deployment. The setting of PSO is as follows: number of
particles is 10, number of iterations is 5000 and number of
runs is 5. The number of sensors is varied from 100 to 250.
For k-coverage problem, with 100 or 150 sensor nodes to be
deployed, the network lifetime computed using ABC and PSO
deployments are the same when k = 1 to 4. When 200 sensor
nodes are to be deployed, ABC performs better than PSO for
k = 1, 2 and 4. But in this case, ABC and PSO attains the
same lifetime when k = 3. ABC consistently achieves better
IEEE SENSORS JOURNAL, VOL. 14, NO. 3, MARCH 2014
Fig. 4. Network lifetime for simple coverage problem after scheduling.
network lifetime than PSO when the number of sensor nodes is
250. It could outperform PSO for all the experimented values
of k.
ABC performs better in a majority of the experiments of
Q-coverage problem. PSO could achieve a network lifetime
same as that achieved by ABC only for Q = [1, 3] when the
number of sensor nodes is 150 and 200. In all the other cases,
ABC outperforms PSO.
5) Other Observations:
a) Varying number of nodes: The number of sensor nodes
to be deployed in the region is varied from 100 to 250. The
network lifetime increases when more number of nodes are to
be deployed.
b) Varying coverage requirement: An increase in network
lifetime is observed but it is evident that the network lifetime
does not increase in proportion to the increase in k. The same
is observed for Q-coverage requirement.
B. Sensor Scheduling
Since the optimal deployment locations are known, now
the sensor nodes have to be scheduled such that each sensor
node need not be awake all the time. Theoretical upper bound
helps to analyze how far the proposed heuristic is successful.
With battery power as the priority deciding factor, though most
instances could obtain network lifetime equal to upper bound,
there are few instances where the network lifetime does not
match the computed upper bound. For these instances, the
second weight deciding factor is considered. An inverse of
this helps to achieve theoretical upper bound. Thus we observe
that there are few cases where if the sensors monitoring more
number of targets are kept in reserve for later use, theoretical
and experimental network lifetime matches.
Fig. 4 shows a comparison of upper bound of network
lifetime and the network lifetime obtained using proposed
approach for simple coverage problem. The proposed approach
could achieve the theoretical upper bound in all the instances.
Fig. 5 and Fig. 6 show the comparison of upper bound
and the network lifetime obtained using proposed approach
for k-coverage and Q-coverage problems respectively. The
network lifetime obtained using proposed approach matches
the theoretical upper bound for all these requirements as
well. Fig. 7 shows a comparison of the proposed method and
Greedy-MSC [42]. Greedy-MSC was proposed to solve simple
MINI et al.: SENSOR DEPLOYMENT AND SCHEDULING FOR TARGET COVERAGE PROBLEM IN WSNs
TABLE I
C OMPARISON OF N ETWORK L IFETIME C OMPUTED U SING ABC AND PSO BASED S ENSOR D EPLOYMENT
Fig. 5. Network lifetime for k-coverage problem after scheduling.
Fig. 6. Network lifetime for Q-coverage problem after scheduling.
Fig. 7. Comparison of proposed approach and G-MSC.
coverage problem. In Greedy-MSC, a critical target (the target
most sparsely covered) is selected initially. Once the critical
target has been selected, the heuristic selects the sensor with
the greatest contribution that covers the critical target. Once a
sensor has been selected, it is added to the current set cover,
and all additionally covered targets are removed from the set
643
of targets to be covered. When all targets are covered, the
new set cover is formed. Our proposed method could achieve
better network lifetime for simple coverage (k = 1) and higher
values of k as shown in Fig. 7.
Computing positions using ABC algorithm outperformed
random deployment, heuristic and PSO. The heuristic pro-
posed to schedule the sensor nodes could meet the theoretical
upper bound in all experimented cases.
VI. C ONCLUSION
In this paper, we compute deployment locations for sensor
nodes using artificial bee colony algorithm such that the
network lifetime is maximum. Artificial bee colony algorithm
performs better than PSO algorithm for this problem. In order
to avoid the battery drain of all nodes at a time, sensor
node scheduling can be done so that only minimum number
of sensor nodes required for satisfying coverage requirement
needs to be turned on. The other nodes can be reserved for
later use. This method helps to prolong the network lifetime.
We use a heuristic which is powerful enough to schedule the
sensor nodes in such a way that the network lifetime matches
the theoretical upper bound of network lifetime. Network
lifetime is extended by using this method of deploying at
optimal locations such that it achieves maximum theoretical
upper bound and then scheduling them so as to achieve the
theoretical upper bound. For future work, we plan to extend
this method of deployment and scheduling for probabilistic
coverage in wireless sensor networks.
R EFERENCES
[1] C.-Y. Chong and S. Kumar, “Sensor networks: Evolution, opportunities,
and challenges,” Proc. IEEE, vol. 91, no. 8, pp. 1247–1256, Aug. 2003.
[2] J. Wang, R. Ghosh, and S. Das, “A survey on sensor localization,” J.
Control Theory Appl., vol. 8, no. 1, pp. 2–11, 2010.
[3] C.-F. Huang and Y.-C. Tseng, “The coverage problem in a wireless
sensor network,” in Proc. 2nd ACM Int. Conf. Wireless Sensor Netw.
Appl., 2003, pp. 115–121.
[4] Y. Gu, H. Liu, and B. Zhao, “Target coverage with QoS requirements
in wireless sensor networks,” in Proc. Intell. Pervas. Comput., 2007,
pp. 35–38.
[5] M. Chaudhary and A. K. Pujari, “Q-coverage problem in wireless
sensor networks,” in Proc. Int. Conf. Distrib. Comput. Netw., 2009,
pp. 325–330.
[6] D. Karaboga and B. Akay, “A survey: Algorithms simu-
lating bee swarm intelligence,” Artif. Intell. Rev., vol. 31,
nos. 1–4, pp. 61–85, 2009.
[7] D. Karaboga and B. Basturk, “On the performance of artificial bee
colony (ABC) algorithm,” Appl. Soft Comput., vol. 8, pp. 687–697, Jan.
2008.
[8] D. Karaboga, B. Gorkemli, C. Ozturk, and N. Karaboga, “A compre-
hensive survey: Artificial bee colony (ABC) algorithm and applications,”
Artif. Intell. Rev., 2012, pp. 1–37.
644
[9] D. Karaboga, C. Ozturk, N. Karaboga, and B. Gorkemli, “Artificial
bee colony programming for symbolic regression,” Inf. Sci., vol. 209,
pp. 1–15, Nov. 2012.
[10] D. Karaboga, S. Okdem, and C. Ozturk, “Cluster based wireless sensor
network routing using artificial bee colony algorithm,” Wireless Netw.,
vol. 18, no. 7, pp. 847–860, 2012.
[11] J. Yick, B. Mukherjee, and D. Ghosal, “Wireless sensor network survey,”
Comput. Netw., vol. 52, pp. 2292–2330, Aug. 2008.
[12] Y. Li and S. Gao, “Designing k-coverage schedules in wireless sensor
networks,” J. Combinat. Opt., vol. 15, no. 2, pp. 127–146, 2008.
[13] A. Krause, R. Rajagopal, A. Gupta, and C. Guestrin, “Simultaneous
placement and scheduling of sensors,” in Proc. Int. Conf. IPSN, 2009,
pp. 181–192.
[14] K. Wu, Y. Gao, F. Li, and Y. Xiao, “Lightweight deployment-aware
scheduling for wireless sensor networks,” Mobile Netw. Appl., vol. 10,
pp. 837–852, Dec. 2005.
[15] J. Wu and S. Yang, “Optimal movement-assisted sensor deployment and
its extensions in wireless sensor networks,” Simul. Model. Pract. Theory,
vol. 15, no. 4, pp. 383–399, 2007.
[16] N. Bartolini, T. Calamoneri, T. La Porta, and S. Silvestri, “Mobile sensor
deployment in unknown fields,” in Proc. IEEE INFOCOM, Mar. 2010,
pp. 1–5.
[17] C.-Y. Chang, C.-T. Chang, Y.-C. Chen, and H.-R. Chang, “Obstacle-
resistant deployment algorithms for wireless sensor networks,” IEEE
Trans. Veh. Technol., vol. 58, no. 6, pp. 2925–2941, Jul. 2009.
[18] T.-L. Chin, “Sensor deployment for collaborative target detection in
the presence of obstacles,” in Proc. IEEE Global Telecommun. Conf.,
Dec. 2009, pp. 1–5.
[19] E. Onur, C. Ersoy, and H. Deliç, “Quality of deployment in surveillance
wireless sensor networks,” Int. J. Wireless Inform. Netw., vol. 12, no. 1,
pp. 61–67, 2005.
[20] X. Bai, S. Kumar, D. Xuan, Z. Yun, and T. H. Lai, “Deploying wireless
sensors to achieve both coverage and connectivity,” in Proc. 7th ACM
Int. Symp. Mobile Ad Hoc Netw. Comput., 2006, pp. 131–142.
[21] Z. Yun, X. Bai, D. Xuan, T. Lai, and W. Jia, “Optimal deployment
patterns for full coverage and k-connectivity (k≤6) wireless sensor
networks,” IEEE/ACM Trans. Netw., vol. 18, no. 3, pp. 934–947,
Jun. 2010.
[22] X. Bai, Z. Yun, D. Xuan, T. Lai, and W. Jia, “Optimal patterns for
four-connectivity and full coverage in wireless sensor networks,” IEEE
Trans. Mobile Comput., vol. 9, no. 3, pp. 435–448, Mar. 2010.
[23] G. Tan, S. Jarvis, and A.-M. Kermarrec, “Connectivity-guaranteed and
obstacle-adaptive deployment schemes for mobile sensor networks,” in
Proc. 28th Int. Conf. Distrib. Comput. Syst., Jun. 2008, pp. 429–437.
[24] Z. Cheng, M. Perillo, and W. Heinzelman, “General network lifetime
and cost models for evaluating sensor network deployment strategies,”
IEEE Trans. Mobile Comput., vol. 7, no. 4, pp. 484–497, Apr. 2008.
[25] C. Ozturk, D. Karaboga, and B. Gorkemli, “Probabilistic dynamic
deployment of wireless sensor networks by artificial bee colony algo-
rithm,” Sensors, vol. 11, no. 6, pp. 6056–6065, 2011.
[26] C. Liu, K. Wu, and V. King, “Randomized coverage-preserving schedul-
ing schemes for wireless sensor networks,” in Proc. Netw., 2005,
pp. 956–967.
[27] L.-H. Yen and Y.-M. Cheng, “Range-based sleep scheduling (RBSS)
for wireless sensor networks,” Wireless Pers. Commun., vol. 48,
pp. 411–423, Feb. 2009.
[28] A. Keshavarzian, H. Lee, and L. Venkatraman, “Wakeup scheduling in
wireless sensor networks,” in Proc. 7th ACM Int. Symp. Mobile Ad Hoc
Netw. Comput., 2006, pp. 322–333.
IEEE SENSORS JOURNAL, VOL. 14, NO. 3, MARCH 2014
[29] A. Makhoul and C. Pham, “Dynamic scheduling of cover-sets
in randomly deployed wireless video sensor networks for surveil-
lance applications,” in Proc. 2nd IFIP Conf. Wireless Days, 2009,
pp. 73–78.
[30] C.-Y. Chang and H.-R. Chang, “Energy-aware node placement, topology
control and MAC scheduling for wireless sensor networks,” Comput.
Netw., vol. 52, pp. 2189–2204, Aug. 2008.
[31] S. Tilak, N. B. Abu-Ghazaleh, and W. Heinzelman, “A taxonomy of
wireless micro-sensor network models,” ACM Mobile Comput. Commun.
Rev., vol. 6, no. 2, pp. 28–36, 2002.
[32] S. Mini, S. K. Udgata, and S. L. Sabat, “Sensor deployment in 3-D
terrain using artificial bee colony algorithm,” in Proc. Swarm, Evol.
Memetic Comput., 2010, pp. 424–431.
[33] C. Ozturk, D. Karaboga, and B. Gorkemli, “Artificial bee colony
algorithm for dynamic deployment of wireless sensor networks,” Turkish
J. Electr. Eng. Comput. Sci., vol. 20, no. 2, pp. 255–262, 2012.
[34] S. K. Udgata, S. L. Sabat, and S. Mini, “Sensor deployment in irregular
terrain using artificial bee colony algorithm,” in Proc. World Congr.
Nature Biol. Inspired Comput., Dec. 2009, pp. 1309–1314.
[35] S. Mini, S. K. Udgata, and S. L. Sabat, “Artificial bee colony based
sensor deployment algorithm for target coverage problem in 3-D terrain,”
in Proc. Distrib. Comput. Internet Technol., 2011, pp. 313–324.
[36] Y. Gu, B.-H. Zhao, Y.-S. Ji, and J. Li, “Theoretical treatment of target
coverage in wireless sensor networks,” J. Comput. Sci. Technol., vol. 26,
no. 1, pp. 117–129, 2011.
[37] S. Mini, S. K. Udgata, and S. L. Sabat, “A heuristic to maximize network
lifetime for target coverage problem in wireless sensor networks,” Ad
Hoc Sensor Wireless Netw., vol. 13, nos. 3–4, pp. 251–269, 2011.
[38] J. Riley, U. Greggers, A. Smith, D. R. Reynolds, and R. Menzel,
“The flight paths of honeybees recruited by the waggle dance,” Nature,
vol. 435, pp. 205–207, Mar. 2005.
[39] V. Tereshko and A. Loengarov, “Collective decision-making in honey
bee foraging dynamics,” Comput. Inform. Syst., vol. 9, no. 3, pp. 1–7,
2005.
[40] J. Kennedy and R. Eberhart, “Particle swarm optimization,” in Proc.
IEEE Int. Conf. Neural Netw., vol. 4, Nov./Dec. 1995, pp. 1942–1948.
[41] R. Eberhart and J. Kennedy, “A new optimizer using particle swarm
theory,” in Proc. 6th Int. Symp. Micro Mach. Human Sci., Oct. 1995,
pp. 39–43.
[42] M. Cardei, M. T. Thai, Y. Li, and W. Wu, “Energy-efficient target
coverage in wireless sensor networks,” in Proc. 24th Annu. Joint Conf.
IEEE INFOCOM, Mar. 2005, pp. 1976–1984.
S. Mini, photograph and biography are not available at the time of publication.
Siba K. Udgata, photograph and biography are not available at the time of
publication.
Samrat L. Sabat, photograph and biography are not available at the time of
publication.