Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: This paper presents a distributed range-free node localization algorithm for three-
Received 22 February 2017 dimensional (3D) wireless sensor networks (WSNs), named as 3D-GAIDV Hop (3D genetic
Revised 15 December 2017
algorithm based Improved Distance Vector Hop). In the proposed algorithm, the average
Accepted 20 December 2017
hop size of anchor nodes is modified by updating a correction factor and the modified
Available online xxx
hop size is further optimized by line search algorithm. The concept of coplanarity is in-
Keywords: troduced to reduce location errors caused by the anchor nodes which are coplanar. The
DV-Hop algorithm localization accuracy is further improved by applying genetic algorithm (GA). To improve
Genetic algorithm the positioning coverage of the network, those target nodes which have been localized
Localization successfully in the first round of the localization process, are upgraded to assistant an-
Optimization chor node. All the processes and calculations are carried out at target node level, which
Wireless sensor networks makes the proposed algorithm energy efficient. Simulation results show that our proposed
algorithm outperforms the similar kind of existing algorithms.
© 2017 Elsevier Ltd. All rights reserved.
1. Introduction
Due to recent advancement in microelectronics and wireless communication technologies, most of the industrial interests
and research have been attracted towards wireless sensor networks (WSNs). These technologies lead to the development of
inexpensive, extremely small, multi-functional, and smart Sensor Nodes (SNs) which have the capabilities to communicate
with each other through a wireless medium [1]. These tiny SNs comprise of processor, dedicated memory, transceiver, ac-
tuator, sensor(s), and power module [2]. Due to size and cost constraint, SNs have limited power and low computational
capabilities. Depending upon the applications, a collection of hundreds or thousands of nodes may be deployed in the area
of interest, which are capable of communicating with each other through a wireless medium and form a network called
wireless sensor network (WSN).
SNs can sense the desired physical phenomena i.e. temperature, pressure, light, humidity, etc. If an event occurs in the
area of interest, SNs present within the vicinity of the event sense it and disseminate the sensed data through multihop
transmission to a centralized unit known as sink node or base station. Base station acts as gateway between WSN and
wired infrastructure which provides a connection to the user end where the sensed data can be collected, processed, and
analysed. A typical WSN is shown in Fig. 1.
R
Reviews processed and recommended for publication to the Editor-in-Chief by Associate Editor Dr. M. H. Rehmani.
∗
Corresponding author.
E-mail addresses: gaurav.sharma@cvr.ac.in (G. Sharma), ashok@nith.ac.in (A. Kumar).
https://doi.org/10.1016/j.compeleceng.2017.12.036
0045-7906/© 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
WSN has a number of prospective applications in various areas such as surveillance, habitat and environmental moni-
toring, military applications, healthcare, structural monitoring, disaster management, etc. [2]. A number of such applications
require accurate location of the sensor nodes to make sensed data meaningful. Some of these applications are forest fire
detection, battlefield surveillance, search and rescue operations, target tracking, etc. [1,2].
Fig. 2 shows the need of localization in WSN, in which SNs sense the event occurred in the sensing field and disseminate
the sensed information to the user through base station and internet/satellite. At this point, user definitely came to know
what event has happened in the sensing field but the first thing in the user’s mind would come –WHERE? Because if the
location of the event is not associated with the sensed data, the user can’t take the prompt action. Since the users need to
know not only what happens but also where the interested event happens. Hence, node localization has become one of the
critical issues in WSNs. Sometimes, it is necessary to take quick actions i.e. in hospitals, when patient is in critical condition.
In such cases, sensed data should be reported to the central coordinator without any traffic delay [3].
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Manual deployment of nodes is one of the easiest methods for location estimation but it is not feasible for a large scale
network particularly when nodes are deployed in inaccessible areas. Addition of Global Positioning System (GPS) [4] receiver
to the node is also another possible way for node localization. But unfortunately, it is not viable for such a large sensor
network as it adds up to the size, power consumption, and cost of network deployment [5]. One possible solution is to
equip only small number of nodes with GPS, known as anchor nodes (ANs) or beacon nodes and localize remaining nodes
known as target nodes with the help of these anchor nodes.
Localization algorithms can be broadly categorized as: range-based and range-free localization.
Range-based localization algorithms need to calculate the distances between ANs and SNs based on the actual range or
angle information and use this information to determine the coordinates of target nodes [6]. Several ranging techniques
exist to estimate the target node distance from three or more anchor nodes. After obtaining ranging results, the position
of target nodes can be estimated through multilateration or triangulation techniques. Though, these schemes show good
localization accuracy but, need an extra hardware for actual range determination. Moreover, these schemes are also easily
affected by the multipath fading [2]. Some popular range-based localization schemes are received signal strength indicator
(RSSI), Angle of Arrival (AoA), Time of Flight (ToF) methods, etc.
ToF and AoA based methods can achieve much better accuracy than RSSI. However, all these methods require additional
hardware. The ToF based methods need high-speed clocks to support ultra-high resolution timing, while the AoA based
methods demand antenna array to effectively detect the angles. Furthermore, rigid synchronization is necessary for simple
ToF methods such as the direct distance calculation method by ToF. On the other hand, the AoA information is sensitive to
multipath fading and noise. Therefore, range-based localization algorithms are not viable for large scale WSN.
In contrast, range-free localization algorithms use connectivity information and hop count values between ANs and SNs
to estimate the position of target nodes. However, these methods suffer from poor localization accuracy. Due to some lim-
itations in range-based schemes, range-free schemes have attracted worldwide researchers’ interest because these are cost
effective alternate methods for location calculation without any extra range determination hardware. A number of range
free localization algorithms are available in literature [5–12].
In literature, localization approaches focus mainly on 2D (two-dimensional) plane but in real practice, nodes are often
deployed in three-dimensional (3D) space in WSNs e.g. in forests, in buildings, on mountains, in deep water, etc. In [13],
a comprehensive survey of WSN deployment in urban areas has been provided, from which it has been noticed that in
almost all practical applications, nodes are deployed in 3D space. Thus the localization in 3D scenario is more popular and
will lead to further improvement of localization technology. The 3D node localization in WSNs is much more complex and
complicated in terms of computations. Thus, research on localization of nodes for 3D space is necessary and more realistic.
Distance Vector Hop (DV-Hop) [6] is one of widely used range-free localization algorithm. However, DV-Hop algorithm
suffers from poor localization accuracy. In this paper, we contribute to improve localization accuracy of 3D-DV Hop algo-
rithm.
The main contributions of the present work are delineated as follows:
• We propose a three-dimensional genetic algorithm based Improved Distance Vector Hop localization algorithm (3D-
GAIDV Hop), which is a range-free distributed cooperative scheme.
• Most of the works reported in literature have not considered the poor distance estimation between nodes if number
of hops between anchor node and target node are more than two. To address such deviation in distance estimation, a
correction factor is proposed to refine the hop size of anchor nodes.
• Line Search Algorithm (LSA) has been used to find the optimum hop size of each anchor node, due to which correct
distance estimation is obtained between nodes.
• The concept of coplanarity is introduced in the present work to reduce the localization errors caused by the anchor
nodes which are coplanar.
• To improve the positioning coverage of the network, concept of promotion of target nodes to assistant anchor nodes has
also been introduced in proposed algorithm. Those target nodes which are successfully localized in the first round of
localization, are upgraded to the assistant anchor nodes. These nodes help in localization during subsequent rounds of
localization process.
• Since location of target nodes can be formulated as an optimization problem. Therefore, to further improve the localiza-
tion accuracy, we integrate 3D-IDV Hop with GA. By integrating GA with 3D-IDV Hop, we get more accurate localization
of the network with fast convergence.
Simulation results establish the superiority of proposed algorithm in terms of localization accuracy, positioning coverage,
and scalability compared to existing range-free localization algorithms.
The remaining paper is structured as follows: Basic 3D-DV-Hop algorithm and some modified 3D-DV-Hop algorithms
are presented in Section 2. In Section 3, system model and problem formulation are presented. In Section 4, our proposed
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
algorithm (3D-GAIDV Hop) is described. Simulation parameters and simulation results are discussed in Section 5. Finally,
conclusion is drawn in Section 6.
2. Related works
In literature, a number of 2D range-free localization algorithms for WSN have been proposed during last two decades
such as Centroid [5], Distance Vector Hop (DV-Hop) [6], Approximate Point In Triangle (APIT) [11], Convex Position Esti-
mation (CPE) [12], Multidimensional Scaling (MDS) [12], etc. Currently, research on localization for 2D environment has
become mature, but the study of three-dimensional localization is still in its embryonic stage. Therefore, in this paper we
focus mainly on three-dimensional localization for WSN. A number of 3D localization algorithms have been reported in lit-
erature [7–9,12,14,15]. Among all these range-free algorithms, DV-Hop localization algorithm has attracted more attention of
researchers due to its simplicity, stability, feasibility and less hardware requirement. However, DV-Hop algorithm exhibits
poor localization accuracy. Therefore, an improved 3D-DV Hop algorithm based on genetic algorithm has been proposed
in this paper. In next subsection, basic 3D-DV Hop algorithm and improved versions of 3D-DV Hop algorithm have been
presented.
DV-Hop localization algorithm was proposed by Dragons Niculescu et al. [6] for 2D WSN. 3D-DV-Hop algorithm is the
extension of basic DV-Hop algorithm in three-dimensional space. Generally, its process divided into following three steps as
follows:
Step 1 Estimation of minimum hop counts between each target node and anchor node
First of all, each anchor node Ai broadcasts a packet containing the location of Ai and hop count field starting with 0.
During broadcasting of packets, hop count field is increased by 1 with every hop. That means, on reception of this packet,
every node J (here, J can be either target or anchor node) records the location of Ai and initializes hopi,J . Here hopi,J is the
minimum hop count between J and Ai . If the received packet is having a lower hop count value compared to previously
stored hopi,J, then J will update hopi , otherwise, J will ignore the packet. By this process, each node gets the minimal hop
counts to all anchor nodes in the network.
Step 2 Estimation of average hop size of each anchor node
Each anchor node receives the value of minimum hop count and location coordinates of other anchor nodes in step 1.
Now Ai calculates its average distance per hop or hop size denoted as HopSizei as per the following expression.
2 2 2
i= j xi − x j + yi − y j + zi − z j
HopSizei = (1)
i= j hopi j
where (xi ,yi ,zi ) and (xj ,yj ,zj ) are the coordinates of anchor nodes i and j respectively. hopij denotes the hop counts between
anchor node i and j (i = j). Once hop size is calculated, it is broadcasted through the network by Ai .
Step 3 Calculation of target nodes coordinates
After receiving the hop size, a target node can calculate the distance dit from the anchor nodes by multiplying the
HopSizei to the hop count from that particular anchor node as follows:
where dit and hopit are the distance and hop counts between anchor node i and target node t.
After calculating the estimated distances dit from each anchor node Ai , i {1, 2, 3…k} where k is the total number of
anchor nodes, target nodes can use multilateration technique to calculate their coordinates (xt ,yt ,zt ), which is the estimated
coordinate of an target node. Here t {k + 1, k + 2 …n}, n is the total number of target nodes.
⎧
⎪
⎪ (xt − x1 )2 + (yt − y1 )2 + (zt − z1 )2 = d12
⎪
⎪
⎪
⎪
⎨(xt − x2 ) + (yt − y2 ) + (zt − z2 ) = d2
2 2 2 2
. (3)
⎪
⎪
⎪
⎪.
⎪
⎪
⎩
(xt − xk )2 + (yt − yk )2 + (zt − zk )2 = dk2
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
. (4)
⎪
⎪
⎪
⎪.
⎪
⎪
⎩
2(xk−1 − xk )xt + 2(yk−1 − yk )yt + 2(zk−1 − zk )zt = dk2−1 − dk2 − x2k−1 + x2k − y2k−1 + y2k − zk2−1 + zk2
Eq. (4) can be expressed in the linear form as:
AX = B; where
⎡ ⎤
x1 − xk y1 − yk z1 − zk
⎢x − x ⎥
⎢ 2 k y2 − yk z2 − zk ⎥
⎢ ⎥
A = −2 × ⎢
⎢.
⎥
⎥ (5)
⎢ ⎥
⎣. ⎦
xk−1 − xk yk−1 − yk zk−1 − zk
⎡ ⎤
d12 − dk2 − x21 + x2k − y21 + y2k − z12 + zk2
⎢d 2 − d 2 − x 2 + x 2 − y 2 + y 2 − z 2 + z 2 ⎥
⎢ 2 ⎥
⎢ k 2 k 2 k 2 k
⎥
B=⎢
⎢ . ⎥
⎥ (6)
⎢ ⎥
⎣. ⎦
dk2−1 − dk2 − x2k−1 + x2k − y2k−1 + y2k − zk2−1 + zk2
⎡ ⎤
xt
⎢ ⎥
X = ⎣yt ⎦ (7)
zt
After solving Eq. (4) based on least square estimations; target node can obtain its estimated coordinate as follows:
−1
X = AT A AT B (8)
where AT is the transpose of the matrix A, and A−1 is the inverse of matrix A.
Since the 3D-DV Hop algorithm is simple to implement and can localize the target nodes which have a few or even
no neighbouring anchor nodes but its localization accuracy needs to be improved. Many improved versions of 3D-DV Hop
algorithm have been proposed in the literature, some of these are discussed below.
Chen et al. [7] presented a 3D-DV Hop localization based on particle swarm optimization (PSO). In the first part of the
paper traditional DV Hop algorithm is presented. In the second part, PSO technique is proposed to improve the localization
accuracy. The proposed method shows good localization accuracy.
Yun et al. [10] proposed two range-free localization approaches based on RSS information. To overcome the inadequacy
of existing range-free schemes, they employed computational intelligence methods. In the first method, the edge weights
are modelled using FLS and further, these are optimized by GA. Localization is decomposed into many individual problems.
While in the second method, localization is considered as a single problem and Neural Networks (NN) is used to approxi-
mate the locations of the nodes.
Xu et al. [14] presented an improved 3D localization algorithm based on the degree of coplanarity in which distance
measurement is done by the RSSI method. The concept of coplanarity is added into the traditional 3D-DV Hop algorithm to
reduce the positioning error caused by anchor nodes which are coplanar. Those target nodes which have been located are
promoted to assistant anchor nodes to improve the positioning coverage. The Quasi-Newton method is used to correct the
position estimated by quadrilateration.
Feng et al. [15] proposed a multihop localization algorithm that incorporates the distance estimation bias for 3D WSN.
The authors observed that multihop distance estimation errors approximately obey normal distribution with various biases.
Bounding cube technique has been used to find the target node feasible region. The target node feasible region has been
obtained by calculating the intersection area of the bounding cubes. Then the Levenberg–Marquardt method is used to solve
distance estimation bias and target nodes’ location. Using theoretical analysis and simulation results, it has been observed
that the biased-incorporated algorithm has better performance in improving multihop localization accuracy for 3D WSNs.
Peng et al. [16] proposed a genetic algorithm (GA) based DV-Hop algorithm for two-dimensional WSN. The whole algo-
rithm has six steps to perform in which starting three steps are same as for traditional DV-Hop algorithm and last three
steps are of genetic algorithm. Though GADV-Hop algorithm improves the localization accuracy up to an extent but by using
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
genetic algorithm, there are some control parameters (crossover, selection, and mutation) to be tuned correctly for getting
better results.
Ahmad et al. [17] proposed a parametric loop division (PLD) algorithm for 3D localization in WSN. In this method, anchor
nodes form a triangle which uses coordinates of anchor nodes. Triangulation method has been used to find the centre node
as well as two consecutive control nodes from ring control matrix which generates parametric nodes. At each parametric
node, RSSI power is measured and compared with the threshold value to find localized nodes. The authors have validated
their findings through Matlab. The algorithm performs well in terms of localization accuracy and stability for 3D deployment
of sensor nodes.
A range-free 3D localization for an anisotropic medium is proposed by Kumar et al. [8]. FLS is used to reduce the compu-
tational complexity and to model the non-linearity between RSS and distance between nodes. Further, biogeography- based
optimization (BBO), H-best PSO (HPSO) and Firefly Algorithm (FA) are used to optimize the edge weights of the anchor
nodes separately. Anisotropic properties and heterogeneous properties of the network are also considered in the proposed
methods. It is observed that HPSO based algorithm gives better accuracy and fast convergence but BBO based algorithm
gives better accuracy, however, convergence is significantly slower than that for HPSO. A choice between HPSO and BBO is
dependent upon the requirement of accuracy or fast convergence. It is also observed that FA based algorithm gives better
localization accuracy than HPSO and BBO based algorithm in anisotropic 3D WSN.
From the above review, we conclude that localization in WSN is the optimization problem and its overall estimation
error needs to be minimized. This encourages us to propose an improved 3D-DV-Hop algorithm based on genetic algorithm.
Wireless sensor network model comprising of N number of static sensor nodes randomly distributed in the three-
dimensional cubic sensing field of length L has been considered. Node distribution is assumed to be the Poisson distribution
with uniform node density λ = L×LN×L [11]. It is assumed that all SNs have equal communication range and unique ID. These
SNs are assumed to be location unaware and such location unaware nodes are termed as target nodes in rest of the paper.
In order to implement range-free cooperative localization within the network, k number of anchor nodes, which is a
small fraction of N is assumed to be distributed randomly within the sensing field. Fig. 3 shows the random deployment
of SNs (anchor nodes (k) and target nodes (n)) in the three-dimensional sensing field of 150 × 150 × 150 m3 , where anchor
nodes and target nodes are represented by squares and circles respectively. All anchor nodes are assumed to have apriori
knowledge of their respective location. (x1 ,y1 ,z1 ), (x2 ,y2 ,z2 ),…(xk ,yk ,zk ) are the given coordinates of anchor nodes respectively.
The essence of localization problem is to calculate the coordinates (xk+1 ,yk+1 ,zk+1 ), (xk+2 ,yk+2 ,zk+2 ),…(xk+t ,yk+t ,zk+t ) of target
nodes, provided coordinates of k anchor nodes. Therefore, WSN localization problem can be described as follows:
where (xt ,yt ,zt ) and (xi ,yi ,zi ) are the coordinates of the target node and anchor node i respectively, di is the distance between
the target node and anchor node i.
Each sensor node is considered to have high processing and computational capabilities. Such sensor nodes are also com-
mercially available e.g. TelosB, MicaZ motes have 8 MHz TI MSP430 microcontrollers with sufficient storage capacity [18,19].
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Table 1
Example of sorted Lookup Table at target node.
A4 (16,22) 1
A2 (19,13) 2
A5 (5,41) 3
A1 (12,27) 4
A6 (3,46) 5
A3 (13,34) 6
It is well known that 3D-DV Hop algorithm is based on Distance Vector (DV) routing. The distance between the nodes
is estimated by the hop count value multiplied by the hop size of the anchor node. However, there is a large amount
of distance error between the calculated and actual value when there are two or more hop counts between anchor node
and the target node. Such deviation have not been considered in the typical 3D-DV Hop algorithm which leads to poor
localization of the network. In order to mitigate this problem, we have proposed 3D-GAIDV Hop algorithm in this paper.
where (xi ,yi ,zi ) and (xj ,yj ,zj ) are the coordinates of anchor nodes i and j respectively. Dij is the actual distance between two
anchor nodes i and j.
Also, the distance between two anchor nodes can be obtained by multiplying hop counts and hop size of particular
anchor nodes. The distance between two anchor nodes i and j (Di j ), can be calculated as per Eq. (2), which is given as:
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Fig. 4. Optimum hop size calculation of an anchor node (N = 100, k = 25, Communication range = 25 m).
This distance error value of an anchor node i can be used to calculate the correction factor; denoted as ψ i , given as:
i= j i j
ψi = (13)
i= j ho pi j
The correction factor is used to revise the hop size of the anchor nodes by adding it to the previous hop size (calculated
in Eq. 1) as follows:
MH Si = (HopSizei + ψi ) (14)
k
Errori = i j (17)
i=1
Optimum Hop Size of ith anchor node, HopSizei,opt is calculated by reducing the Errori by using the Line Search Algorithm.
This optimum Hop Size is used by target nodes to find their respective location. Fig. 4 illustrates the optimum hop size
calculation of an anchor node.
Now, each target node stores the modified hop size of the respective anchor node in Lookup Table. After calculating
the optimum hop size of anchor nodes, each target node calculates the distance from all anchor nodes according to the
following equation, as:
dit = HopSizei,opt + ψi × hopit (18)
where dit is the modified distance between anchor node i and target node t.
After distance calculation, target nodes can estimate their position by the multilateration technique. For this technique,
a target node needs to select a set of optimal anchor nodes for better location estimation. The selection of optimal anchor
nodes is also proposed in this paper.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
sufficient enough to be considered a good solution. Otherwise, such chromosome will need to go through the selection,
mutation, and crossover process in as many times as possible to be able to select the best result.
In WSN, the main goal of localization algorithms is to determine the location of target nodes. Optimization techniques
can be used to localize the network with minimum localization error. Many localization algorithms based on optimization
techniques have been proposed in the literature [8,16].
The accuracy of location estimation in the proposed method depends on anchor nodes deployment, hop size of anchor
nodes, and hop counts between the nodes. In the proposed method, the distance obtained between anchor nodes and target
nodes (by Eq. (18)) is the calculated value, which may not be accurate. In order to minimize the error, GA has been applied
on 3D-IDV Hop. To solve the node localization problem by GA, a mathematical optimization model has been established
by using the distance between anchor nodes and target nodes. Since node localization in WSN is an optimization problem
whose overall estimation error can be minimized. The objective function and fitness function of localization problem can be
expressed as:
⎛ ⎞
⎜
f (xt , yt , zt ) = Min ⎝ (xt − xi )2 + (yt − yi )2 + (zt − zi )2 − dit ⎟
⎠ (22)
i=1,2...k
t=k+1....k+t
α
f itness(xt , yt , zt ) = (23)
f (xt , yt , zt )
where (xi ,yi ,zi ) are the coordinates of anchor nodes, i = 1,2…k. (xt ,yt ,zt ) are the estimated coordinates of target nodes,
t = k + 1,k + 2....k + t and dit is the distance between anchor nodes and target nodes, calculated according to Eq. (18) and
α is the positive real coefficient.
Generation of initial population:
To minimize the localization error of proposed algorithm, a population feasible region has been created. We have as-
sumed homogenous network for operation, therefore the communication radius of all nodes is considered as R meters. To
understand the generation of initial population for localization problem of the proposed algorithm, let us consider an ex-
ample as shown in Fig. 5, where Nt is the target node whose coordinate is (xt ,yt ,zt ), A1 ,A2 ,A3, and A4 are four anchor nodes
which are the neighbour of target node Nt . (x1 ,y1 ,z1 ), (x2 ,y2 ,z2 ) (x3 ,y3 ,z3 ) and (x4 ,y4 ,z4 ) are the coordinates of A1 ,A2 ,A3 , and
A4 respectively and h1,Nt, h2,Nt ,h3,Nt and h4,Nt are the minimum hop count values between target node Nt and anchor nodes
A1 ,A2 ,A3 , and A4 respectively.
Population feasible region of target node Nt is represented by inner cube in Fig. 5, which bounds the intersection volume
of four spheres. These spheres are constructed with a radius of R × hi,Nt , where hi,Nt is the hop count value from a particular
anchor node i (here i = 1,…4) of target node Nt . Inner cube expresses the possible population feasible region, where Nt lies.
Initially, the random population is generated in this population feasible region such that the coordinate of target node Nt
(xt ,yt ,zt ) must lie in this feasible region in order to achieve the least error. The randomly generated population will search
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
for the accurate coordinate of the target node bounded by upper bound and lower bound, which are defined as follows:
⎧
⎨maxi=1,2..k (xi − R × hi,Nt ) ≤ xt ≤ mini=1,2..k (xi + R × hi,Nt )
⎪
maxi=1,2..k (yi − R × hi,Nt ) ≤ yt ≤ mini=1,2..k (yi + R × hi,Nt ) (24)
⎪
⎩
maxi=1,2..k (zi − R × hi,Nt ) ≤ zt ≤ mini=1,2..k (zi + R × hi,Nt )
Crossover: Arithmetic crossover operator has been used in which linear combination of two chromosomes takes place
for crossover. Suppose (x1 ,y1 ,z1 ) and (x2 ,y2 ,z2 ) are two parents chromosomes selected randomly for crossover, which may
produce (x1 , y1 , z1 ) and (x2 , y2 , z2 ) as two offspring respectively given as:
⎧
⎨x1 = x1 δ + x2 (1 − δ )
⎪
y1 = y1 δ + y2 (1 − δ ) (25)
⎪
⎩
z1 = z1 δ + z2 ( 1 − δ )
⎧
⎨x2 = x2 δ + x1 (1 − δ )
⎪
y2 = y2 δ + y1 (1 − δ ) (26)
⎪
⎩
z2 = z2 δ + z1 ( 1 − δ )
where δ is a random number in a range [0,1].
Mutation: It is a genetic operator used to maintain genetic diversity at a certain mutation rate from one generation
to next generation of chromosomes population. In mutation, the solution may change entirely from the previous solution.
Hence, GA can provide the better solution by using mutation. In the proposed method, if (xt ,yt ,zt ) is chosen for mutation.
Then new coordinate (xt , yt , zt ) is defined as follows:
•F or cos (θ ) ≥ 0
⎧
⎨xt = xt + |xmax − xt |. cos (θ )
⎪
(27)
yt = yt + |ymax − yt |. sin (θ )
⎪
⎩
zt = zt + |zmax − zt |. cos (θ )
•F or cos (θ ) < 0
⎧
⎨xt = xt + |xmin − xt |. cos (θ )
⎪
(28)
yt = yt + |ymin − yt |. sin (θ )
⎪
⎩
zt = zt + |zmin − zt |. cos (θ )
where (xmax ,ymax ,zmax ) and (xmin ,ymin ,zmin ) are the upper bounds and lower bounds of (xt ,yt ,zt ) respectively and θ is the
angle randomly produced in the range of [0, 2π ].
Following steps are carried out for location estimation in the proposed algorithm:
1. Each anchor node floods the packets in the network containing anchor’s ID and anchor’s coordinates.
2. Each target node estimates minimum hop count values from each anchor node.
3. A sorted Lookup Table is maintained by each target node (like as Table I).
4. Each target node calculates the average hop size of every anchor node.
5. Modification of hop size and calculation of optimum hop size of each anchor node are carried out at target node using
the information available in Lookup Table.
6. Then the target nodes select the best localizing unit based on DCP value.
7. Target nodes calculate their location using the multilateration technique and refine their location using genetic algorithm.
8. Localized target nodes are upgraded to assistant anchor nodes for next round of localization.
5. Performance evaluations
In this section, we present the simulation of proposed algorithm viz. 3D-GAIDV hop in MATLAB 8.1 to evaluate the
performance in terms of localization accuracy, computational efficiency, and positioning coverage with respect to different
parameters i.e. Node density, Anchor node ratio, Communication range, and Network connectivity. In this paper, we have
considered the sparse random topology for algorithm operation.
It is assumed that all sensor nodes in the network have equal communication radius as R meters. The values of simulation
parameters are shown in Table 2.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Table 2
Simulation Parameters.
Parameters Value
To evaluate and compare the performance of proposed algorithm with other existing algorithms under similar conditions,
following performance metrics have been considered:
• Localization Error (LE): It is the difference between actual coordinates and estimated coordinates of nodes. LE of each
target node can be calculated as follows:
LE = (xt − xa )2 + (yt − ya )2 + (zt − za )2 (29)
where (xt ,yt ,zt ) and (xa ,ya ,za ) are the estimated and actual coordinates of target nodes respectively.
• Average Localization Error (ALE): It is the ratio of total localization error to the number of target nodes (n). In this paper,
ALE is taken as the performance metric to measure the localization accuracy of the algorithms with respect to different
parameters i.e. node density, anchor node ratio, communication range and sensing area size. It is taken as the average of
total localization error of 200 independent simulation runs for each value of different parameters. Mathematically, it can
be expressed as:
n
t,a=1 (xt − xa )2 + (yt − ya )2 + (zt − za )2
ALE = (30)
n
• Localized Node Proportion (LNP): It is the ratio of the number of successfully localized nodes (nSL ) to the number of
target nodes (n). LNP is the measure of positioning coverage. It exhibits the proportion of target nodes which are suc-
cessfully localized. Mathematically, LNP can be expressed as:
nSL
LNP = (31)
n
• Bad Node Proportion (BNP): It is the ratio of the number of bad nodes (nbadnodes ) to the number of successfully localized
nodes (nSL ). BNP is the measure of stability of algorithm. Bad nodes refer to those nodes which have localization error
larger than their communication range. Mathematically, BNP can be expressed as:
nbadnodes
BNP = (32)
nSL
To evaluate the performance of the proposed algorithm, simulations have been conducted in MATLAB 8.1. In this section,
simulation results are presented to compare the performance of the proposed algorithm with 3D-DV Hop algorithm (ex-
tension of traditional DV-Hop [6] in 3D), 3D-GADV Hop (extension of GADV-Hop [16] in 3D) and 3D-PSODV Hop [7] under
similar conditions. Simulation results evaluate the performance of our proposed algorithm with above mentioned algorithms
with variations of different parameters i.e. anchor node ratio, node density, communication range, and network connectivity.
Instantaneous localization error of each target node is shown in Figs. 6 and 7. The simulation is carried out for 100
sensor nodes randomly spread in the field of 100 × 100 × 100 m3 with anchor node ratio of 30%. Each node is assumed to
have the radio range of 25 m.
Fig. 6 shows the localization error of each target node of 3D-DV Hop algorithm and 3D-improved DV Hop algorithm
(which is also proposed in this paper without application of GA). It can be seen from Fig. 6 that localization accuracy of
3D-improved DV Hop is better as compared to 3D-DV Hop algorithm. Fig. 7 shows the comparison of localization error for
each target node of 3D-GADV Hop, 3D-PSODV Hop, and 3D-GAIDV Hop algorithm. It is observed from the results that our
proposed algorithm outperforms the other variants of 3D-DV hop algorithm.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Fig. 6. Localization error of each target node of 3D-DV Hop and 3D-Improved DV Hop for single simulation run at any instant.
Fig. 7. Localization error of each target node of 3D-GADV Hop, 3D-PSODV Hop, and 3D-GAIDV Hop for single simulation run at any instant.
Table 3
Localization error comparisons.
Localization Algorithm Max. location error(m) Min. location error(m) Avg. location error(m)
Maximum, minimum and average localization error obtained for each localization algorithm is tabulated in Table 3. It can
be seen from Table 3 that 3D-GAIDV Hop performs better compared to other localization algorithms in terms of localization
accuracy.
Localization accuracy of an algorithm solely depends on the number of anchor nodes deployed in the sensing field.
200 number of sensor nodes, each having the communication range of 25 m are randomly deployed in the sensing field of
150 × 150 × 150 m3 . To examine the performance of our proposed algorithm, ALE is calculated with variation in anchor node
ratio from 5% to 40%. Fig. 8 shows the average localization error of different algorithms with variation in the percentage
of anchor nodes deployed in the sensing field. It can be seen from the results that the proposed algorithm performs better
compared to 3D-DV Hop, 3D-GADV Hop and 3D-PSODV Hop algorithm.
It is observed from the Fig. 8 that as the percentage of anchor nodes increases, average localization error of all algorithms
decreases. This is due to the fact that hop count value between the target nodes and anchor nodes decreases with increasing
the number of anchor nodes. Hop size of an anchor node depends on the hop count value. The smaller value of hop count
leads to accurate hop size calculation. Hop size of anchor nodes is modified with increasing number of anchor nodes. Hence,
average localization error of all algorithms decreases with increase in anchor nodes ratio.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Fig. 8. Average localization error with variation in percentage of anchor nodes deployed.
Fig. 9 shows the frequency of error occurrence for different localization algorithms. For this simulation, 200 num-
ber of sensor nodes, each having the communication range of 25 m are randomly deployed in the sensing field of
150 × 150 × 150 m3 with 20% anchor nodes. 200 simulation runs are conducted to measure the average localization errors
of all algorithms. It is observed from the result that large errors of each algorithm occur with very low probability whereas;
mean error of each algorithm occurs very frequently. It can be seen from the Fig. 9 that curve for the frequency of occur-
rence of our proposed algorithm is narrower as compared to other algorithms, which indicates the stability and consistency
of proposed method.
Sensor nodes deployment in the sensing field is application specific. However, there is a significant impact of node den-
sity on localization accuracy. Fig. 10 shows the average localization error for different sensor node density. ALE is calculated
with variation in node density from 50 to 400 in the sensing field of 150 × 150 × 150 m3 with 20% anchor node. The com-
munication range of each node is considered as 25 m.
It is observed from the Fig. 10 that average localization error decreases with the increase in number of sensor nodes.
It is due to the fact that as the number of sensor nodes increases, network connectivity also increases. Due to which large
information about the location of nodes can be collected, which improves the localization accuracy. However, if node density
is increased beyond a certain limit, variation in localization accuracy is marginal which leads to minor variations in ALE. It
can be seen from the result that our proposed algorithm performs better in terms of localization accuracy as compared to
other algorithms.
Communication range of the nodes also affects the localization accuracy of an algorithm. In order to study the effect
of variation in communication range on localization accuracy, 200 number of sensor nodes with 20% anchor nodes are
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
randomly deployed in the sensing field of 150 × 150 × 150 m3 . To examine the performance of our proposed algorithm, ALE
is calculated with variation in communication range of nodes from 15 m to 45 m.
Fig. 11 shows average localization error for different communication ranges of the sensor nodes. Simulation result shows
that the localization error decreases with increase in communication range. It is due to the fact that as communication range
increases, network connectivity also increases, which leads to better coverage and good location estimation of nodes in the
network. It can be seen from the results that our proposed algorithm performs better in terms of localization accuracy in
comparison of other localization algorithms.
To evaluate the effect of variation in sensing area size on localization accuracy of the proposed algorithm, ALE is calcu-
lated with variation in sensing field size from 100 × 100 × 100 m3 to 600 × 600 × 600 m3 . In the simulation, 200 number
of sensor nodes are randomly deployed with 20% of anchor nodes. The communication range of all nodes is considered
as 25 m. Fig. 12 shows the ALE with different sizes of sensing field, in which A1 denotes 100 × 100 × 100 m3 ; A2 denotes
200 × 200 × 200 m3 and so on up to 600 × 600 × 600 m3 , which is denoted by A6. It is observed from the results that ALE
increases of all algorithms with increase in the size of sensing field. It is due to the fact that network connectivity decreases
with the increase in the size of sensing area. As the fixed number of nodes are deployed randomly in the static environ-
ment with fixed communication range, due to this node density decreases with increase in the size of sensing field, which
leads to poor localization accuracy. It can be seen from the results that our proposed algorithm performs better in terms of
localization accuracy as compared to other algorithms due to up-gradation of target nodes to assistant anchor nodes in the
localization process.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Fig. 12. Average localization error with variation in the size of sensing field.
Fig. 13. Localized node proportion (LNP) with respect to network connectivity of the different localization algorithms.
Positioning coverage of an algorithm is also an important parameter to be studied. To investigate the effect of network
connectivity on positioning coverage of proposed algorithm, localized node proportion (LNP) is calculated with variation in
network connectivity from 4 to 16. In the simulations, 200 number of sensor nodes, each having the communication range
of 25 m are randomly deployed in the sensing field of 150 × 150 × 150 m3 .
Figs. 13 and 14 show the successfully LNP and bad node proportion (BNP) with respect to network connectivity of
above mentioned algorithms. In our proposed method, when the network connectivity approaches 10, the location cover-
age reaches to 1, which means all target nodes are successfully localized. The proposed algorithm shows better positioning
coverage compared to other localization algorithms.
Fig. 14 shows the comparisons of bad node proportion of different localization algorithms. There are fewer bad nodes
in the proposed algorithm compared to other methods. When the network connectivity approaches 10, then there is no
bad node in the proposed method. It can be seen from the results that the proposed algorithm shows better positioning
coverage of the nodes compared to other localization algorithms due to up-gradation of target nodes to assistant anchor
nodes in the localization process.
Communication cost of a localization algorithm is measured by calculating the total number of transmitted and received
packets used to localize complete network. To examine the communication cost of all algorithms, TTRP (Total number of
Transmitted and Received Packets) is calculated [24]. TTRP of all algorithms is presented in Table 4, where N is the total
number of nodes in the network, k is the number of anchor nodes and Cavg is the average connectivity.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Fig. 14. Bad node proportion (BNP) with respect to network connectivity of different localization algorithms.
Table 4
Communication cost of all the algorithms.
In step 1 of all algorithms (3D-DV Hop, 3D-GADV Hop, 3D-PSODV Hop, and 3D-GAIDV Hop), anchor node communicates
to all the target nodes one time to broadcast its location. At this point, there is equal communication cost of all algorithms.
In step 2 of 3D-DV Hop and 3D-GADV Hop algorithms, anchor node floods its hop size in the network. Each target node
receives it and forwards only first arrived hop size, while in step 2 of 3D-PSODV Hop, each target node receives hop size
and forwards all arrived hop size information. In step 2 of the proposed algorithm, no communication between nodes takes
place. Since in proposed algorithm, hop size calculation, hop size modification, location optimization, optimal selection of
anchor nodes, etc. are carried out at target node level. Therefore, communication used for step 2 is completely reduced in
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
our proposed algorithm, which is a significant achievement in terms of energy which is also one of the most critical issues
in WSN [22]. From Table 4, it can be observed that TTRP of the proposed algorithm is lesser than other algorithms.
Computational cost is the measure of complexity of an algorithm. It is measured by calculating the time taken by an
algorithm to perform the specific task. To evaluate the computational cost of all algorithms, localization time is calculated.
This is the computational time to compute the location of nodes. Communicational time is not considered during the calcu-
lation of computational time [23,24].
Fig. 15 shows the localization time for all the algorithms with variation in the number of nodes. From results, it can be
seen that localization time for 3D-DV Hop algorithm is lowest and remains consistent with increase in the number of nodes.
3D-GADV Hop is a bit slower than 3D-DV Hop algorithm due to slow convergence of GA. Localization time for proposed
algorithm is slower than 3D-DV Hop and 3D-GADV Hop algorithms but a bit faster than 3D-PSODV Hop. It is due to the fact
that in proposed algorithm, modification of hop size, optimal selection of anchor nodes, location optimization, and nodes
up-gradation to assistant anchor nodes consume more time but it achieves high localization precision. Therefore, in practical
scenarios, we need to keep the balance between localization accuracy and localization time. Overall, the computational cost
of the proposed algorithm is O(k3 ), where k is the number of anchor nodes. Such computational complexity can be easily
handled by commercially available sensor nodes which are having high processing power, large storage capacity, speedily
processing, etc. [18,19,23,25].
In this paper, a distributed range-free localization algorithm for three-dimensional WSN has been proposed viz. 3D-
GAIDV Hop algorithm. In the proposed method, hop size of anchor nodes is modified using the correction factor. The line
search algorithm has been used to find the optimal hop size of the anchor nodes. The distance between target nodes and
anchor nodes is calculated using modified optimal hop size. To reduce the localization errors; the concept of degree of
coplanarity has been introduced, in which coplanar anchor nodes have been ruled out for localization process. For further
improvement in localization accuracy, the genetic algorithm has been applied. With the help of bounded population feasible
region, 3D-GAIDV Hop locates the target nodes accurately. Those target nodes are upgraded to assistant anchor nodes, which
have localized successfully in the first round of localization process and these assistant nodes work as anchor nodes in the
subsequent rounds, which will increase the localization coverage and localization accuracy. In the proposed algorithm, all
the calculations are carried out at target node level due to which communication between the anchor nodes and the target
nodes is significantly reduced and make the algorithm energy-efficient. Simulation results show that proposed algorithm
performs better in terms of localization accuracy, positioning coverage, energy efficiency, and scalability as compared to 3D-
DV Hop, 3D-GADV Hop, and 3D-PSODV Hop algorithms. Future work will be concentrated on the evaluation of the proposed
algorithm in the terrains causing multiple anisotropies due to holes, non-uniform distribution of nodes, sparsity in the
network and irregular radio patterns.
Acknowledgement
This work is partially supported by the National Institute of Technology, Hamirpur, Himachal Pradesh of India (No. B-198)
and Ministry of Human Resource Developments (MHRD) of India with Fundamental Research Funds (No. 2K13-Ph.D-ECE-
227).
Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.compeleceng.
2017.12.036.
References
[1] Akyildiz IF, Su W, Sankarasubramaniam Y, Cayirci E. A survey on sensor networks. IEEE Commun Mag 2002;40(8):102–14.
[2] Boukerche A, Oliveira HA, Nakamura EF, Loureiro AA. Localization systems for wireless sensor networks. IEEE Wirel Commun 2007;14(6):6–12.
[3] Umer T, Amjad M, Afzal MK, Aslam M. Hybrid rapid response routing approach for delay-sensitive data in hospital body area sensor network. In:
Proceedings of the 7th international conference on computing communication and networking technologies. ACM; 2016. p. 3.
[4] Hofmann-Wellenhof B, Lichtenegger H, Collins J. Global positioning system: theory and practice. Springer Science & Business Media; 2012.
[5] Bulusu N, Heidemann J, Estrin D. GPS-less low-cost outdoor localization for very small devices. IEEE Pers Commun 20 0 0;7(5):28–34.
[6] Niculescu D, Nath B. DV based positioning in ad hoc networks. Telecommun Syst 2003;22(1):267–80.
[7] Chen X, Zhang B. 3D DV–hop localisation scheme based on particle swarm optimisation in wireless sensor networks. Int J Sens Netw 2014;16(2):100–5.
[8] Kumar A, Khosla A, Saini JS, Sidhu SS. Range-free 3D node localization in anisotropic wireless sensor networks. Appl Soft Comput 2015;34:438–48.
[9] Chaurasiya VK, Jain N, Nandi GC. A novel distance estimation approach for 3D localization in wireless sensor network using multi dimensional scaling.
Inf Fus 2014;15:5–18.
[10] Yun S, Lee J, Chung W, Kim E, Kim S. A soft computing approach to localization in wireless sensor networks. Expert Syst Appl 2009;36(4):7552–61.
[11] Sharma G, Kumar A. Improved DV-Hop localization algorithm using teaching learning based optimization for wireless sensor networks. Telecommun
Syst 2017:1–16. doi:10.1007/s11235- 017- 0328- x.
[12] Sharma G, Kumar A. Fuzzy logic based 3D localization in wireless sensor networks using invasive weed and bacterial foraging optimization. Telecom-
mun Syst 2017:1–14. doi:10.1007/s11235- 017- 0333-0.
[13] Rashid B, Rehmani MH. Applications of wireless sensor networks for urban areas: a survey. J Netw Comput Appl 2016;60:192–219.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
[14] Xu Y, Zhuang Y, Gu JJ. An improved 3D localization algorithm for the wireless sensor network. Int J Distrib Sens Netw 2015;11(6):315714.
[15] Feng RJ, Guo XL, Wan JW, Wu YF, Yu N. Multihop localisation with distance estimation bias for 3D wireless sensor networks. Electronics letters
2012;48(14):884–6.
[16] Peng B, Li L. An improved localization algorithm based on genetic algorithm in wireless sensor networks. Cognit Neurodyn 2015;9(2):249–56.
[17] Ahmad T, Li XJ, Seet BC. 3D localization based on parametric loop division and subdivision surfaces for wireless sensor networks. In: Wireless and
optical communication conference (WOCC), 2016 25th. IEEE; 2016. p. 1–6.
[18] Crossbow Technology [Online]. http://www.xbow.com/.
[19] Amjad M, Sharif M, Afzal MK, Kim SW. TinyOS-new trends, comparative views, and supported sensing applications: a review. IEEE Sens J
2016;16(9):2865–89.
[20] Mitrinovic DS, Pecaric J, Volenec V. Recent advances in geometric inequalities. Brill Archive; 1989.
[21] Holland J. Adaptation in natural and artificial systems. Ann Arbor: University of Michigan Press; 1975.
[22] Akhtar F, Rehmani MH. Energy replenishment using renewable and traditional energy resources for sustainable wireless sensor networks: a review.
Renewable Sustain Energy Rev 2015;45:769–84.
[23] Sharma G, Kumar A. Modified energy-efficient range-free localization using teaching–learning-based optimization for wireless sensor networks. IETE J
Res 2017:1–15. doi:10.1080/03772063.2017.1333467.
[24] Kumar S, Lobiyal DK. Power efficient range-free localization algorithm for wireless sensor networks. Wirel Netw 2014;20(4):681–94.
[25] Sharma G, Kumar A. Dynamic range normal bisector localization algorithm for wireless sensor networks. Wirel Pers Commun 2017;97(3):4529–49.
doi:10.1007/s11277- 017- 4736- 8.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036
JID: CAEE
ARTICLE IN PRESS [m3Gsc;December 29, 2017;19:29]
Gaurav Sharma obtained Ph.D from National Institute of Technology (NIT), Hamirpur of Himachal Pradesh (India) in 2017. He is currently working as
Assistant Professor in CVR College of Engineering, Hyderabad (India). He has published more than 35 research papers in reputed international journals and
conferences including IEEE. His main research area focuses on Localization in WSN, Routing Protocols, and Optimization.
Ashok Kumar obtained Ph.D from NIT, Hamirpur and currently working as Associate Professor in Department of ECE, NIT, Hamirpur since 1996. He has
published more than 60 research papers in reputed international journals and conferences. His main research area focuses on Wireless Communications,
WSN, Localization techniques, etc. He has 22 years and 10 years of teaching and research experience respectively.
Please cite this article as: G. Sharma, A. Kumar, Improved range-free localization for three-dimensional
wireless sensor networks using genetic algorithm, Computers and Electrical Engineering (2017),
https://doi.org/10.1016/j.compeleceng.2017.12.036