Content uploaded by Essa Q. Shahra
Author content
All content in this area was uploaded by Essa Q. Shahra on Feb 19, 2020
Content may be subject to copyright.
Vol.:(0123456789)
1 3
J Ambient Intell Human Comput
DOI 10.1007/s12652-017-0451-2
ORIGINAL RESEARCH
Perfomance comparison ofthree localization protocols inWSN
using Cooja
TarekR.Sheltami1· EssaQ.Shahra1· ElhadiM.Shakshuki2
Received: 10 August 2016 / Accepted: 20 November 2016
© Springer-Verlag Berlin Heidelberg 2017
1 Introduction
Nowadays, wireless sensor network has turned into an
essential research need. This is because of their extensive
applications that includes civil, manufacturing, farming,
and military applications (Son et al. 2006). A sensor net-
work comprises of sensor devices that are small in size,
inexpensive, and have short transmission range. A sensor
device includes four fundamental components including
processing, sensing, transmission, and power (Hamdi etal.
2008; Antoine-Santoni etal. 2009).
Researchers have focused on various parts of WSN that
includes equipment outline, routing, protection, and locali-
zation (Taleb et al. 2009). One of the basic viewpoints
taken into account is localizing mobile nodes by deploying
sensor networks. The goal of node localization is to find the
physical position of a mobile node (a node with obscure
position) utilizing reference nodes (nodes with predefine
locations) (Alhmiedat and Yang 2009; Abusaimeh and
Yang 2009).
Numerous localization approaches have been proposed
in the literature to provide location data of nodes (Alh-
miedat etal. 2013; Feng etal. 2012; Honma etal. 2015;
Jia etal. 2016). The localization protocols are classified
based on different aspects such as estimation of location,
range-free and range-based, centralized and decentral-
ized (Shen etal. 2005). In range-based approach, nodes
decide their position taking into account angle or dis-
tance calculation to some anchor nodes with known posi-
tions. These estimations may be obtained using different
methodologies such as receive signal strength indicator
(Honma etal. 2015), time of arrival (Zhang etal. 2016),
time difference of arrival (Schmitz etal. 2016), and the
angle of arrival (Kułakowski etal. 2010). There are two
fundamental types of range-free localization protocols
Abstract Wireless sensor networks are applied in a large
number of applications. Indoor localization is drawing sig-
nificant attention in recent years. This is due to the need of
achieving high localization accuracy at a low power con-
sumption. The existing localization protocols are classified
into two main types, including range-free and range-based
protocols. Cooja, which is the Contiki network simulator, is
used to evaluate the performance of different localization
protocols. It facilitates the simulation of Contiki motes in
small and large networks. Moreover, it emulates the motes
at the hardware level. In this paper, we evaluate the accu-
racy and power consumption performance of three known
localization protocols, namely: fingerprint, centroid, and
DV-Hop using Tmote sky in Cooja. This is the first time
this study is conducted in Cooja. The results conform to
the theory that fingerprint protocol has better performance
than centroid and DV-Hop protocols in terms of localiza-
tion accuracy. On the other hand, DV-hop and centroid pro-
tocols outperform fingerprint protocol in terms of stability
and power consumption.
Keywords COOJA· Fingerprint· Centroid· DV-hop·
Wireless sensor network· Localization
* Tarek R. Sheltami
tarek@kfupm.edu.sa
Essa Q. Shahra
g201302670@kfupm.edu.sa
Elhadi M. Shakshuki
elhadi.shakshuki@acadiau.ca
1 King Fahd University ofPetroleum andMinerals, Dhahran,
SaudiArabia
2 Acadia University, NovaScotia, Canada
T.R.Sheltami et al.
1 3
that are recommended for sensor networks (Chen etal.
2013): (1) local strategies that depend on a high thickness
of points of interest so that each sensor node can hear a
few historic points. This is represented by centroid algo-
rithm. (2) Hop based strategies that depend on flooding
the connectivity information in the network, such as hop
count and range-free algorithms. This includes Centroid
and DV-hop (Feng etal. 2012).
One of the commonly known localization methodologies
is the received signal strength (RSS). RSS-based localiza-
tion systems are one of the most known and cheap meth-
ods. They are progressively acknowledged as a localizing
solution for positioning mobile nodes in both indoor and
outdoor environments. RSS-based localization approach
operates by transforming the signal strength to a trans-
mitter-beneficiary, utilizing separate distance estimations.
Walls and obstacles that may reflect and spread the signal
can influence the RSS value. Therefore, this methodology
provides a non-straight change between the RSS values and
the distance. Because of the previously stated impediments,
conveying RSS-based localization approach in indoor situ-
ations turns into a convoluted task, which is hard to design-
ers who utilizing established numerical models. The RSS
data can be utilized to assess the distance between the
transmitter and the recipient, using two methods. In the
first method, the signal proliferation model transforms the
signal strength to distance estimation, utilizing past infor-
mation about the reference point nodes’ location. Secondly,
it sends a geometry approach to process the location of
mobile nodes. This is called a triangulation localization
approach (Smailagic etal. 2000).
The fingerprinting localization approach depends on
predetermined RSS readings. RSS values are accumulated
at a few locations to construct a database of location fin-
gerprints. The fingerprinting-based localization approach
is usually divided into the following two stages: (1) offline
stage: this stage incorporates measuring the position of
mobile target focus in a few ranges. Then, storing the col-
lected RSS values at all points with the corresponding posi-
tion in a database file. (2) Online stage: the versatile target
collects a few RSS values from various guide nodes within
its proximity and sends the information to a server. The
server then processes this information and evaluates the
versatile target’s area based on the produced results (Cheng
etal. 2005).
This paper main aim is to study and compare the accu-
racy as well as the power consumption of three different
protocols of indoor localization algorithms, namely: Fin-
gerprint, (range-based), Centroid and DV-Hop (range-
free). The rest of the paper is organized as follows: Sect.2
explores the fingerprint localization approach. Section 3
discusses the centroid localization approach. Section 4,
introduces DV-hop. Section 5 discusses the simulation
results and the performance evaluation using Cooja simula-
tor. Finally, Sect.6 provides the conclusion.
2 Fingerprint localization approach
In this section, we introduce a fingerprinting localization
approach keeping in mind the end goal is to reduce the
localization error accomplished in the trilateration based
methodology (Milioris et al. 2014). Fingerprinting locali-
zation protocol is one of the most encouraging approach,
because of its simplicity and its high exactness regarding
localization (Gogolak etal. 2011). Fingerprinting protocol
requires the accumulation of a substantial number of ref-
erence points in the tracking range to accomplish sensible
localization precision. There are two fundamental issues to
build up a fingerprinting. The first issue is collecting the
RSS values and storing them in a database that requires a
large amount of time when the localization area is large.
The second issue is how to figure out the location using the
values stored in the database which is considered a trouble-
some (Ficco 2014).
In this section, a fingerprinting-based localization meth-
odology is explained. This methodology diminishes the
aggregate number of reference nodes, which needs to be
gathered in the offline stage while accomplishing low local-
ization error between 1 and 3.5m. An indoor fingerprinting
methodology is incorporated into two main stages. In the
first stage, the database is created. In the second stage, fea-
ture identification and estimation are performed. The first
stage is executed during the offline phase, while the second
stage is executed during the online stage.
2.1 Stage 1: fingerprint database creation
This phase starts by dividing the study area into grid points.
Each of which has its own X and Y coordinates, defined by
P = (X, Y). This phase involves the following two steps:
1. For each grid point, the RSS values are collected from
the three beacons {b1, b2, b3} within its transmission
range and then stored in the database.
2. For each subarea, the number of grid points for each
subarea and the range of each subarea is determined
based on the collected RSS values.
In the first step, the mobile target moves manually
through the grid points one-by-one and collects the RSS
values from the beacons. An RSS vector is formed at each
grid point and represented by: rssbi, rssbj, and rssbk. These
vectors are collected to build the database. In the second
step, the main objective is to determine the number of grid
points for each subarea. The reasons of knowing the range
Perfomance comparison ofthree localization protocols inWSN using Cooja
1 3
of each subarea are: (1) any change in the network topology
can be recovered by the RSS for the same grid points in
the same subarea. (2) Knowing the number of grids points
in each subarea, the search space is reduced and enhanced.
The RSS values are stored in the database as shown in
Table1.
2.2 Stage 2: dividing
In this stage, a group of beacon IDs addresses is used as
an identifier for each subarea. Three beacons IDs determine
each subarea. Assume that the first subarea is represented
by Ac, and its identifier beacons are (B1, B2, B3). Therefore,
all RSS values that are received from these three beacons
belong to subarea
Ac
and each subarea has its own range
of RSS values. For example, subarea
Ac
has the range from
1 to 30. A mobile target uses this range in the estimation
phase to get the nearest three grid points from the database.
During this stage, the location of a mobile node is calcu-
lated. This stage involves the following two steps:
1. Determining the subarea
Ac
where the mobile target is
located.
2. Finding the nearest three grid points to the target point
that depends on the RSS values read from the beacons
in the same subarea.
In the first step, to compute the location of a mobile
node, the three anchors start measuring the RSS values of
the mobile node. The ID address of the received beacons
is used as an identifier to determine in which subarea the
mobile target is found. In the second step, to locate the
mobile node in the subarea, it is necessary to find the near-
est three grid points to mobile target. This is achieved by
comparing the mobile RSS values with RSS values stored
in the database for the same subarea. In addition, we need
to split the RSS values of each beacon in the same subarea
into two vectors (Vmax and Vmin). The Vmax vector includes
all RSS values that are greater than the RSS value of the
mobile node; whereas Vmin vector includes all RSS values
that are less than or equal to the mobile RSS value. Then,
the smallest RSS value from Vmax and the largest RSS value
from Vmin are selected. Then, the difference between these
two values and mobile value is calculated to get the nearest
one. This process is repeated for all three beacons in the
subarea. The three nearest points to mobile target are cen-
troid to get the position of the mobile node or target.
3 Centroid localization approach
The Centroid localization approach depends on a great
thickness of references so that each mobile node receives
notification from a few beacons (Dong and Xu 2014).
Depending on the round radio propagation presumption,
every mobile node computes its location by determining
the center position of all received anchor nodes. The inter-
esting point of the centroid localization methodology is that
it does not require coordination between references nodes.
This methodology offers reasonable localization precision.
The implementation of its algorithm includes two stages. In
the first stage, all anchors are required to send their coordi-
nates,
Bj
(x,y), j = 1,…,n), to all mobile sensor nodes within
their transmission area. In the second stage, all mobile
nodes compute their location M(x,y) by deriving the aver-
age for the coordinates of all n locations of the anchors in
range, using Eq.(1).
where, M(
x, y
) is the coordinate of the mobile target, n is
the total number of beacons in the transmission area, and
Bj(x,y) is the coordinates of the beacon node j.
4 DV‑hop localization
In the DV-hop algorithm, the anchor broadcasts a packet
in its transmission range in the network (Gui etal. 2015).
The transmitted packet contains the coordinate of this
anchor node. It also contains a hop metric with an initial
value equals to one. The packet is broadcasted in the net-
work with increment in hop count. The receiver saves the
packet with a small number of hops and discards the packet
with a large number of hops. The DV-hop implementation
(1)
M
(x,y)= 1
n
n
∑
j
=
1
Bj(x,y
)
Table 1 Fingerprint database Grid no. Coordinates RSS from
Bi
RSS from
Bj
RSS from
Bk
Subarea
Identifier
X Y
1 x1y1rssbi1 rssbj1 rssbk1 A1
. . . . . . .
. . . . . . .
. . . . . . .
N
xn
yn
rssbin
rssbjn
rssbkn
An
T.R.Sheltami et al.
1 3
includes three steps. In the first step, every anchor broad-
casts its position data and hops count in the network. This
way ensures that all anchor nodes receive location informa-
tion with the minimum hop count in the network topology.
In the second step, each anchor node computes the average
distance per hop, using Eq.(2):
where
(xi,xj)
and
(yi,yj)
are the actual and coordinates of
anchor
i
and
j
, and
hi,j
is the hop metric value between two
anchor nodes
i
and
j
and
S
represents the set of anchors.
After computing the average distance, each anchor
broadcasts its average distance of hops (ADH) in its trans-
mission range. Before this step finishes, each beacon node
will have a routing table that includes the destination node
ID,
x
coordinate,
y
coordinate and minimum hop count.
Consequently, each beacon node will contain all the nec-
essary information about all beacons in the network. The
unknown node receives the first ADHs from the beacons
and discards the later ones. After that, the unknown node
uses the received ADHs to calculate the distances to the
first three anchors by multiplying each ADH with a number
of hops to its anchor. In the third step, the positions of the
unknown nodes are computed using trilateration method,
using Eq.(3).
5 Simulation results
5.1 The simulation platform
Cooja is a sensor network simulator for the Contiki OS
(Bagula and Erasmus 2015). The Contiki OS is a portable
OS designed for restricted resource devices, such as sen-
sor nodes. It is constructed around event-driven kernel;
however, it supports multi-threading. Likewise, it supports
full TCP/IP stack by means of uIP and programming proto-
threads. The main objective of Cooja is its extendibility to
utilize the interface and plug-ins. The interface represents
the devices or motes and the plug-ins facilitate the inter-
action with the simulator such as to control the speed of
simulation or watch network traffic between nodes.
The mobile model describes the movement pattern of
mobile nodes and how their locations change. In our work,
the mobile model is used in which the mobile nodes ran-
domly and periodically change their location. We used
Tmote sky, shown in Fig.1. Tmote sky is a wireless sensor
(2)
ADH
=j∈S,j≠i
xi−xj2+yi−yj
2
j∈S,j≠i
h
i,j
(3)
d
2
1=(X−X1)
2
+(Y−Y1)
2
d
2
2=(X−X2)2
+(Y−Y2)
2
d
2
3
=
(
X−X
3)
2
+
(
Y−Y
3)2
module that offers high data rate with ultra-low power con-
sumption. Some other features to look for are high reliabil-
ity and ease of development. It is widely proven platform
for wireless sensor systems deployments.
5.2 Performance evaluation offingerprint technique
The simulation setup is a network with 15 nodes deployed
in an area of 40 × 20m, including five anchor devices and
ten unknown devices. The unknown devices can be local-
ized inside the study area. The percentage of the beacon
is 33%, and the wireless communication range is 30m.
The study area is divided into three subareas; each has its
own identifier and numbers of grid points. The topology
is a grid of 30 points that cover the total study area. Dur-
ing our experiments, we start by collecting the RSS val-
ues of grid points and store them in the database, during
an offline stage. A snapshot from Cooja simulator inter-
face is shown in Fig.2.
The execution of the algorithm starts by broadcasting
a HELLO message from the mobile node to all anchor
nodes in its transmission range. The mobile node receives
responses from the anchors and consequently measures
the RSS values for all beacons. In addition, the identi-
fier of the subarea of the mobile node is determined. To
select the nearest three grid points to the mobile node, all
RSS values of grid points in that subarea are compared
with RSS values measured between the mobile node and
the anchors. Finally, the location of the mobile node is
computed and sent to the gateway.
5.2.1 Localization accuracy
The fingerprint localization system is evaluated by ana-
lyzing the error in the derived location. We compute the
location error between the exact coordinates and the esti-
mated location of the mobile node, using Eq.(4):
The results of the fingerprint localization system
are presented in Table2. It contains the node ID, exact
(4)
Error
=
√(
Xexact −Xestimated
)
2
+
(
Yexact −Yestimated
)
2
Fig. 1 TmoteSky
Perfomance comparison ofthree localization protocols inWSN using Cooja
1 3
coordinates of all unknown nodes, and the estimated
coordinates of unknown nodes computed by fingerprint
localization system. Table2 also shows that fingerprint
localization gets the accuracy ranging between 2m (in
the best case) and 3.162m (in the worst case). The aver-
age error is 2.4m. This means that the accuracy is excel-
lent (less than 3 m). These results accepted by applica-
tions that need a high accuracy.
5.2.2 Power consumption
To estimate the power consumption of the fingerprint
system, we use the PowerTrce plug-in provided by Cooja
simulator. It allows us to monitor the number of ticks per
seconds for real time (rtimer), which is a structure that
represents a real-time task. The collected data is shown in
Table3
The raw data presented in Table3 is used to calculate
the power consumption of the mobile node at any time,
using Eq.(5).
where rd1, rd2 are the two sequence values from the raw
data. Current and voltage represent the current and voltage
values of TmoteSky sensor taken from the data sheet. The
RTIMER represents the number of ticks per second repre-
sented in Contiki OS. Table4 vides the power consumption
of the fingerprint system. It contains the power consump-
tion of CPU, LPM, Tx, Lx and total power consumption.
Figure 3 presents the power consumption for Finger-
print system. It shows that the transmissions (Tx) reach the
peak value at 6 and 14s, because the mobile node replies
(5)
Power
(mw)=
(
rd2−rd1
)
×current ×
voltage
RTIMER
Fig. 2 Fingerprint localization system
Table 2 Fingerprint localization accuracy
Node ID Exact coordinates Estimated coordi-
nates
Error (M)
X Y X Y
1 19 14 17 15 2.236
2 12 6 15 5 3.162
3 40 15 42 13 2.828
4 27 10 27 8 2
5 39 3 39 2 2
6 43 10 41 10 2
7 31 10 29 9 2.236
8 23 7 20 8 3.162
9 15 16 17 15 2.236
10 44 3 45 5 2.236
AVG 2.422
Table 3 Raw data of fingerprint (ticks)
Time (s) CPU LPM TX LX
0 2382 63,099 0 390
2 5583 125,428 0 806
4 8789 187,753 0 1222
6 16,460 245,615 1865 1612
8 31,891 295,715 1865 2637
10 35,195 357,943 1865 3053
12 38,490 420,180 1865 3469
14 46,550 477,650 3721 4111
16 50,210 539,518 3721 4953
18 53,505 601,755 3721 5369
20 56,799 663,993 3721 5785
T.R.Sheltami et al.
1 3
to the anchor nodes that request to measure the RSS of the
mobile. The listening (Lx) reach the peak value at 8 and
16s because it is the time of the mobile node to move from
one position to another. Low power mode (LPM) is almost
remain with a constant value.
The duty cycle is calculated using Eq.(6), by utilizing
the data presented in Table3.
(6)
Duty cycle
(%)=
(rd
2
−rd
1
)
Tx
(
(rd2−rd1)CPU +(rd2−rd1)LPM
)
Figure4 presents the duty cycle status for the Transmis-
sion and Listening states. It shows the duty cycle as per-
centage or ratio of the time required to change the status of
Tx to Lx or vice versa.
5.3 Performance evaluation ofcentroid technique
Our experiments are performed in a network with 20 nodes
deployed in an area of 90 × 17. In this network, there is
eight anchor devices and 12 unknown devices. The per-
centage of the beacon is 40%, while the wireless commu-
nication range is 28 m. The deployment of nodes for the
localization systems is shown in Fig.5. When we start run-
ning Cooja simulator, the anchor nodes send their location
information. Then, the unknown mobile node receives the
beacons’ information of the first three beacons for which it
calculates its position, as shown in Fig.5. Nodes with IDs
4, 12 and 15 have the coordination of (33, 52), (39, 55), and
(58, 55) respectively.
5.3.1 The localization accuracy
To evaluate the localization accuracy of the centroid sys-
tem, the location error between the exact locations and
estimated locations are computed by centroid localiza-
tion, using Eq. 4. Table5 shows the achieved results and
a summary of location errors for the Centroid localization
system. Table 5 contains the node ID, exact coordinates
of all unknown nodes and the estimated coordinates of
unknown nodes that are computed by centroid localization
system. Based on the achieved results, the centroid locali-
zation accuracy ranges from 2.975785m (in the best case)
to 5.8180495 m (in the worst case). The average error is
4.518m. The accuracy should be accepted by less number
of applications than that of fingerprint localization system
method. Also, the results show that all unknown nodes
located in the same block that is covered by the same three
anchors get the same location. This is because the average
value of the three anchors has the same value.
5.3.2 Power consumption
To estimate the power consumption of the Centroid sys-
tem, we utilize the PowerTrce plug-in provided with Cooja
simulator. This allows us to monitor the number of ticks per
seconds for rtimer for the CPU, LPM, Tx and Lx; similar to
what we did in fingerprint approach. The collected data is
shown in Table6.
We use Eq.5 with the data listed in Table6 to calculate
the power consumption of the mobile node. Figure6 is a
representation of the data provided in Table7.
We can see that Tx still remains constant with zero
value. This is because no transmission is performed from
Table 4 Power consumption (power in mW)
Time (s) CPU LPM TX LX Total
2 0.263754 0.155499 0 0.358008 0.777261
4 0.264166 0.155489 0 0.358008 0.777663
6 0.632071 0.144355 1.792831 0.335632 2.904889
8 1.271475 0.12499 0 0.882111 2.278576
10 0.272241 0.155247 0 0.358008 0.785496
12 0.2715 0.15527 0 0.358008 0.784777
14 0.664124 0.143377 1.78418 0.552502 3.144183
16 0.301575 0.154349 0 0.724622 1.180545
18 0.2715 0.15527 0 0.358008 0.784777
20 0.271417 0.155272 0 0.358008 0.784697
0
1
2
3
4
246810 12 14 16 18 20
Power (mW)
Time (Seconds)
CPU LPM TX LX Total
Fig. 3 Power consumption graph of fingerprint system
Fig. 4 Duty cycle of fingerprint system
Perfomance comparison ofthree localization protocols inWSN using Cooja
1 3
the mobile node; also Lx state is constant with value of
0.0716. On the other hand, LPM and CPU have a little
change in their values.
The duty cycle can also be calculated from the data pro-
vided in Table 6, using Eq. 6. Figure 7 presents that Lx
status has a full time 100%, as compared to Tx due to no
transmission from the unknown nodes to the anchor nodes.
5.4 Performance evaluation ofDV‑hop technique
The deployment of nodes for localization systems is shown
in Fig.8, utilizing the same scenario used in Sect.5.3. By
running Cooja simulator, nodes start to exchange their
information packets with the anchors for building their
routing tables, as shown in Fig.8. It shows the final routing
table for all anchor nodes. These tables contain all the coor-
dinates as well as the minimum hops of all other anchor
nodes in the network.
Figure9 depicts the beacons broadcast the ADH within
the proximity of their transmission area. This allows the
unknown nodes to calculate the distance to the anchors
to be used in the trilateration process of computing the
Fig. 5 Centroid localization system
Table 5 Centroid localization accuracy
Node id Exact coordination Estimated coor-
dination
Error (M)
X Y X Y
1 14.76 55 19 55 4.24
2 26.65 60.37 23 61 3.7039708
3 40.44 57.74 39 55 3.0953514
4 49.92 63.83 49 61 2.9757856
5 64.81 55.64 59 55 5.8451433
6 21.18 59.85 19 55 5.3174148
7 25.49 65.64 29 61 5.8180495
8 38.65 60.37 39 55 5.3813939
9 51.18 58.48 49 61 3.3320864
10 60.65 60.22 59 55 5.4745685
AVG 4.518
Table 6 Raw data of centroid system (ticks)
Time (s) CPU LPM TX LX
2 2500 63,040 0 390
4 5809 125,262 0 806
6 9138 187,464 0 1222
8 12,488 249,646 0 1638
10 15,862 311,804 0 2054
12 19,236 373,961 0 2470
14 22,610 436,118 0 2886
16 25,984 498,275 0 3302
18 29,358 560,432 0 3718
20 32,733 622,589 0 4134
Fig. 6 Power consumption of centroid system
T.R.Sheltami et al.
1 3
location of unknown nodes. Figure9 also shows the final
coordinates of some unknown nodes. These nodes include
node ID 7 with coordinate (47, 39) and node ID 10 with the
coordinate of (29, 55).
5.4.1 Localization accuracy
To analyze and evaluate the accuracy of this system, we
use the data presented in Table8 and calculate the location
errors using Eq.4. Table8 provides a summary of the loca-
tion errors in the network. This table contains the node
ID, the exact coordinates of all unknown nodes and the
estimated coordinates of unknown nodes computed by the
DV-Hop localization system. Table8 also shows that DV-
Hop localization accuracy range between 0.9669023m (in
the best case) and 3.5846897 m (in the worst case). The
average error is 2.551m. The distance is the same for all
unknown nodes in the same block (covered by same three
anchor nodes).
5.4.2 Power consumption
To calculate the power consumption, we use the PowerTrce
plug-in in Cooja simulator. Table 9 shows the achieved
results from the monitored number of ticks per seconds for
rtimer for the CPU, LPM, Tx, and Lx.
The data presented in Table 9 is used to calculate the
power consumption of the mobile node at different times,
using Eq.(5). The results presented in Table10 is also rep-
resented in Fig.10.
The duty cycle is calculated from the data shown in
Table9, using Eq.(6). Figure11 presents that Lx status has
a 100% full time compared to Tx as there is no transmis-
sion from the unknown nodes to the anchor nodes.
6 Conclusions andfuture work
A performance evaluation in terms of localization accuracy
and power efficiency between fingerprint, centroid, and
DV-hop localization protocols is conducted using Cooja
simulator. TmoteSky sensors were used in all experiments.
The anchor nodes were arranged in triangle topology and
Table 7 Power consumption (mW) of Centroid system
Time (S) CPU LPM TX LX Total
2 0.00999 0.0006266 0 0.0716 0.0822
4 0.01005 0.0006264 0 0.0716 0.0822
6 0.01012 0.0006262 0 0.0716 0.0823
8 0.01019 0.0006259 0 0.0716 0.0824
10 0.01019 0.0006259 0 0.0716 0.0824
12 0.01019 0.0006259 0 0.0716 0.0824
14 0.01019 0.0006259 0.0716 0.0824
16 0.01019 0.0006259 0 0.0716 0.0824
18 0.01019 0.0006259 0 0.0716 0.0824
20 0.0.010197 0.00062597 0 0.0716 0.0824
Fig. 7 Duty cycle of centroid system
Fig. 8 DV-Hop routing table
Perfomance comparison ofthree localization protocols inWSN using Cooja
1 3
moving targets were within the vicinity of at least three
anchors. The results showed that the fingerprint localiza-
tion provides accuracy range between 2 and 3.162m with
an average error of 2.422 m. For centroid localization,
the accuracy range is between 2.976 and 5.818m with an
average error of 4.518m. Finally, for DV-hop localization,
the accuracy range is between 0.967 and 3.585m with an
average error of 2.551m. We conclude that the fingerprint
Fig. 9 DV-hops results
Table 8 DV-Hop localization data
Node id Exact coordinates Estimated coor-
dinates
Error (M)
1 13.24 41.21 14 38 3.3016616
2 16.25 35.85 14 38 3.1120733
3 29.97 52.57 29 55 2.616448
4 29.4 52.66 29 55 2.3739419
5 45.47 40.99 47 39 2.5101793
6 46.82 39.95 47 39 0.9669023
7 47.47 40.78 47 39 1.8410052
8 10.83 38.83 14 38 3.2768583
9 12.78 39.5 14 38 1.9334942
10 45.2 42.1 47 39 3.5846897
AVG 2.551
Table 9 Data of DV-hop system (ticks)
Time (s) CPU LPM TX LX
2 2395 63,116 0 390
4 10,215 120,822 0 1486
6 19,559 177,005 0 2580
8 28,902 233,188 0 3674
10 38,270 289,346 0 4768
12 47,638 345,504 0 5862
14 57,006 401,662 0 6956
16 66,374 457,820 0 8050
18 75,743 513,978 0 9144
20 85,111 570,136 0 10,238
Table 10 Power consumption (mW) of Centroid system
Time (S) CPU LPM TX LX Total
2 0.023626 0.000581 0 0.188643 0.21285
4 0.02823 0.000566 0 0.188298 0.217095
6 0.028227 0.000566 0 0.188298 0.217092
8 0.028303 0.000566 0 0.188298 0.217167
10 0.028303 0.000566 0 0.188298 0.217167
12 0.028303 0.000566 0 0.188298 0.217167
14 0.028303 0.000566 0 0.188298 0.217167
16 0.028306 0.000566 0 0.188298 0.21717
18 0.028303 0.000566 0 0.188298 0.217167
20 0.028342 0.000565 0 0.188298 0.217206
Fig. 10 Power consumption of DV-Hop system
T.R.Sheltami et al.
1 3
protocol has the best average localization accuracy fol-
lowed by DV-hop protocol. Nevertheless, fingerprint pro-
tocol has an overhead of setting the network and measuring
all the points in the grid. For the power performance, both
DV-hop and centroid protocol outperform fingerprint pro-
tocol in terms of stability and power consumption. In the
future, we plan to perform our experiments with different
types of sensors such as Zolertia and MICAz.
Acknowledgements The authors would also like to thank King
Fahd University of Petroleum and Minerals and Acadia University for
their support.
References
Abusaimeh H, Yang S-H (2009). Reducing the transmission and
reception powers in the AODV. Paper presented at the Net-
working, Sensing and Control, 2009. ICNSC’09. International
Conference
Alhmiedat TA, Yang S (2009). Tracking multiple mobile targets
based on ZigBee standard. In: Proceedings of the 35th Annual
Conference of the IEEE Industrial Electronics Society
Alhmiedat T, Samara G, Salem AOA (2013) An indoor fingerprint-
ing localization approach for ZigBee wireless sensor networks.
arXiv preprint arXiv:1308.1809.
Antoine-Santoni T, Santucci J-F, De Gentili E, Silvani X, Morandini
F (2009) Performance of a protected wireless sensor network in a
fire. Analysis of fire spread and data transmission. Sensors 9(8):
5878–5893
Bagula B, Erasmus Z (2015) Iot emulation with cooja. Paper pre-
sented at the ICTP-IoT Workshop
Chen C-C, Chang C-Y, Li Y-N (2013) Range-free localization scheme
in wireless sensor networks based on bilateration. Int J Distrib
Sens Netw 2013
Cheng BH, Hudson RE, Lorenzelli F, Vandenberghe L, Yao K (2005)
Distributed gauss-newton method for node loclaization in wire-
less sensor networks. Paper presented at the IEEE 6th Workshop
on Signal Processing Advances in Wireless Communications
Dong Q, Xu X (2014) A novel weighted centroid localization algo-
rithm based on RSSI for an outdoor environment. J Commun
9(3):279–285
Feng W-J, Bi X-W, Jiang R (2012) A novel adaptive cooperative loca-
tion algorithm for wireless sensor networks. Int J Autom Comput
9(5):539–544
Ficco M (2014) Calibration-less indoor location systems based on
wireless sensors. J Ambient Intell Human Comput 5(2):249–261
Gogolak L, Pletl S, Kukolj D (2011) Indoor fingerprint localization in
WSN environment based on neural network. Paper presented at
the 2011 IEEE 9th International Symposium on Intelligent Sys-
tems and Informatics
Gui L, Val T, Wei A, Dalce R (2015) Improvement of range-free
localization technology by a novel DV-hop protocol in wireless
sensor networks. Ad hoc Netw 24:55–73
Hamdi M, Boudriga N, Obaidat MS (2008) WHOMoVeS: an opti-
mized broadband sensor network for military vehicle tracking.
Int J Commun Syst 21(3):277–300
Honma N, Ishii K, Tsunekawa Y, Minamizawa H, Miura A (2015)
DOD-based localization technique using RSSI of indoor bea-
cons. Paper presented at the 2015 International Symposium on
Antennas and Propagation (ISAP)
Jia M, Sun J, Bao C (2016) Real-time multiple sound source locali-
zation and counting using a soundfield microphone. J Ambient
Intell Human Comput 1–16
Kułakowski P, Vales-Alonso J, Egea-López E, Ludwin W, García-
Haro J (2010) Angle-of-arrival localization based on antenna
arrays for wireless sensor networks. Comput Electr Eng
36(6):1181–1186
Milioris D, Tzagkarakis G, Papakonstantinou A, Papadopouli M,
Tsakalides P (2014) Low-dimensional signal-strength fin-
gerprint-based positioning in wireless LANs. Ad hoc Netw
12:100–114
Schmitz J, Hernández M, Mathar R (2016) Real-time indoor localiza-
tion with TDOA and distributed software defined radio: demon-
stration abstract. Paper presented at the Proceedings of the 15th
International Conference on Information Processing in Sensor
Networks
Shen X, Wang Z, Jiang P, Lin R, Sun Y (2005) Connectivity and
RSSI based localization scheme for wireless sensor networks.
Paper presented at the International Conference on Intelligent
Computing
Smailagic A, Small J, Siewiorek DP (2000) Determining user location
for context aware computing through the use of a wireless LAN
infrastructure. Institute for Complex Engineered Systems Carn-
egie Mellon University, Pittsburgh, p15213
Son B, Her Y-S, Kim J-G (2006) A design and implementation of for-
est-fires surveillance system based on wireless sensor networks
for South Korea mountains. Int J Comput Sci Netw Secur IJC-
SNS 6(9):124–130
Taleb AA, Pradhan DK, Kocak T (2009) A technique to identify and
substitute faulty nodes in wireless sensor networks. Paper pre-
sented at the Sensor Technologies and Applications, 2009. SEN-
SORCOMM’09. Third International Conference
Zhang H, Seow CK, Tan SY (2016) Virtual reference device-based
narrowband TOA localization using LOS and NLOS path. Paper
presented at the 2016 IEEE/ION Position, Location and Naviga-
tion Symposium (PLANS)
Fig. 11 Duty cycle of DV-Hop system
A preview of this full-text is provided by Springer Nature.
Content available from Journal of Ambient Intelligence and Humanized Computing
This content is subject to copyright. Terms and conditions apply.