Você está na página 1de 7

2009 11th IEEE International Conference on High Performance Computing and Communications

A High Energy Efficient Localization Algorithm


for Wireless Sensor Networks Using Directional Antenna

Baoli Zhang Fengqi Yu Zusheng Zhang


Department of Integrated Electronics
Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences
Shenzhen, Guangdong Province, 518067, China
bl.zhang@siat.ac.cn & fq.yu@siat.ac.cn

significant advantages of knowing the location


Abstract information of sensor nodes [1]. First, location
information is required to identify the location of an
Directional antenna offers various benefits for event of interest. For instance, the location of an
wireless sensor networks, such as increased spatial intruder, the location of fire, or the location of enemy
reuse ratio and reduced energy consumption. tanks in battlefield is of critical importance for
Numerous localization schemes using anchor nodes deploying rescue and relief troops. Second, location
equipped with omni-directional antenna to transmit awareness facilitates numerous application services,
beacon information have been proposed. However, the such as location directory services that provide doctors
radiated RF signal from an omni-directional antenna with the information of nearby medical equipment and
is easier to be interfered by wide range of environment personnel in a smart hospital, target-tracking
noise, which leads to greater localization error. In this applications for locating survivors in debris, or enemy
paper, we tackle the problems in estimating location of tanks in a battlefield. Third, location information can
randomly deployed sensor nodes with low gain omni- assist in various system functionalities, such as
directional antennas. Since directional antenna geographical routing, network coverage checking, and
concentrates energy on a particularly direction with a location-based information querying. Hence, with
high gain and narrow covering area, we propose a these advantages and much more, localization should
high accuracy, energy efficient, and low cost be considered as an implicit feature of a sensor
localization scheme using a mobile anchor node network [1].
equipped with a directional antenna. After Some localization schemes were proposed to have a
investigating the effect of directional antenna on GPS receiver or other hardware to obtain distance. To
wireless sensor networks, we show how the location reduce hardware cost, some people suggest solutions
estimation problem can be solved using geometric of using anchor nodes which already know their
characteristics on a 2D plane with directional antenna. absolute locations via GPS or manual placement and
Our localization scheme is verified by OPNET other sensor nodes estimate their locations based on
simulation software. The result shows that the the information provided by these anchor nodes [2].
proposed localization algorithm has higher accuracy As an alternative solution, some people proposed
and energy efficiency than other localization schemes. localization algorithms by mounting omni-directional
antennas on anchor nodes due to simplicity and cost
efficiency [3]. However, the radiated energy in all
1. Introduction
directions is easier to be interfered by wide range of
environment noise, which leads to big localization
Wireless sensor networks (WSNs) greatly extend
errors. In contrast, directional antenna concentrates
our ability to monitor and control the physical world,
energy on a particular narrow direction with a higher
as they have become the mark of pervasive technology.
gain, the reduction in interference to surrounding
WSNs have an endless array of potential applications,
environment noise can improve localization accuracy.
such as target tracking, military, habitat sensing, and
By using directional antenna, a node may also be able
fire detection. Most of these applications require
to selectively receive signals only from a certain
knowledge on the position of every node in the WSNs.
desired direction. This enables the receiver node to
Therefore, sensing data without knowing the sensor’s
avoid most of interference that comes from unwanted
location is meaningless. In any WSNs, the location
directions, thereby increasing signal to interference
information of node plays a crucial role in
and noise ratio (SINR). Due to higher gain, directional
understanding the application context. There are three

978-0-7695-3738-2/09 $25.00 © 2009 IEEE 230


DOI 10.1109/HPCC.2009.15
antenna has a greater transmission range than omni- message transmitted by a single mobile node. After
directional antenna. Therefore anchor node with receiving the beacon message, a sensor node computes
directional antenna may transmit beacon information its location by applying RSSI technique. The accuracy
more effectively than that with omni-directional can be improved when the sensor node receives more
antennas. beacons. In [12], the distance from a sensor node to a
Inspired by the Self-Organizing Medium Access mobile anchor node is computed based on ultrasonic
Control for Sensor Networks (SMACS) [4], we technology. Once node to node distance is available,
propose a simple, distributed, low-cost, high energy the sensor location can be estimated by a range based
efficient, reliable, and accurate localization scheme by localization scheme. Reference [13] presents a scheme
employing a mobile anchor node equipped with a to reduce the position uncertainty of a sensor node
directional antenna and a GPS receiver. caused by radio connectivity constrains. Reference [14]
The rest of the paper is organized as follows. proposes probabilistic localization schemes with a
Section 2 provides an overview of existing location mobile beacon. The approaches use TOA technique for
algorithms using anchor nodes. In Section 3, our ranging and centroid formula with distance
proposed localization scheme is discussed in-depth. information for calculating sensor position.
The simulation and analysis are presented in Section 4.
In Section 5, we draw our conclusion. 2.3. Our consideration
In the localization schemes using stationary anchor
2. Related work node, each sensor node needs to collect the location
information of its neighbouring anchor nodes and then
In this section we provide an overview of current compute its location. Therefore a certain percentage of
research in WSNs localization and describe the work anchor nodes are necessary, which increases the cost
that is mostly related to ours. The localization of the network. On the other hand, the localization
algorithm can be classified into two categories, namely, schemes using mobile anchor node are cost effective
based on stationary anchor node and based on mobile because only one or a few mobile anchor nodes are
anchor node. needed. However their performance is seriously
affected by environment noise. Directional antenna
2.1. Localization scheme using stationary offers a potential solution to the problem. It has the
anchor node advantages of counteracting interference, long
transmission range, and spatial reuse. Many of these
Localization scheme using stationary anchor node
advantages come at cost of increasing complexity in
can be further classified into range based and range
communication protocols, such as DMAC [15].
free localization scheme [5].
Localization with a mobile anchor node using
2.1.1. Range based localization scheme directional antenna, some common problems in
Range based localization scheme use range (distance conventional communication protocols, such as hidden
or angle) information to estimate a sensor location. terminal and deafness, are automatically resolved.
Range estimation consists of identifying the distance or
angle between two communication nodes by using 3. Algorithm design
TOA [6], TDOA [7], AOA [5], or RSSI [8].
2.1.2. Range free localization scheme In this section, we propose a distributed and energy
Considering the drawbacks of range based efficient localization scheme that allows a sensor node
localization scheme, some people have proposed to estimate its location with high accuracy by using a
location algorithms without using distance or angle mobile anchor node equipped with a directional
information, Such as centroid algorithm [9], APIT [10], antenna and GPS receiver.
and DV-HOP [2].
3.1. Preliminaries
2.2. Localization scheme using mobile 3.1.1. Antenna Model
anchor node According to beam pattern (beam-radius, beam-
width, and beam orientation), antenna models are
One or more anchor nodes fly over or move through
classified into two types: omni-directional antenna (as
a sensor area to broadcast beacon information and
shown in Fig. 1(a)) and directional antenna (as shown
unlocalized sensor nodes estimate their locations using
in Fig. 1(b)). An omni-directional antenna transmits
received beacon information [11]. In [11], the authors
and receives data in all directions with a gain (Go). On
proposed a localization method based on beacon

231
the contrary, a directional antenna transmits and transmit beacon message. The mobile anchor node has
receives data only in one direction with a gain of Gd, a transmission range R and an open angle ȕm, as shown
where Gd is typically greater than Go. Directional in Fig. 3. For a sensor node Si in the open area ȕm, Si
antenna can point its main lobe towards a specified can receive beacon message from the mobile anchor
direction. We approximate the radiation pattern of the node. On the contrary, Si cannot receive any beacon
side lobes as a sphere with the node at its center, the message.
gain of the side lobes is assumed to be very small
3.1.3. Communication model
(approximate to zero).
In our proposed localization scheme, a sensor node
only need to receive beacon messages from anchor
node and does not need to receive any information
Main lobe
from its neighbours. In order to resolve location
ȕm estimation problem, we propose D-O (directional -
antenna
Omni directional) communication model, as shown in
Side lobes
Fig. 3. For a sensor node S in the area ȕm, it can receive
(a) (b) beacon message from AS, and for other sensors S1, S2 ,
S3, S4, they cannot.
Fig. 1 Antenna model: (a) omnidirectional antenna (b) single beam
directional antenna.
In omni-antenna, signal electromagnetic energy is
spreaded over a large region of space, while only a
small portion is received by an intended receiver.
Directional antenna can concentrate electromagnetic
energy into a certain direction (main lobe) while cancel
the energy in other directions, resulting in an amplified Fig. 3 D-O communication model
signal in a certain direction.
Traditionally, both the Omni and directional mode 3.1.4. The radiation pattern of a single beam
may be used to transmit or receive message. But in our directional antenna
proposed border line intersection localization (BLI) As shown in Fig. 4, a sensor node can receive
scheme, the omni mode is used only by sensor nodes to beacon message from directional antenna if it is within
receive beacon messages from mobile anchor node. an open angle ȕm, where ȕm is formed by border line L1
The directional mode is used only by a mobile anchor and L2. The greatest gain happens in the middle of
node to transmit beacon messages. Using this antenna, open angle ȕm, i.e. L0, and sensor node can get
the best channel quality can be achieved to avoid noise maximum RSSI. A sensor node cannot receive any
interface. beacon message if it is in the area of Ƹ=3600-ȕm.
3.1.2. Network model 3.2. Location determination
Each mobile anchor node broadcasts beacon
message with single directional antenna, which
contains: 1) The mobile anchor node’s coordinates and
timestamp when the beacon message is broadcasted; 2)
The sector boundary lines of angle ȕm are defined by
the directional antenna and angle Į is defined by
border line L2 and X axis, as shown in Fig. 4. Based on
the collected beacon messages, sensor nodes determine
their locations by using the geometric characteristics of
Fig.2 Network model. the confined area. To achieve this goal, we suggest a
localization scheme, namely BLI, which will be
The network model in the proposed localization
discussed in detail in the follow sections.
scheme is shown in Fig. 2. We assume two-tier
Our wireless sensor network is shown in Fig. 2,
network architecture with a set of sensor nodes S
where the mobile anchor node is carried on a vehicle
which are randomly deployed in a certain area A. A
(or aircraft). The mobile anchor node moves through
mobile anchor node AS, equipped with a single beam
the sensor area and transmits beacon messages to
directional antenna and GPS, moves through or flies
sensor nodes for localization. We here make two
over the area A with a certain or random speed to

232
assumptions. First, the mobile anchor node has a GPS ­ L1 : y  y1 k1 ( x  x1 )
receiver to get its location coordinate and sufficient ® ` (1)
energy to broadcast beacons. Second, the mobile ¯ L 2 : y  y2 k2 ( x  x2 )
anchor node is equipped with a directional antenna and
the sensor nodes are equipped with omni directional From equation (1), the coordinates of S can be
antenna, as shown in Fig. 1. calculated as x = (a-b)/(k2-k1), y = (a×k2-b×k1)/(k2-k1),
where, a = y1-k1×x1, b = y2-k2×x2 , k1 = tan(Į), k2 =
tan(Į + ȕm).
Y
L2 L`2 L1 L`1

ȕm ȕm

Į
N1 (x1, y1) N2 (x2, y2) Į
Fig. 4 The radiation pattern of a single beam directional antenna. 0
X

Fig. 5 Sensor location estimation by BLI.


3.2.1. Beacon point selection
BLI localization scheme is based on the following
principle: a sensor node S receives the first and last 3.3. Performance Analysis
beacon message from a mobile anchor node. When
sensor S finds two beacon points, it looks for border 3.3.1. Localization error analysis
lines of ȕm, which are L1 an L’2, respectively. The
intersection point of two lines is the location of sensor
node S. In BLI localization scheme, we use two beacon
points to obtain a sensor node location. One of them,
denoted as N1, is the position for the sensor node to
receive the first beacon message from the mobile
anchor node, which is used to define border line L1.
Another position is where the sensor node to receive
the last beacon message, denoted as N2, which is used
to define second border line L’2. The mobile anchor
node moves through the sensor area and broadcasts Fig. 6 Localization error analysis.
beacon messages periodically. The beacon message
includes the anchor node’s position, timestamp, Į, and
ȕm. The lifetime must be long enough to assure the
anchor node can pass through the sensor area. Each
sensor node maintains a visiting list and a localization
list. The visiting list contains the RSSI values and
packet information at the time when the sensor node
receives the packet from the anchor node. The
localization list includes beacon points N1, N2, Į, and
ȕm.
Fig. 7 Short distance movement pattern.
3.2.2. Location estimation
After a sensor node S obtains two beacon points and The proposed localization scheme works with high
defines two border lines, it can compute its location. accuracy if the selected beacon points are ideal, i.e.
As shown in Fig. 5, the beacon points are N1 and N2, their real positions match what are calculated. In
which location coordinates are (x1, y1) and (x2, y2), practice, however, incorrect beacon points could be
respectively. The coordinates of S can be obtained by chosen due to collision or inappropriate beacon
message interval. The error in location of a beacon

233
point directly leads to the location error of sensor node. node S locates behind obstacle O and S cannot receive
As shown in Fig. 6, localization error is caused by any beacon message. A solution to the problem is
incorrect beacon points. The real border line is L1, proposed, as shown in Fig. 8 as well. When mobile
while the estimated border line is L’1 and the estimated anchor node AS reaches point P1, P2, P3, or P4, it stops
location is S’. We propose a method to reduce the error and broadcasts beacon message with stronger
caused by environment noise interfere. transmission power. Sensor node S will receive at least
We set a threshold (Ȝ) for the minimum RSSI value. two beacon messages and use them to estimate its
From Fig. 6, we can approximate distance error h as: location.
h d u sin(T ) (2)
4. Simulation
where d is the distance between sensor node S and
beacon point N1 For small angle ș, one can get
In this section, we discuss simulation setup,
hĝd (3) localization error, execution time, and throughput, and
then compare with Yu’s [16] and Ssu’s localization
Based on (3), the localization accuracy can be algorithm [1]. OPNET with wireless LAN module is
improved by decreasing the distance between sensor adopted in the simulation. The radio model is based on
node S and mobile anchor node. Thus Short Distance IEEE 802.11 wireless LAN product with 1 Mbps
Movement Pattern (SDMP) is proposed, as shown in bandwidth. The IEEE 802.11 wireless LAN distributed
Fig. 7. Every sensor node can localize itself with high coordination function (DCF) was used for medium
accuracy if the mobile anchor node moves following access control (MAC) model. Each sensor node is
SDMP. forced to stay active during localization process. After
it obtains its location, the sensor node could schedule
3.3.2. Location retrieval
its state for power efficiency.
Generally speaking, any localization algorithm
cannot guarantee all sensor nodes in a network to get 4.1. Simulation environment
their locations. In addition a new sensor node needs to
localize itself when it joins a network. Therefore some TABLE I
kind of location retrieval technique must be SIMULATION PARAMETERS
considered in every localization scheme. In the Parameter Value
Beam width 50, 100, 150, 200, 250
proposed localization algorithm, every beacon Distance between two adjacent line D (m) 25
message includes a timestamp. If localization process Directional antenna gain (dB) 10
is over and a sensor node has not localized itself, it Omni directional antenna gain (dB) 0
will broadcast location request to its neighbours. After Packet transmission period (sec) 0.1, 0.3, 0.5, 0.7, 0.9
receiving the request, the neighbours will send out Max. moving speed (m/sec) 10, 20, 30, 40, 50
their location information. The unlocated sensor node The size of sensor field is 500x500 m2 and 200
takes this information to estimate its location using sensor nodes are randomly deployed in the area. The
APIT [12] or other anchor-based localization schemes. mobile anchor node is equipped with a GPS receiver
3.3.3. Obstacle immunization and a single beam directional antenna. It moves in
straight lines across the sensor area. Each sensor node
is equipped with omni-directional antenna receiver.
The mobile node has no energy restriction and each
stationary sensor node is energy restrained. The other
simulation parameters are listed in TABLE I

4.2. Metric
Three metrics are employed to evaluate the
performance of our localization algorithm.

Fig. 8 Obstacle immunity.


4.2.1. Average localization error
A sensor node may not receive any beacon message It is defined as average difference between
when a big obstacle stands between an anchor node estimated location (Xei, Yei) and actual location (Xi, Yi)
and the sensor node, as shown in Fig. 8, where sensor of all sensor nodes [16]

234
N
proposed algorithm is less than those of the other two
¦ ( X ei  X i )2  (Yei  Yi )2
algorithms. In addition, it is less affected by beacon
'Eav i 1 (4)
N interval.
where N denotes the total number of stationary sensor 4.3.2. Average localization error vs. average
nodes. throughput
In a localization scheme, the number of packets
4.2.2. Average execution time received by sensor nodes determines network energy
It is defined as the time needed for localize all consumption. Therefore it should be minimized to
sensor nodes. reduce energy consumption for localization. The
N average throughput for the three localization
¦t i algorithms are shown in Fig. 10, where the mobile
(5)
ATav i 1
anchor node moves at a speed of 10m/s and beam
N
width is 5 degree. We can clearly see that our
where tj denotes the time required for sensor node i to
localization scheme outperforms the other two
obtain its location and N denotes the total number of
schemes. This is because, in our localization algorithm,
sensor nodes.
each sensor node receives less number of beacon
4.2.3. Average throughput messages than Yu’s and Ssu’s algorithm.
It is defined as
N
4.3.3. Average localization error vs. average energy
¦R i
consumption
Rav i 1 (6)
N

Average throughout(number of packets)


where Ri denotes the number of packets received by 400
stationary sensor node i to estimate its location. 300

4.3. Simulation Result 200 BLI


Ssu's scheme
4.3.1. Average localization error vs. beacon interval 100 Yu's scheme
0
18 0.1 0.2 0.3 0.5 0.9
Average localization error (m)

16 BLI Beacon interval (s)

14
Ssu's scheme
Yu's scheme
12
10 Fig. 10 Average localization error vs. throughout.
8
The major energy consumption is due to packet
6
4
reception. The energy consumption cased by
2 computation is relatively small and can be ignored. So
0 the average throughout may reflect the average energy
0.9 0.7 0.5 0.3 0.1 consumptions in localization. If we assume the energy
Beacon interval (s)
required for receiving a packet is a constant, the energy
consumption of our proposed localization scheme is
much less than that of the other two schemes, as shown
Fig. 9 Average localization error vs. beacon interval. in Fig. 10. This is because in our proposed localization
The beacon interval is the time between transmitting scheme a stationary sensor node receives less number
two adjacent beacon messages. In Yu’s and Ssu’s of beacon messages, while other localization schemes
schemes, sensor location is computed based on beacon constantly receive messages for determining beacon
points where sensor node receives packets from a points.
mobile anchor node. Beacon interval determines the 4.3.4. Average localization error vs. anchor moving
maximum number beacon messages that sensor node speed
can receive when the mobile anchor node moves at a In order to collect enough beacon messages for a
constant speed. So beacon interval greatly affects sensor node to localize itself, a mobile anchor node
localization accuracy, which is shown in Fig. 9, where with a faster moving speed must lower its beacon
the mobile anchor node moves at a speed of 10m/s and interval. Fig. 9 presents localization error vs. beacon
beam width is 5 degree. The localization error of our interval, where anchor node moves at a speed of 10m/s

235
and beam width is 5 degree. Localization error for B. Furht (ed.), CRC Press, Taylor and Francis Group,
different moving speed and beacon interval is shown in 2008.
TABLE II. The performance of all localization [2] D. Niculescu and B. Nath, “Ad Hoc Positioning System
schemes is not affected by moving speed. (APS) using AOA,” in Proceedings of IEEE
International Conference on Computer Communication
TABLE II (INFOCOM), pp. 1734-1743, March 2003.
AVERAGE LOCALIZATION ERROR vs. AVERAGE [3] A. Joshua, P.Lee, Sensor Network Localization via
EXECUTION TIME Received Signal Strength Measurements with
Anchor moving speed Directional Antennas, Proc. 42nd Annual Allerton
10 20 30 40 50
(m/sec)
Conference on Communication, Control, and
Beacon interval (sec) 0.1 0.05 0.033 0.025 0.02
Yu’s
Computing, pp 1861-1870, Monticello, IL, Sept. 2004.
0.82 0.81 0.81 0..80 0.80 [4] K. Sohrabi and G. J. Pottie, “Performance of a novel
algorithm
Average
Ssu’s self-organization protocol for wireless ad hoc sensor
localization 5.34 5.35 5.33 5.34 5.33
error (m)
algorithm networks,” in Proc. IEEE 50th Vehicular Technology
BLI Conf., pp. 1222–1226, 1999.
0.48 0.48 0.49 0.46 0.47
scheme
[5] B. Azzedine, A.B.F. Oliveira, F.Nakamura, A.F.L
Yu’s
128.16 75.63 65.35 57.26 50.65 Loureiro Localization System for Wireless Sensor
algorithm
Average Networks, IEEE Wireless Communications, Dec. 2007.
Ssu’s
execution 127.89 74.88 66.11 57.13 50.70
algorithm [6] B. Hofmann-Wellenho, H. Lichtenegger, and J. Collins,
time (sec)
BLI
63.26 35.10 31.11 28.09 25.08 Global Positioning System: Theory and Practice, 4thed.,
scheme Springer-Verlag, 1997.
4.3.5. Average localization error vs. average [7] B. Priyantha, C. Anit, and B. Hari,, The Cricket
Location-Support system, Proc. 6th ACM MOBICOM,
execution time
Boston, MA, Aug. 2000.
Average execution time is an important criterion in [8] D. Niculescu and B. Nath, “DV Based Positioning in Ad
the performance evaluation of a localization scheme. Hoc Networks,” J. Telecomm. Systems, vol. 22, pp.
TABLE II shows localization error as a function of 267-280, Jan.-Apr. 2003.
average execution time. The average execution time of [9] N. Bulusu, J. Heidemann, and D. Estrin, “GPS-less low
BLI scheme varies less than that of Ssu’s and Yu’s cost outdoor localization for very small devices,” IEEE
scheme. In BLI scheme, a sensor node can get enough Pers. Commun., vol. 7, no. 5, pp. 28–34, Oct. 2000.
beacon messages when a mobile anchor node passes [10] X. Li, H. Shi, and Y. Shang, “A Partial-Range-Aware
though it. Therefore less number of beacon messages Localization Algorithm for Ad-Hoc Wireless Sensor
Networks,” Proc. IEEE Int’l Conf. Local Computer
is received by the sensor node in BLI scheme.
Networks, pp. 77-83, Aug. 2004.
[11] M.L. Sichitiu and V. Ramadurai, “Localization of
5. Conclusion Wireless Sensor Networks with A Mobile Beacon,”
Proc. First IEEE Int’l Conf. Mobile Ad-Hoc and Sensor
Given inherent constrains of a sensor device Systems (MASS ’04), pp. 174-183, Oct. 2004.
envisioned and estimation accuracy desired by [12] P. Dutta and S. Bergbreiter, MobiLoc: Mobility
location-dependent applications, we proposed a enhanced localization.
[13] A. Galstyan, B. Krishnamachari, K. Lerman, and S.
distributed and range free localization scheme (BLI) by
Patterm, “Distributed Online Localization in Sensor
using directional antenna. Based on beacon messages Networks Using A Mobile Target,” Proc. Third Int’l
sent by a mobile anchor node, sensor nodes can Symp. Information Processing in Sensor Networks
estimate their locations. The BLI scheme can be used (IPSN ’04), pp. 61-70, Apr. 2004.
for any size and any density of sensor networks. By the [14] M.L. Sichitiu and V. Ramadurai, “Localization of
use of a single mobile beacon, all computation is Wireless Sensor Networks with A Mobile Beacon,”
accomplished locally. Simulation shows the proposed Proc. First IEEE Int’l Conf. Mobile Ad-Hoc and Sensor
BLI has higher localization accuracy and lower power Systems (MASS ’04), pp. 174-183, Oct. 2004.
consumption compared to Ssu’s and Yu’s localization [15] R.R. Choudhury, X.Yang, R.rRamanathan and
H.Vaidya, “Using Direntional Antennas for Medium
scheme.
Access Control in Ad Hoc Networks”, MOBICOM’02.
[16] T. He, C. Huang, B. Lum, J. Stankovic, and T.
6. References Adelzaher, “Range-free localization schemes for large
scale sensor networks,” in Proc. ACM MobiCom, San
[1] A. Srinivasan and J. Wu, "A Survey on Secure Diego, CA, pp. 81–95, Sept. 2003.
Localization in Wireless Sensor Networks,"
Encyclopedia of Wireless and Mobile Communications,

236