You are on page 1of 5

(IJCSIS) International Journal of Computer Science and Information Security,

Vol. 8, No. 8, November 2010

Localization Accuracy Improved Methods Based on

Adaptive Weighted Centroid Localization Algorithm
in Wireless Sensor Networks

Chang-Woo Song, Jun-Ling Ma, Kyung-Yong Chung Kee-Wook Rim

Jung-Hyun Lee Department of Computer Department of Computer and
Department of Information Information Engineering, Sangji Information Science, Sunmoon
Engineering, INHA University, University, Wonju, Korea University, Asan, Korea
Incheon, Korea,

Abstract—Generally, see Localization of nodes is a key technology kind of monitoring, tracking, or controlling. Specific
for application of wireless sensor network. Having a GPS applications include habitat monitoring, object tracking,
receiver on every sensor node is costly. In the past, several nuclear reactor control, fire detection, and traffic monitoring. In
approaches, including range-based and range-free, have been a typical application, a WSN is scattered in a region where it is
proposed to calculate positions for randomly deployed sensor meant to collect data through its sensor nodes.
nodes. Most of them use some special nodes, called anchor nodes,
which are assumed to know their own locations. Other sensors A sensor network is composed of a large number of sensor
compute their locations based on the information provided by nodes that are densely deployed in a field. Each sensor
these anchor nodes. This paper uses a single mobile anchor node performs a sensing task for detecting specific events. The sink,
to move in the sensing field and broadcast its current position which is a particular node, is responsible for collecting sensing
periodically. We provide an adaptive weighted centroid data reported from all the sensors, and finally transmits the data
localization algorithm that uses coefficients, which are decided by to a task manager. If the sensors can’t directly communicate
the influence of mobile anchor node to unknown nodes, to with the sink, some intermediate sensors have to forward the
prompt localization accuracy. We also suggest a criterion which data [1].
is used to select mobile anchor node which involve in computing
the position of nodes for improving localization accuracy. The There are several essential issues (e.g., localization,
localization accuracy of adaptive weighted centroid localization deployment, and coverage) in wireless sensor networks.
algorithm is better than maximum likelihood estimation which is Localization is one of the most important subjects for wireless
used very often. sensor networks since many applications such as environment
monitoring, vehicle tracking and mapping depend on knowing
Keywords-component; Weighted Centroid Algorithm; Wireless the locations of the sensor nodes. In addition, with location-
Sensor Networks; Localization; based routing protocols, both routing and data forwarding are
determined based on the geographic location [2].
To solve the localization problem, it is natural to consider
A wireless sensor network (WSN) consists of spatially placing sensors manually or equipping each sensor with a GPS
distributed autonomous sensors to cooperatively monitor receiver. However, due to the large scale nature of sensor
physical or environmental conditions, such as temperature, networks, those two methods become either inefficient or
sound, vibration, pressure, motion or pollutants. The costly, so researchers propose to use a variety of localization
development of wireless sensor networks was motivated by approaches for sensor network localization.
military applications such as battlefield surveillance. They are
now used in many industrial and civilian application areas, These approaches can be classified as range-based and
including industrial process monitoring and control, machine range-free. Firstly, the range-based approach uses an absolute
health monitoring, environment and habitat monitoring, node-to-node distance or angle between neighboring sensors to
healthcare applications, home automation, and traffic control. estimate locations. Common techniques for distance or angle
estimation include received signal strength indicator (RSSI),
A sensor network normally constitutes a wireless ad-hoc time of arrival (TOA), time difference of arrival (TDOA), and
network, meaning that each sensor supports a multi-hop routing angle of arrival (AOA). The approaches typically have higher
algorithm (several nodes may forward data packets to the base location accuracy but require additional hardware to measure
station). In computer science and telecommunications, wireless distances or angles. Secondly, the range-free approach does not
sensor networks are an active research area with numerous need the distance or angle information for localization, and
workshops and conferences arranged each year. The depends only on connectivity of the network and the contents
applications for WSNs are varied, typically involving some

ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 8, November 2010
of received messages. For example, Centroid method, APIT estimates using offline hop-distance estimations through
method, DV-HOP method, Convex hull, Bounding box, and neighbor information exchange. Another existing range-free
Amorphous algorithm have been proposed [3][4][5]. Although scheme is an APIT algorithm. APIT resolves the localization
the range-free approach cannot accomplish as high precision as problem by isolating the environment into triangular regions
the range-based [6], they provide an economic approach. Due between anchor nodes. A node uses the point-in-triangle test to
to the inherent characteristics (low power and cost) of wireless determine its relative location with triangles formed by anchors
sensor networks, the range-free mechanism could be a better and thus narrows down the area in which it probably resides.
choice to localize a sensor’s position, so we pay more attention APIT defines the center of gravity of the intersection of all
to range-free approach in this paper. triangles that a node resides in as the estimated node location
This paper uses a single mobile anchor node as the
reference node, which is required to move in the sensing field Based on these analyses, localization using a single mobile
and broadcast its current position periodically. Sensor nodes anchor node would be more economical. In addition,
receive the position information of the mobile node and considering the constraints in computing and memory power of
localize themselves to the centroid of these positions by using sensors, we adopted the weighted centroid method with a
adaptive weighted centroid algorithm. The algorithm based on single mobile anchor to locate sensors in wireless sensor
the Received Signal Strength Indication (RSSI). The results of networks. Depending on the method used for ranging, an
simulations show that the method is a practical method that can appropriate localization technique is applied in the second
be used in real-world system, and is also a method whose phase. The following localization strategies have been
principle is simple, less computing and communication, is low proposed.
cost, and provides flexible accuracy.
A. Trilateration
II. RELATED WORK This is one of the more popular strategies and is used when
In the past several years, extensive research has been done the exact distances between known points and an object to be
on localization for wireless sensor networks. A general survey located are available. Fig. 1 shows when the distance between
is found in. Here we provide only a brief survey about range- an object and three points are given, the object's location x can
free approaches and localization method, which involve mobile be computed as the intersection of three circles centered at the
reference nodes. Some nodes are equipped with special known points.
positioning devices that are aware of their locations. These
nodes are called anchor nodes or reference nodes. Other nodes
that do not initially know their locations are called unknown
nodes or sensor nodes. Generally, an unknown node can
estimate its location by range-based or range-free methods if
three or more anchors are available in its coverage field.
Obviously, the number and position of anchor nodes have a
noticeable influence on the localization precision.
The main idea of localization with a mobile anchor node is Figure 1. Example of Trilateration
as follows: After sensor deployment, a mobile anchor node
traverses the sensor network while broadcasting anchor
B. Bounded Intersection
packets, which contain the coordinates of the anchor node.
Sensor nodes receiving anchor packets could infer their The trilateration technique works well when the three
distance from a mobile anchor node and use these circles intersect at a single point, but this is rarely the case
measurements as constraints to construct and maintain position when estimates are used in ranging. For example, when using
estimates. These methods have a common feature: they use incremental stepping of transmission power for ranging,
range-based approaches. Though they can reach fine resolution, maximum values can be used for estimating the distances. Fig.
either the required hardware is expensive (ultrasound devices 2 shows The object to be located would fall into a geometric
for TDOA, antenna arrays for AOA) or the results depend on region that is the intersection of three circles.
other unrealistic assumptions about signal propagation (for
example, the actual received signal strengths of radio signals
can vary when the surrounding environment changes). Due to
the hardware limitations of sensor devices, range-free
approaches are a cost effective alternative to a more expensive
range-based approach. A simple algorithm proposed, computes
location as the centroid of its proximate anchor nodes. An
alternate solution, DV-Hop, extends the single hop broadcast to
multiple-hop flooding, so that sensors can find their distance Figure 2. Localization with Maximum Bounds
from the anchors in terms of hop counts. An amorphous
positioning scheme adopts a similar strategy as DV-Hop; the
major difference is that Amorphous improves location

ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 8, November 2010
C. Triangulation nodes. This can be done by driving the mobile anchor node to
The triangulation method is useful if the angle between two move in a spiral trajectory. Obviously there are many other
objects can be measured. Fig. 3 shows an example. Suppose P1 options to moving trajectory. Finding an optimal trajectory to
and P2 are points with known locations and X is an object to be cover all sensor nodes can be a research topic for our future
located. Nodes P1 and P2 can measure angles a1 and a2 , and, work. No matter which trajectory is used, the location of the
with known distance Sx , one can easily compute ax , S1 and mobile anchor node on the trajectory should be known. At the
S2. same time, we assume that the mobile anchor has sufficient
energy for moving and broadcasting its information during the
localization process. The speed of the mobile anchor is
adjustable and unrestricted, but uniform in the process of
We used an idealized radio model for wireless
communication because it was simple and easy to reason
mathematically. We assumed that our idealized model is
perfect spherical radio propagation and has identical
transmission range (power) for all radio positions as shown in
Fig. 5. It is a sphere with the anchor as its center and the
Figure 3. Example of Triangulation broadcasting radius R as its radius. Only the sensors within the
range are assumed capable of receiving the information sent by
the anchor.
D. Maximum Likelihood
When estimates are used for ranging, it is possible that the
region of intersection is empty. This will occur if at least one
ranging estimate is too small. One method that overcomes this
problem selects the point for localization that gives the
minimum total error between measured estimates and
distances. In Fig. 4, distance estimates (d1, d2, d3) are made
between the object to be located and three points (P1, P2, P3) .
The errors (e1, e2, e3) are computed by finding the difference Figure 5. System Environment with a Mobile Anchor Node
between the actual Euclidean distances and the ranging
estimates. In this paper, we proposed the location of mobile anchor
node influence: In the localization algorithm, location of
mobile anchor node has influence to the unknown nodes, RSSI
bigger location, and the greater influence on the location of
sensor nodes. When Unknown node received multiple mobile
anchor node position signal then unknown node by the impact
of these locations. Location of largest RSSI has the greatest
power to decide to the position of sensor node.
Figure 4. Localization with Maximum Likelihood Signal selection principle: An unknown node may receive
multiple signals of positions from the mobile anchor node.
III. ADAPTIVE WEIGHTED CENTROID LOCALIZATION RSSI value should be the largest of several signals position
METHOD calculation. Location computed to ensure that the signals
involved in more than three. Will be distances of more than R
A. Method of Localization the location of mobile anchor node removed, so as to avoid the
This method can be used in large-scale field environment. expansion of the positioning error. Behind the simulation
Figure 1 illustrates the system environment where a sensor proves this point.
network consists of a mobile anchor node and unknown nodes
that could be scattered from a plane or from a mortar shell. The B. Adaptive Weighted Centroid Localization Algorithm
mobile anchor is a human operator or an unmanned vehicle Through the front of the Analysis, can find common
deployed with the sensor network. If the network is deployed centroid algorithm, did not reflect the mobile anchor node's
by plane scattering, this anchor can be even the plane itself. influence, affecting the localization accuracy. To enhance the
The unknown nodes are the nodes of initially unknown localization accuracy, in this paper we used the adaptive
positions. Once the nodes are deployed, they will stay at their weighted centroid localization algorithm. Its main idea: In the
locations to conduct the sensing task. The mobile anchor, algorithm, mobile anchor node confronts the right to decide the
which is a node aware of its location (e.g. equipped with GPS), location of the centroid through weighted factor to reflect. The
and is able to traverse for assisting the sensors to determine use of weighted factor reflected the intrinsic relationship
other node locations [10]. The mobile anchor node needs to between them.
traverse over the entire region in order to cover all sensor

ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 8, November 2010
We embody this relationship through the formula of the
weighted factor: 6) Calculates the mean value(X,Y) of C_n^3 coordinate.
The (X,Y) is Unknown node coordinate.
( + ) ( + + )
X = d1 d 2 d 3 , Y = d1 d 2 d 3 (1) A. Simulation Environment
1 1 1 1 1 1
( + + ) ( + + ) The key metric for evaluating a localization technique is the
d1 d 2 d 3 d1 d 2 d 3 accuracy of the location estimates versus the communication
and deployment cost. To evaluate this proposed method we use
Fig. 6 illustrates Known 3 mobile anchor nodes coordinate UNIX, programs with the C language. We have carried on the
(X1, Y1), (X2, Y2), (X3, Y3), unknown node to anchor nodes computer simulation to the above algorithm. Simulation
distance d1, d2, d3. According to the formula can be calculated condition: The mobile anchor node reference MICA2 mote;
unknown node coordinates (X, Y). Compared to ordinary Uses outdoor launches the radius 200 to 300m; Deployment
centroid algorithm, 1/ d1, 1/ d2, 1/ d3 is weighted factor. The area is 200*200. The unknown node arranges stochastically;
factor 1/ d1, 1/ d2, 1/ d3 indicates that mobile anchor node with the unknown node is 220. The mobile anchor node has 6 kinds
a shorter distance to unknown nodes has a larger infect its of situations: 9, 12, 16, 20, 25, and 30 positions.
coordinates. We can improve the localization accuracy from
these inner relations. B. Results and Analysis
The simulation uses adaptive weighted centroid localization
algorithm and maximum likelihood estimation method.
Localization accuracy mainly depends on the numbers of the
mobile anchor node broadcasting its positions or the anchor
density. It is very easy for our method to change anchor density
by adjusting the interval time or the moving length of the
mobile anchor node broadcasting its positions or by changing
Figure 6. Scheme of the adaptive weighted centroid localization algorithm the moving interval of spiral line. In comparison with other
methods, this is one of the advantages with our method, and it
Adaptive Weighted Centroid Localization Algorithm does not require additional hardware. Figure 7 and Figure 8
process: show the simulation result. In the figure 7 error of adaptive
weighted centroid localization algorithm is 16.2m and error of
1) The mobile anchor node periodically sends its own maximum likelihood estimation is 24.2m when mobile anchor
information. node is 9. In the figure 8 error of adaptive weighted centroid
localization algorithm is 21.5m and error of maximum
2) Unknown node received information, only records the likelihood estimation is 40.5m when mobile anchor node is 30.
same location of the mobile anchor node average RSSI. As can be seen from the figure adaptive weighted centroid
localization algorithm has better localization accuracy. Has the
3) Unknown node received over threshold m in the obvious superiority, if the anchor density is low. Adaptive
position information then RSSI value in accordance with the Weighted Centroid Localization Algorithm is simple, and no
communication is needed while locating. It does not require
smallest sort of mobile anchor node location .And to establish
additional hardware. The mobile anchor node can be used
the mapping between RSSI value and the distance from many times. So it is very inexpensive.
unknown node to the mobile anchor node. The establishment
of three sets: mobile anchor node_set={a1 , a2 , …, am};
Distance_set={d1 ,d2 , … , d m}; Mobile anchor node
position_set={(X1 ,Y1 ),(X2 ,Y2 ),…,(Xm ,Ym )};

4) RSSI value with the first few large location of mobile

anchor node of the calculation:
a) Based on the preceding analysis, In the mobile
anchor node_set Select RSSI value of large node location then
the composition of the triangle set. This is very important.
Triangle_set={( a1 , a2,a3),( a1,a2 , a4),… ( a1,a3 ,
Figure 7. Average Error
a4),( a1,a3 , a5) … };

5) n location of mobile anchor nodes can be composed of

C_n^3 triangles. The use formula (1) calculates C_n^3

ISSN 1947-5500
(IJCSIS) International Journal of Computer Science and Information Security,
Vol. 8, No. 8, November 2010
[5] Nagpal, R., Shrobe, H., and Bachrach, J.:Organizing a global coordinate
system from local information on an ad hoc sensor network. In IPSN’03,
[6] Niculescu, D., Nath, B.:DV based positioning in ad hoc networks.
Journal of Telecommunication Systems, Vol. 22, No. 4, pp.267-280,
[7] Kim, Y. C., Kim, Y. J., Chang, J. W.:Distributed Grid Scheme using S-
GRID for Location Information Management of a Large Number of
Moving Objects. Journal of Korea Spatial Information System Society,
Vol. 10, No. 4, pp.11-19, 2008.
[8] Lee, Y. K., Jung, Y. J., Ryu, K. H.:Design and Implementation of a
System for Environmental Monitoring Sensor Network. In Proc. Conf.
Figure 8. Maximum Error APWeb/WAIM Workshop on DataBase Management and Application
over Networks, pp. 223-228, 2008.
[9] Hammad, M. A., Aref, W. G., Elmagarmid, A. K.:Stream window join:
V. CONCLUSION AND FUTURE WORKS Tracking moving objects in sensornetwork databases. In SSDBM, 2003.
The Many wireless sensor network applications depend on [10] Niculescu, D., and Nath, B.:Position and orientation in ad hoc networks.
Ad hoc Networks, Vol. 2, No. 2, pp. 133-151, 2002.
nodes being able to accurately determine their locations. This is
the first work to study range-free localization in the presence of
mobility. One of our ideas is that a mobile anchor can improve AUTHORS PROFILE
the localization accuracy and coverage because it can move to
every point of wireless sensor networks. Another factor is that Chang-Woo Song received M.S. degree, from the Inha University, in 2007,
range-free requires no extra hardware or data communication respectively. He is currently working as a Lecturer in the Department of
Computer System Enginnering, affiliated to Inha Technical College. He
and reduces the costs of localization. Our simulation research interest includes ubiquitous/embedded system, computer architecture,
experiments reveal that our method can provide accurate mobile programming.
localization even when memory limits are severe, the seed
density is low, and network transmissions are highly irregular.
Jun-Ling Ma received M.S. degree, from the Inha University, in 2010,
Many issues remain to be explored in future work including respectively. He is doing his research in Wireless Sensor Network. His area of
how to select a moving path to improve the locating interest includes operating systems and object analysis and design.
performance, how to apply this to real-world sensor networks
and how to expend our method to other applications.
Professor Kyung-Yong Chung received the B.S., M.S. and Ph.D. degrees from
the Inha University, Korea, in 2000, 2002, and 2005, respectively. Currently, he
ACKNOWLEDGMENT is a Professor in the School of Computer Information Engineering, Sangji
University, Korea. His research interest includes data mining, HCI, information
“This research was supported by the MKE(The Ministry of retrieval, and sensibility engineering.
Knowledge Economy), Korea, under the ITRC(Information
Technology Research Center) Support program supervised by
the NIPA(National IT industry Promotion Agency)” (NIPA- Professor Jung-Hyun Lee received the B.S., M.S. and Ph.D. degrees from the
2010-C1090-1031-0004). Inha University, Korea, in 1977, 1980 and 1988, respectively, all in Electrical
Engineering. Since 1989, he is a Professor in the School of Computer Science
& Engineering, Inha University. In 1979-1981, he was a researcher at the Korea
REFERENCES Institute of Electronics Technology. In 1984-1989, he was an Associate
Professor at the Kyonggi University. His research interests are in computer
[1] Ssu, K. F., Ou, C. H., and Jiau, H. C.:Localization with mobile anchor architecture, speech recognition, data mining, HCI, information retrieval, and
points in wireless sensor networks. IEEE Trans. on Vehicular sensibility engineering.
Technology, Vol. 54, No. 3, pp. 1187-1197, 2005.
[2] Hu, L., and Evans, D.:Localization for mobile sensor networks. in Proc.
of ACM MobiCom, 2004. Professor Kee-Wook Rim received the B.S., and Ph.D. degrees, from the Inha
[3] He, T., Huang, C., Blum, B. M., Stankovic, J. A., and Abdelzaher, University, in 1987 and 1994; the M.S. degree from Hanyang Univercity,
T.:Range-free localization schemes for large scale sensor networks. in respectively. Since 2000, he is a Professor in the School of Computer Science
Proc. ACM Int. Conf. Mobile Computing Networking (MOBICOM), and Engineering, Sunmoon University. In 1977-1999, he was a Senior
San Diego, CA, pp. 81-95, 2003. Researcher at ETRI and TICOM Development Manager. His research interests
are in Real-time Database Systems, Operating Systems and System.
[4] Niculescu, D., and Nath, B..:DV based positioning in ad hoc networks.
Kluwer J. Telecommun. Syst, Vol. 22, No. 1, pp. 267-280, 2003.

ISSN 1947-5500