2. Therefore, we try to select the relay nodes that can sustain the nodes. At last, all sensor nodes and anchor nodes exist in two-
position of mobile anchor nodes to achieve the above goals. dimensional space.
Our paper has the following key features. First, the moving Under these assumptions, a sensor node hears all the
direction of the anchor node is used to choose the relay nodes. information from the anchor nodes within its communication
It allows the relay nodes to keep the position of the mobile range. In other words, the anchor nodes should exist within a
anchor nodes. Second, the selected relay nodes can send their one-hop radius of a sensor node. But a sensor node can’t
information on the anchor nodes to the sensor nodes within estimate its position if there are no anchor nodes inside its
the communication range. As a result, it reduces energy communication range. So the relay nodes have their
consumption because only the selected relay nodes are positional information from the anchor node inside their one-
activated. hop communication range. The relay nodes can send their
This paper is organized as follows. We discuss some information received from the anchor node to a sensor node.
related works in Section 2 and present an overview and That is, we should choose the best relay nodes to track the
discussion of our method in Section 3. In Section 4, the mobile anchor node and to save energy for communication.
performance of the proposed scheme is simulated and It is important to select the relay node based on the
verified. Finally, we conclude the paper in Section 5. moving direction of the anchor node, to choose relay nodes
according to some conditions, and to activate the least
number of relay nodes. The reason for this statement comes
II. RELATED WORKS from the following perspectives. The first thing is that a
The previous works can be divided into two categories sensor node gets the information of the anchor nodes from
based on their computational methodology: the centralized many relay nodes more than one time in order to locate itself
methods and the distributed methods. Centralized localization because of the loss of noise and wireless communication
techniques require the condition of inter-node ranging and channel characteristics. Hence, it is desirable to choose the
connectivity. In this technique, the positional information is optimal relay node that is capable of sustaining the latest
sent to a central base station for localization [11]. Then, all position of the mobile anchor nodes taking into account its
the nodes receive their positional information from a base moving direction. Secondly, lots of communication time
station. Doherty proposed a semi-definite programming demand greater overhead and delay during communication
approach [12]. In his algorithm, the convex optimization is between two sensor nodes. So, we have to choose relay nodes
used to estimate the positions based on the connectivity according to some conditions which will be discussed in the
constraints provided some nodes have their positions. Shang following section, and they will decrease the total number of
presented a method from mathematical psychology called relay communications between the two nodes. Finally, we
multidimensional scaling [13]. should minimize the number of nodes participating in
The distributed algorithms don’t require a central base communication and activate the least number of relay nodes
station. Localization could be done through the node-to-node to save energy for WSNs. Yet, we need to make sure there are
communication. There are two approaches in the distributed enough relay nodes to ensure that all sensor nodes are located.
localization algorithms: the anchor node-based approach and A. Relay node selection
the anchor node-free approach. Anchor node-based
distributed algorithms take advantage of some nodes called There are many candidates for relay nodes that are within
anchor nodes which know their absolute locations by a GPS the anchor node’s communication range. To select the
or preset placement. The other nodes estimate their positions optimal relay node among those candidates, we consider its
using the positional information provided by the neighboring distance from the anchor node and its proximity from the
anchor nodes [14]. Anchor node-free distributed algorithms anchor node’s moving direction. We use an internal angle
use an unrefined method to localize a sensor node in the between the anchor node’s moving direction and the relay
network. So, applying the refinement algorithm customizes its node’s position in order to determine the proximity between
position to optimize a local error metric [15]. Another the two nodes. The optimal relay node is obtained from the
approach of the anchor node-free distributed method uses a projected distance of a sensor node along the anchor node’s
coordinate system to optimize a network wide metric in a moving direction, which is expressed as follows:
distributed manner [16].
DISTANCE × cos(θ i ) where
(1)
III. OUR PROPOSED SCHEME DISTANCE = ( X A − X si ) 2 + (YA − Ysi ) 2
Our scheme assumes the following four things. First, a
Let us assume that the anchor node is moving to the right
GPS-free sensor node estimates its own position by using the
and three sensor nodes exist near that area. In Fig. 1, the node
relative distance from the anchor nodes with GPS. Also, the
An is an anchor node, and S1, S2, and S3 are the sensor nodes,
anchor node sends the positional information of the sensor
which can be candidate nodes for a relay node. d1, d2, and d3
nodes within its one-hop communication range to the sensor
express the distance from each sensor node to the anchor
nodes using a received signal strength indicator (RSSI).
node and θ1, θ2, and θ3 indicate each internal angle between
Second, the anchor nodes know the position, speed and
each sensor node and the anchor node.
direction of their movements. Third, all sensor nodes are not
static and move very slowly, more slowly than the anchor
775
Authorized licensed use limited to: SRM University. Downloaded on July 19,2010 at 08:13:39 UTC from IEEE Xplore. Restrictions apply.
3. t2~t5 (4)
S2
t14~t20 (7)
t8~t11 (4) S6
An
t1 t2 t3 S3
t4
t5 An t15
S1 t6 t14 t16
An t12 t13 t17 An
~t5 (3) t7 t10 t11 t18 t
t8 t9
S5 19
t20
rs = rt S4
t14~t17 (4)
t8~t13 (6)
Figure 2. Selecting a relay node based on proximity.
B. Localization scheme
Figure 1. Three candidate nodes for a relay node. The anchor node is able to know its position using its
GPS. It can send the positional information of sensor nodes
within its one-hop communication range using RSSI
TABLE I. PROJECTED DISTANCE OF EACH SENSOR NODE technology. The selected relay node with the positional
Projected information notifies its location to other sensor nodes within
Node Distance Internal angle Priority
distance its one-hop communication range. This process is done
S1 3.5 20 1.76 2 repeatedly until a sensor node knows the locations of three
S2 7 40 2.03 1 neighbor relay nodes. After that, the sensor node calculates its
location by the triangulation method using three positional
S3 9 70 1.03 3
coordinates.
In Fig. 3, the sensor node S has the coordination (X, Y)
To decide the optimal relay node, we first calculate the and d1, d2 and d3 denote the Euclidean distances between the
projected distance of each sensor node shown in Table I. unknown nodes and the sensor nodes S1, S2, and S3
According to the values of projected distance, the node S2 is respectively. The locations of the three relay node candidates
chosen as the relay node because it has the highest value of are already known. They are (x1, y1), (x2, y2) and (x3, y3)
projected distance and maintains proximity with the anchor respectively.
node the longest as the anchor node moves. The node S3 has
a larger angle θ3 and a longer distance from the anchor node
than any other node. However, the node S3 has less
possibility of being inside the one-hop range of the anchor
node when the anchor node is moving. Therefore, it is
desirable to choose the node S2 than the node S3, which is far
(x1,y1)
from the anchor node because θ2 is smaller than θ3. S1 (x2,y2)
Fig. 2 shows three examples for how to select a relay node d1
(X,Y) d2 S2
among sensor nodes when an anchor node moves. The first S1
S
example is that when the anchor node is at time slot t3, the
node S1 stays about 3 slots and the node S2 stays over 4 slots d3
within the one-hop radius of an anchor node when it moves to
S3
the right. In a similar fashion, the second example is the time
slot t8. The node S3 has duration of about 4 slots within one- (x3,y3)
hop of the anchor node, and the node S4 has duration of about
6 slots. The third example is for the case of S5 and S6 at the
time slot of t14. The duration of the node S5 to stay within the
one-hop communication range is longer than that of the node Figure 3. Localization of a sensor node using triangulation.
S6. These clarify that choosing a relay node based on the
sensor node with the longest time in proximity to the anchor
node is better than selecting the nearest sensor node from the The distances from each relay node candidate to the
anchor node. unknown node can be calculated as follows [17]:
776
Authorized licensed use limited to: SRM University. Downloaded on July 19,2010 at 08:13:39 UTC from IEEE Xplore. Restrictions apply.
4. positions, speed and direction. All sensor nodes are not static
⎧( X − x1 ) 2 + (Y − y 1 ) 2 = d 1 2
⎪ and move more slowly than the anchor nodes.
⎪ 2 2
⎨( X − x 2 ) + (Y − y 2 ) = d 2
2
(2)
⎪ 2 2 2
⎪( X − x 3 ) + (Y − y 3 ) = d 3
⎩
Solving for X and Y, we get the coordinates of the
position-unknown node by
2 2 2 2 2 2 2 2 2 2 2 2
( y 2 − y1 )(x 2 − x3 + y 2 − y 3 − d 2 + d 3 ) − ( y3 − y 2 )(x1 − x2 + y1 − y 2 − d1 + d 2 )
X=
2 ((x 2 − x1 )( y3 − y 2 ) − ( x3 − x2 )( y 2 − y1 ))
(3)
2 2 2 2 2 2 2 2 2 2 2 2
( x − x )(x − x3 + y 2 − y 3 − d 2 + d 3 ) − ( x3 − x 2 )(x1 − x2 + y1 − y 2 − d1 + d 2 )
Y= 2 1 2
2 ((x3 − x2 )( y 2 − y1 ) − ( x2 − x1 )( y 3 − y 2 ))
C. Node state transition diagram
Fig. 4 shows a sensor node state transition diagram.
Figure 5. The number of rounds when the nodes die.
Fig. 5 shows the number of rounds when a sensor node
dies. The x-axis represents the number of dead sensor nodes,
and the y-axis represents the time in rounds which are
obtained by descending order. We simulate these in two
cases: One is the case of deploying relay nodes in the network
and the other is a scheme of not having relay nodes. A sensor
node in the case of deploying relay nodes has a longer
lifetime than the other. Our scheme of using the relay nodes
has more desirable energy expenditure than the method of not
Figure 4. State transition diagram for a sensor node. using the relay nodes.
Fig. 6 shows how long the relay nodes and the sensor
nodes stay in one-hop communication range from an anchor
There are three states; the sleep state, the discovery state node. As shown in the figure, the duration of selected relay
and the active state. We define sleep state as a state where nodes to stay within one-hop communication range from an
communication and sensing are turned off to reduce energy anchor node is mostly larger than that of the sensor nodes.
consumption. Discovery state is defined as receiving signals This makes the proposed scheme spend less power than the
from other sensor nodes while sensing. If a sensor node is in scheme without relay nodes.
active state, it has four roles; monitoring the anchor node,
broadcasting its information to other neighboring sensor
nodes, determining whether a node should become a relay
node and estimating its own location as well.
IV. SIMULATION RESULTS
We consider a 200 x 200 network configuration with 200
nodes throughout the simulation. Some important simulation
parameters are listed in Table II.
TABLE II. SIMULATION PARAMETERS.
Parameter Value Parameter Value
Transmission
Number of AN 5 720mW
energy
Speed of Receiving
2 m/s 369mW
AN(mean) energy
Transmission
20m Initial energy 500kW
range Figure 6.The duration of a node staying within one-hop radius of an anchor
In our simulations, each node moves at a constant unit node.
speed in random directions, and the anchor nodes know their
777
Authorized licensed use limited to: SRM University. Downloaded on July 19,2010 at 08:13:39 UTC from IEEE Xplore. Restrictions apply.
5. [1] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci,
“Wireless Sensor Networks: A Survey,” Computer Networks, vol. 38,
no. 4, pp. 393-422, 2002.
[2] B. Karp and H. Kung, “GPSR: Greedy Perimeter Stateless Routing for
Wireless Networks,” Proc. ACM MobiCom ’00, pp. 243-254, Aug.
2000.
[3] Y.B. Ko and N.H. Vaidya, “Location-Aided Routing (LAR) in Mobile
Ad Hoc Networks,” Wireless Networks, vol. 6, Issue 4, pp. 307-321,
2000.
[4] A. Howard, M. Mataric, and G. Sukhatme, “Localization for Mobile
Robot Teams Using Maximum Likelihood Estimation,” Proc.
IEEE/RSJ Int’l Conf. Intelligent Robots and Systems, Oct. 2002.
[5] N. Sundaram and P. Ramanathan, “Connectivity Based Location
Estimation Scheme for Wireless Ad Hoc Networks,” Proc. IEEE
Global Telecomm. Conf. (Globecom ’02), vol. 1, pp. 143-147, Nov.
2002.
[6] Qingjiang Shi, Chen He, Lingge Jiang and Jun Luo, “Sensor Network
Localization via Non differentiable Optimization,” Proc. IEEE
Figure 7. The time needed to locate each node using three relay nodes’ Telecomm. Conf. (Globecom ’08), pp.1-5, Dec. 2008.
positions.
[7] B. Hull, V. Bychkovsky, Y. Zhang, K. Chen, M. Goraczko, A. K. Miu,
E. Shih, H. Balakrishnan, and S. Madden. “Car-Tel: A Distributed
Fig. 7 shows the time needed to find out the location of Mobile Sensor Computing System,” Proc. 4th ACM SenSys, pp. 125-
each node using three relay nodes’ positions. We find that 138, Nov. 2006.
75% of all nodes successfully estimate their position within [8] Qiao-ling Du, Zhi-hong Qian, Hong Jiang, Shu-xun Wang,
200 rounds, and 13% of the nodes find their positions “Localization of Anchor Nodes for Wireless Sensor Networks,” Proc.
New Technologies, Mobility and Security, NTMS '08, pp.1-5, Nov.
between 200 and 500 rounds. The remaining 12% of the 2008.
nodes need more than 500 rounds to find their positions. In
[9] Liqiang Zhang, Qiang Cheng, Yingge Wang, Sherali Zeadally, “A
general, most of the nodes can successfully find their Novel Distributed Sensor Positioning System Using the Dual of Target
positions within 4% of their entire lifetime. Tracking,” IEEE TRANSACTIONS ON COMPUTERS, vol. 57, no. 2,
pp. 246-260, Feb. 2008.
V. CONCLUSIONS [10] Mark R. Morelande, Bill Moran and Marcus Brazil, “Bayesian node
In this paper, we propose a self localization scheme which localisation in wireless sensor networks,” Proc. Acoustics, Speech and
Signal Processing, ICASSP 2008, pp.2545-2548, 2008.
deploys the relay nodes properly to maintain estimations with
[11] Ganggang Yu, Fengqi Yu, “A Localization Algorithm for Mobile
high accuracy and to reduce the energy consumption for Wireless Sensor Networks,” Proc. IEEE International Conference on
MWSN. The main idea in our scheme is to select a relay node Integration Technology, pp.623-627, Mar. 2007.
as the one which maintains proximity with an anchor node [12] L. Doherty, K. Pister, and L.-E. Ghaoui, “Convex position estimation
longest along its moving direction. An internal angle between in wireless sensor works,” Proc. IEEE Infocom, vol. 3, pp.1655-1633,
the anchor node’s direction and the position of a sensor node Apr. 2001.
is used to determine the proximity between the two nodes. [13] Y. Shang, W. Ruml, Y. Zhang, and M. Fromherz, “Localization from
Our scheme clearly enhances the accuracy of locating Mere Connectivity,” Proc. 4th ACM international symposium on
Mobile ad hoc networking & computing, MobiHoc 2003, pp. 201-212,
mobile nodes and reduces the total energy consumption for 2003.
MWSN. Finally, we found that adopting the relay node is
[14] T. He, C. Huang, B. Blum, J. Stankovic and T. Abdelzaher, “Range-
beneficial to accurate locating in MWSN but needs to be free localization schemes in large scale sensor networks,” Proc. 9th
improved for various environment or applications in this Annual International Conference on Mobile Computing and
initial study. As part of our future work we will extend our Networking, pp. 81-95, 2003.
algorithm to improve the accuracy by diversifying the [15] N. Priyantha, H. Balakrishnan, E. Demaine, and S. Teller, “Anchor-
conditions of the relay node qualification for a self free distributed localization in sensor networks,” Proc. 1st
localization scheme in MWSN. International Conference on Embedded Network Sensor Systems, pp.
340-341, Nov. 2003.
ACKNOWLEDGMENT [16] S. Capkun, M. Hamdi and J.-P. Hubaux, “GPS-free positioning in
mobile ad-hoc networks,” Proc. 34th Annual Hawaii International
This research was supported by Handong Global Conference on System Sciences, vol. 9, pp. 3481-3490, Jan. 2001.
University Research Grants 2008. [17] X. Yu, K. Niyogi, S. Mehrotra, and N. Venkatasubramanian,
“Adaptive Target Tracking in Sensor Networks,” Proc.
Communication Networks and Distributed Systems Modeling and
REFERENCES Simulation Conference, Jan. 2004.
∗
Correspondent Author
778
Authorized licensed use limited to: SRM University. Downloaded on July 19,2010 at 08:13:39 UTC from IEEE Xplore. Restrictions apply.