Wireless Sensor Network, 2011, 3, 1-9
doi: 10.4236/wsn.2011.31001 Published Online January 2011 (http://www.SciRP.org/journal/wsn)
Distributed Sensor Network Based on RFID System
for Localization of Multiple Mobile Agents*
Byoung-Suk Choi1, Joon-Woo Lee1, Ju-Jang Lee1, Kyoung-Tai k Park2
1Department of Electrical Engineering, KAIST, Daejeon, Korea (South)
2Korea Institute of Machinery and Materials (KIMM), Daejeon, Korea (South)
Received December 10, 2010; revised December 29, 2010; accepted January 13, 2011
Abstract
This paper presents a distributed wireless sensor network for multiple mobile agents localization. Localiza-
tion of mobile agents, such as mobile robots, humans, and moving objects, in an indoor space is essential for
robot-robot interaction (RRI) and human-robot interaction (HRI). The standard localization system, which is
based on sensors installed in the robot body, is not suitable for multiple agents. Therefore, the concept of
sensor network, which uses wireless sensors distributed in a specified space, is used in this study. By ana-
lyzing related studies, two solutions are proposed for the localization of mobile agents including humans: a
new hardware system and a new software algorithm. The first solution focuses on the architectural design of
the wireless sensor network for multiple agent localization. A passive RFID system is used, and then the ar-
chitecture of the sensor network is adapted to suit the target system. The second solution centers on a locali-
zation algorithm based on the sensor network. The proposed localization algorithm improves the accuracy in
the multiple agent localization system. The algorithm uses the displacement conditions of the mobile agents
and the recognition changes between the RFID tags and RFID reader. Through experiments using a real
platform, the usefulness of the proposed system is verified.
Keywords: Multiple Robot Localization, Distributed Sensor Network, RFID System
1. Introduction
Recently, ubiquitous computing [1,2], where networks
can be accessed anywhere and anytime, has garnered
increasing attention. The infrastructure to build a sensor
network has been developed rapidly as a result of
advances in semiconductors, MicroElectroMechanical
Systems (MEMS), and communications technologies
[3,4]. Sensor networks are a basic element of ubiquitous
computing environments and have features such as low
power, efficiency, low cost, robustness, flexibility, and
distribution. Sensor networks with these features have
been applied to several areas such as intruder surveill-
ance, position tracking, and motion monitoring in indoor
spaces [5-8].
In the near future, multiple mobile agents are expected
to co-exist and interact with each other in indoor spaces.
The mobile agents include humans and objects, as well
as robots. The following two examples are considered to
be the types of interactions with mobile agents that will
increase: robots will interact with other robots and
perform cooperative functions; robots will offer services
to humans and track objects. To facilitate smooth and
safe interactions, more accurate and robust localization
methods for these mobile agents are needed. That is, the
states of all mobile agents within a specified space
should be detectable.
Classic localization algorithms use sensors installed in
mobile agents, and the positions of the mobile agents are
estimated by analyzing the data obtained from these
sensors, such as cameras, encoders, and ultrasonic sensors
[9]. If multiple mobile agents are located closely in the
same space, however, interference between the sensors
data obtained is generated. Therefore, accurate data for
localization cannot be obtained from these sensors. Also,
the use of sensors is limited according to the features of
the mobile agents. For example, human agents do not
have general sensors such encoders or cameras for
localization. Thus, the classic localization system is not
suitable for multiple mobile agent localization that
*This work is supported by the Technology Innovation Program funde
d
by the Ministry of Knowledge Economy (MKE, Korea).

B. S. CHOI ET AL.
2
includes human agents. Recently, studies on mobile
agent localization have been undertaken using wireless
sensor networks: the sensors distributed in a specified
space are used in the wireless sensor network system.
Therefore, the problems found in the classic localization
system are examined with reference to wireless sensor
networks as a possible solution.
In this paper, a distributed wireless sensor network is
used to effectively estimate the position of multiple
mobile agents. The sensor network is organized using
passive Radio Frequency IDentification (RFID) technology
in order to improve efficiency and solve interference
problems in the multiple agent localization process.
However, there are some unknown factors in the locali-
zation process; therefore, the purpose of this study is to
reduce uncertainty and to improve accuracy in the
multiple agent localization system.
The remainder of the paper is organized as follows.
Section 2 states the related works on the localization
system using sensor network and research objectives.
Section 3 presents the architecture of the sensor network.
Section 4 presents the proposed algorithm for effective
localization in a sensor network environment. The ex-
periment results are presented in Section 5. Finally, con-
clusions are drawn in Section 6.
2. Related Works
Many researchers have attempted to improve the
accuracy of localization systems based on sensor net-
works.
Zhang et al. [10] proposed a distributed sensor
network using infrared sensors: the infrared sensors were
suspended from the ceiling and human agents wore
infrared transmitters. The transmitters spread the signal
with unique data for identification, and the positions of
the human agents were detected. Lee and Kim [11,12]
proposed an active beacon system to fuse the RF signals
and ultrasonic sensors. Their system calculated the dis-
tance between the mobile agents and the node using the
time of arrival (TOA) of the transmitted signal. Villa-
dangos et al. [13] proposed a sensor network composed
of ultrasonic sensors. The positions of the robot agents
were estimated using a combination of the distance data
obtained from the ultrasonic sensors. Han et al. [14]
proposed a localization system for mobile agents using
passive RFID tags that were arranged on the floor of an
indoor space.
Mehta et al. [15] used a wireless camera sensor
network platform. Data was obtained using grayscale
images from the camera sensor nodes over a wireless
channel: they proposed simple arithmetic calculations,
and implemented and evaluated the physical camera
sensor network platform. Shenoy et al. [16] addressed
the localization algorithm for wireless sensors in a hybrid
sensor network. The hybrid sensor network consisted of
a large number of static sensors and GPS signals. The
static sensors were able to estimate the robots’ positions
based on the messages received. Their proposed algorithm
worked simultaneously for both the sensor node locali-
zation and mobile robot navigation. Sheng et al. [17]
proposed a new sensor network architecture consisting of
an active sensor network. A potential-based robot area
partition algorithm and a distributed localization algo-
rithm were also developed.
The previous works are limited by issues of mobile
robot localization in indoor spaces. For accurate locali-
zation, therefore, the existing research might fuse sensor
networks and general sensors such as ultrasonic sensors
or cameras installed on the robot body. In Section 1, HRI
and RRI have been mentioned. The localization for the
human agents and moving objects should also be
addressed. As shown in the previous studies, robots can
have extra sensors for localization on their body, except
in sensor network systems. However, sensors such as
encoders, ultrasonic sensors, and cameras cannot be
installed in human agents and objects for localization.
The position of the human agents should be estimated
using a distributed sensor network. Therefore, a locali-
zation system based on sensor network for the several
types of mobile agents (robots, humans, and objects)
must be addressed. To date, previous studies have not
considered these issues.
In this paper, the problems shown in existing studies
are considered and possible solutions are proposed. The
sensor network using an RFID system is organized for
effective localization of multiple and different types of
mobile agents. Even If they coexist, localization can be
processed irrespective of this. The architecture of the
sensor network is presented in Section 3. Furthermore,
an algorithm to improve accuracy in the localization
process is proposed based on the designed sensor
network. Because the algorithm does not require extra
sensors, it can be applied to human agents, moving
objects, and robots alike. The algorithm is explained in
Section 4.
3. Architectural Design of Sensor Network
A distributed sensor network system has an architecture
where several sensor agents are connected and net-
worked with each other, as shown in Figure 1. In a sen-
sor network system, the data obtained from each sensor
is transferred through a wired or wireless network. In this
paper, a distributed wireless sensor network is applied to
the mobile agent localization as shown in Figure 2. The
proposed system is composed of nodes, mobile agents,

B. S. CHOI ET AL.
3
Figure 1. General concept and architecture of sensor network
system that consists of several small sensors agents.
Figure 2. Application of sensor network for the localization
of multiple mobile agents including humans, robots, and
moving objects.
and the main network. The position of multiple mobile
agents is estimated by the nodes, which consist of small
sensors arranged in the specified space. One node covers
a local assignment space and nodes are placed so that
they overlap slightly; thus, the total coverage by the
nodes is equates with the overall global space. The mo-
bile agents are all human agents, objects, or mobile ro-
bots that exist in the specified space. When a mobile
agent exists in a certain local space, the position of the
agent is estimated by the node that covers that local
space. If several agents are located near the same node,
the positions of the agents are obtained independently
using their own IDs. The main network with network
servers stores and manages the IDs and states of every
agent. All agents have knowledge of the position of each
agent in the network.
The distributed wireless sensor network is organized
using RFID technology. An RFID system has several
features and functions that depend on the frequency band,
nature of the power source, tags, etc [18]. A passive RFID
system with an operation frequency of 13.56 MHz is used
to obtain the following practical features [19,20]. Passive
RFID tags use induction power through RF waves from
RFID readers. Because an external power source is not
used, semi-permanent use without power replacements is
possible after the initial installation. The passive RFID
system recognizes passive RFID tags without contact.
Damage or incorrect readings caused by obstacles do not
occur. Because the ID and position data are stored in the
RFID tags, the positions of several agents can be ad-
dressed simultaneously without confusion.
Figure 3 shows the concept of the sensor network
based on the RFID system for localization of multiple
mobile agents. Passive RFID tags are arranged with a
constant gap on the floor and a pre-stored coordinated
value. That is, the sensor network node is a passive RFID
tag in this study. The mobile agents have independent
RFID readers and antennas to communicate with the
RFID tags. Mobile robots have antennas installed on the
bottom of their body and human agents install the
antennas on the bottom of their shoes. The antenna, which
is connected to a reader, radiates electromagnetic waves
with a finite range in the direction of the floor. RFID tags
are activated using electromagnetic waves and can then
transmit their coordinate values to the RFID reader. The
packet with the obtained position data and ID of the
mobile robots is uploaded to the main network. Because
passive RFID tags do not use a separate external power
source, the position data yielded by the RFID system is
indirectly uploaded through the RFID reader installed in
the robot agent. As shown in Figure 4, the mobile agents
have RFID readers and antennas, and the data obtained
from the RFID tags is transferred to the main network.
The protocol used in the network communication is
organized as shown in Table 1. Through this network,
each robot agent and user knows the position of every
robot agent at any time.
4. Algorithm for Localization Based on
Sensor Network
As mentioned in the previous section, the RFID reader is
Figure 3. Concept of the sensor network based on the RFID
system for localization of multiple mobile agents: distributed
RFID tags in a space.

B. S. CHOI ET AL.
4
Figure 4. Localization of mobile agents using the RFID
system.
Table 1. Communication protocols between the RFID
reader and network server.
ABCDEFGHIJKLMNO
A: Mobile agent ID 0: robot-1, 1: robot-2,2: human
B: Detection by any Node (On/Off) 0: detection, 1: not detected
C: empty/dummy
D: First Activated Node ID integer
E: x-coordinate of Node integer
F: y-coordinate of Node integer
G: Second Activated Node ID integer
H: x-coordinate of Node integer
I: y-coordinate of Node integer
J: Third Activated Node ID integer
K: x-coordinate of Node integer
L: y-coordinate of Node integer
M: Fourth Activated Node ID integer
N: x-coordinate of Node integer
O: y-coordinate of Node integer
installed in mobile agents and passive RFID tags are
arranged on the floor. When a mobile agent approaches a
specified tag, the coordinated value stored in the tag is
read by the RFID reader in the mobile agent. That is, the
mobile agent is located around a coordinate value stored
in the tag. However, an accurate position value of the
mobile agent cannot be obtained. The distance between
the node (i.e. RFID tag) and the mobile agent is
unknown due to the technical limitations of passive
RFID systems. Only limited position information can be
obtained: that is, the mobile agent is located near a
certain node. Previous works have used other sensor
systems with a sensor network to improve the accuracy
of localization. While the sensors such as encoders and
cameras can be installed on the robot body, human
agents do not have these sensors for localization. Thus,
for successful localization of the mobile agents including
human agents, a scheme that uses only the distributed
sensor network must be considered.
Therefore, a localization algorithm that uses the
motion of the mobile agents and recognizes changes in
the RFID tags under a sensor network environment is
proposed. The approach using the motion of the mobile
agents is as follows. If the mobile agents move in the
space, displacement for a constant time is generated
according to the agents’ velocity. If the velocity of
agents is controlled, the displacement will be within a
constant range. Therefore, the range of the next
position from the initial position of the agents can be
predicted and estimated. The approach using the
recognition change is as follows. In the RFID system,
communication and recognition between the RFID
reader and tags are always conducted. However, the
maximum distance for communication and recognition
is fixed. If the mobile agent equipped with the RFID
reader moves, the distance between the RFID reader
and tag varies continuously. Therefore, two cases-case
I where the RFID reader recognizes the tags and case II
where the RFID reader does not recognize the tags-
occur repetitively. Using the recognition change in the
RFID system, the position of mobile agents can be
estimated.
It is assumed that mobile agents move with a limited
velocity in a specified indoor space. As shown in Figure
5, the RFID tags are arranged regularly and defined as
the node1, node2, and node3. The coordinate values stored
in each node are CN.1, CN.2 and CN.3., respectively.
...
(, ),1,2,3
NkNk Nk
Cxyk

 (1)
The nodes cover a constant territory. If a mobile agent
exists in the recognition territory (RT), the position of the
mobile agent is estimated using the corresponding node.
The territory recognized by nodek is represented as RTk.
For example, RT1 is approximated as a circle with
constant radius, γ, and the center is node1 [21].
Figure 5. Localization for the movement of a mobile agent
from RT1 to RT2.

B. S. CHOI ET AL.
5
222
.1 .1
()( )
NN
xx yy

 (2)
Note that a mobile agent is always recognized by
node1 within the RT1; however, the recognition rate for a
mobile agent is low near the boundary of RT1. That is, the
reader communicates with tags in the recognition
territory, and the communication is unstable at the
boundary of the recognition territory. Eventually, the
communication is prevented due to the mobile agent
moving out of the recognition territory. However, two
RFID tags simultaneously communicate with the RFID
reader in the overlap region of RT1 and RT2. A localization
algorithm is proposed using the communication state
between the tags and the reader. The algorithm is
explained below for the mobile agent movement from
RT1 to RT2, as shown in Figure 5.
Step 1.
Assume that the mobile agent is detected by node1. It
can be considered that the mobile agent is located within
RT1. However, the exact position within the RT1 cannot
be known. Therefore, it is assumed that there are some
points that can be expected to contain the current
position of mobile agent in RT1. As shown in Figure 6,
m points are scattered in this area. Each point has an x-y
coordinate value in a 2D plane. The points are represented
as follows:
11
(, ),1~
ss
ii
Exyi m
(3)
The number of points, m, is subject to the resolution of
the localization and the dimension of RT1. The point,
1
s
i
E, satisfies the following condition in Step 1.
11
.1 .2
|| ||
ss
Ni Ni
CE ANDCE

   (4)
The probability that each 1
s
i
E is the real position of a
mobile agent is the same. If the real position of the
mobile agent is PM, the sum of the position estimation
errors for 1
s
i
E is represented as follows:
11
1
||
m
s
s
Mi
i
ePE


 (5)
11
(, ),1~
ss
ii
Exyi m
1
ˆ()( )1~
s
Mi
P
kAvgE forim
11
.1 .2
||()||
ss
Ni Ni
CE ANDCE

  
Figure 6. Shape of the recognition territory of node1 in
Step 1.
Because the weight of 1
s
i
E for the real position of the
agent is the same, the average of 1
s
i
E minimizes the
error. The estimated position of the mobile agent,
ˆ()
M
Pk
, at time k is determined by the average value.
11
11
ˆ()
ˆ()/ /
ˆ()
T
mm
Mss
Mii
ii
M
xk
Pkx my m
yk 










 (6)
Step 2.
The mobile agent moves toward node2 within RT1. The
mobile agent is continuously detected by node1. The
recognition change, where node2 detects the mobile
agent, is generated. In this step, it can be considered that
the mobile agent can be located at the boundary of RT2 as
shown in Figure 7. An arbitrary region along the
boundary of recognition territory is set; at some points
along this boundary, 2
s
i
E, it can be expected as that the
position of the mobile agent is selected within an
arbitrary region using the same method as explained in
Step 1. The points, 2
s
i
E, satisfy the following conditions
in Step 2.
2
.2 1
()| |(),
s
iN
ECwithinRT     (7)
The probability that 2
s
i
E is the real position of
mobile agent is the same as in Step 1.
The estimated position of the mobile agent at time k is
determined by the average of 2
s
i
E.
22
11
ˆ()
ˆ()/ /
ˆ()
T
mm
Mss
Mii
ii
M
xk
Pkx my m
yk 










 (8)
The mobile agent has a limited maximum velocity.
Therefore, the maximum displacement of the mobile
agent is calculated during the movement time from Step
1 to Step 2.
|2 1|max21max
()
step stepssM
dvTTforvv


  (9)
22
(, ) ,1~
ss
ii
E
xy im
2
.2
()| |()
s
iN
EC

  
2
ˆ()( )1~
s
Mi
P
kAvgE forim
121
|2 1|
11
:,| |
()( ),1~
sss
iiistep step
ss
Mi i
Eremovedif EEd
P
kAvgEfor remaining Eimr




Figure 7. Shape of the recognition territory of the node1 and
node2 in Step 2.

B. S. CHOI ET AL.
6
The distances from 1
s
i
E to 2
s
i
E are calculated; then,
the 1
s
i
E values that are not satisfied with the maximum
displacement are removed.
121
|2 1|
:,| |
sss
iiistep step
Eremovedunless EEd  (10)
The position of the mobile agent is estimated again
using the remaining 1
s
i
E.
11
ˆ()(),1, ,
ss
Mi i
PkAvgEfor remainingEimr
(11)
Step 3.
If a mobile agent is stably recognized by node1 and
node2, it is considered that the mobile agent is located in
the RT1 and RT2 overlap region as shown in Figure 8.
Next, the points, 3
s
i
E, where the position of the mobile
agent can be expected are determined. Point 3
s
i
E
satisfies the following conditions in Step 3.
33
.1 .2
||,||
ss
iN iN
ECand EC

  
(12)
The estimated position of the mobile agent at time k is
determined by the average of 3
s
i
E. The distance from
2
s
i
E to 3
s
i
E is calculated, and the 2
s
i
E values that are
not satisfied with the maximum displacement are then
removed.
232
|3 2|
:,| |
sss
iiistep step
Eremovedunless EEd  (13)
The position of the mobile agent is estimated again
using the remaining 2
s
i
E.
22
ˆ()(),1, ,
ss
Mi i
PkAvgEfor remaining Eimr
(14)
Step 4.
As this step is the same as that of Step 2 albeit with RT2
and RT3, the recognition change by node1 within RT2 is
3
ˆ()( )1~
s
Mi
P
kAvgE forim
232
|3 2|
22
:,| |
()( ), 1~
sss
iiistep step
ss
Mi i
Eremovedif EEd
P
kAvgEfor remainingEimr



33
(, ) ,1~
ss
ii
Exyi m
33
.1 .2
|| ||
ss
iN iN
EC ANDEC

  
Figure 8. Shape of the recognition territory by the node1
and node2 in Step 3.
generated. Instead, the mobile agent is continuously de-
tected by node2. As shown in Figure 9, it can be as-
sumed the mobile agent is located in the boundary of RT1.
An arbitrary region is set and some points, 4
s
i
E, are se-
lected as in the previous step. Point 4
s
i
E satisfies the
following conditions in Step 4.




4
.1 2
||,
s
iN
EC withinRT     (15)
The estimated position of the mobile agent is
calculated as in Step1 to Step 3.
33
ˆ()(),1, ,
ss
Mi i
PkAvgEfor remainingEimr


(16)
According to the movement of the mobile robot, the
recognition change of the mobile agent by the node is
generated. The position of the mobile agent is continuously
estimated using the aforementioned steps.
5. Experiment and Result
The proposed sensor network for localization is applied
in a real system. The positions of the multiple mobile
agents are obtained using the sensor network including
the algorithm introduced in Section 4.
Figure 10 shows the RFID system and mobile robots
used in this experiment. The RFID system, KISR300H,
used for the sensor network is a passive RFID system
and it operates at a frequency of 13.56 MHz. As shown
in Figure 10(a), the passive RFID tags fabricated from
epoxy have dimensions of 0.03 m × 0.03 m. The RFID
reader and antenna are installed in the mobile robots and
human, as shown in Figures 10(b) and 10(c). The size of
the robots’ antennas is 0.15 m × 0.15 m and the size of
human agent’s antenna is 0.07 m × 0.07 m. The commu-
nication between the RFID reader and network server is
conducted using a wireless LAN. Figure 10(c) shows the
mobile agents: mobile robot 1 and mobile robot 2. Both
robots have two-wheel differential drive and RFID read-
44
(, ) ,1~
ss
ii
E
xy im
4
.1
()| |()
s
iN
EC



  
4
ˆ()()1~
s
Mi
P
kAvgE forim
343
|4 3|
33
:,| |
()( ),1~
sss
iiistepstep
ss
Mi i
EremovedifEEd
P
kAvgEforremaining Eimr




Figure 9. Shape of the recognition territory by node1 and
node2 in Step 4.

B. S. CHOI ET AL.
7
(a) (b)
(c)
Figure 10. Experimental environment: (a) RFID reader,
antenna and tags; (b) human agent with RFID reader and
antenna; (c) two mobile robots in the wireless sensor
network using RFID tags.
ers. The antennas (connected to the RFID reader) are
installed on the bottom of body. The human agent has
also a small portable RFID reader and antenna as shown
in Figure 10(b).
The mobile agents are driven as follows. The RFID
tags are arranged on the floor and the gap between the
tags is 0.5 m, as shown in Figure 10(c). Figure 11(a)
shows the driving path of the mobile agents in the speci-
fied indoor space. In order to prevent the mobile agents
colliding, the driving paths were predefined. The linear
velocity of each agent during the movement along the
path is represented in Figure 11(b). The maximum ve-
locity of the mobile robots is limited to 25 cm/s. The
human agent moves with the maximum instantaneous
velocity, 8 cm/s, and steps forward one step at a time.
The sensor network estimates the position of the moving
mobile agents using the RFID system. The position of
the mobile agents is estimated during approximately 60 s,
and the sampling period of the position measurement is
0.25 s.
(a)
(b)
Figure 11. The predefined path and linear velocity of the
mobile agents: (a) the path of all agents; (b) the linear
velocity of mobile robot 1, mobile robot 2, and the human
agent.
The error between the real position value and the
measured value of the mobile agents is defined as the
performance index. Figure 12 represents the position
error of the mobile agents measured using the proposed
algorithm for the driving time from the starting point to
the finishing point. Figures 12(a) and 12(b) are the
position errors of mobile robot 1 and mobile robot 2,
respectively. The average value of the position error
during driving was 12.06 cm and 12.62 cm for mobile
robot 1 and mobile robot 2, respectively. Figure 12(c)
shows the position error for the human agent’s motion.
Because the human agent repeats a stopping and starting/
moving motion, the human agent’s error pattern is
different to that of the mobile robots. The average value
of error is 14.24 cm for the human agent.
The proposed algorithm from Section 4 and an RFID-
base localization scheme introduced in a previous work
were used to evaluate the utility and advantages of the

B. S. CHOI ET AL.
8
(a)
(b)
(c)
Figure 12. Error generated in the localization process based
on the proposed sensor network using the proposed
algorithm: (a) error for mobile robot 1; (b) error for mobile
2; (c) error for the human agent.
Table 2. Comparison for experimental result between
proposed algorithm and previous algorithm.
Average value of error
(proposed algorithm)
Average value of error
(previous algorithm in [14])
Mobile
robot 1 12.06 cm Mobile robot 1 23.16 cm
Mobile
robot 2 12.62 cm Mobile robot 2 25.81 cm
Human 14.24 cm
algorithm proposed in this study. The results were then
compared for each algorithm. In the RFID-based sensor
network, the distribution of RFID tags affects the
accuracy of localization. If the number of RFID tags is
large, the accuracy generally rises.
The previous method proposed in [14] only depends
on the data of the node (i.e. RFID tag) that is close to
mobile agents. If a mobile agent is near the node, the
position of the mobile agent is estimated using a
coordinated value stored in the node. If the mobile agent
is not located near the node, however, it is difficult to
obtain a coordinated value from the node. When a small
number of RFID tags is used, the accuracy of the
previous algorithm [14] is drastically reduced.
The method proposed in this study uses the
displacement of the mobile agent and recognition of the
change between the RFID reader and tags, and can reduce
uncertainty in localization systems based on a sensor
network. That is, the proposed system is less affected by
the distribution of RFID tags than the previous algorithm.
The position of the mobile agents is estimated using
factors obtained by their movement. If the mobile agents
are not located near the node, the position of the mobile
agents is indirectly calculated using other factors. Even
when a small number of RFID tags is used, a reasonable
level of accuracy can be maintained.
For the same density distribution of RFID tags, the
performance of the previous algorithm and the proposed
algorithm is compared. The gap between RFID tags is
uniformly 0.4 m. The experiments were conducted with
two mobile robots. The results from the experiments in
this study are listed in Table 2. Through the experiments,
it was verified that the performance of the proposed
sensor network using the proposed algorithm exceeds
that of previous research.
6. Conclusion
This work studied the localization of multiple mobile
agents using a distributed sensor network. The distri-
buted sensor network that has been used for the locali-
zation of mobile agents in previous studies had limited
functions. This paper presented a sensor network and
localization algorithm that can be applied to multiple
mobile agents and the distributed wireless sensor net-
work was organized using RFID technology. This con-
cept addressed issues such as sensor interference and
position estimation confusion among multiple agents.
Furthermore, an algorithm that estimates more accurate
position of agents was proposed using the recognition
change between the RFID tag and RFID reader. This
algorithm can be applied to many types of agents includ-
ing humans.
In the future, the system will be organized for interac-
tion and motion planning among agents using the pro-
posed localization scheme based on a sensor network.
7. References
[1] D. Crasto, A. Kale and C. Jaynes, “The Smart Bookshelf:
A Study of Camera Projector Scene Augmentation of an
Everyday Environment,” Proceedings of the 7th IEEE
Workshop on Application of Computer Vision, Colorado,
January 2005, pp. 218-225.
[2] O. Lazaro, A. Gonzalez, L. Aginako, T. Hof, F. Filali, R.
Atkinson, S. Maza, P. Vaquero, B. Molina, J. O. Flaherty,
and R. Mazza, “Multinet: Enabler for Next Generation
Pervasive Wireless Services,” Proceedings of the 16th
IST Mobile and Wireless Communication Summit,
Budapest, Hungary, July 2007, pp. 1-5.
[3] Y. Wang, Y. Wang, X. Qi, L. Xu, J. Chen and G. Wang,
“L3SN: A Level-Based, Large-Scale, Longevous Sensor
Network System for Agriculture Information Monitor-

B. S. CHOI ET AL.
9
ing,” Wireless Sensor Network, Vol. 2, No. 9, 2010, pp.
655-660. doi:10.4236/wsn.2010.29078
[4] X. Zhao, K. Mao, S. Cai and Q. Chen, “Data-Centric
Routing Mechanism Using Hash-Value in Wireless Sen-
sor Network,” Wireless Sensor Network, Vol. 2, No. 9,
2010, pp. 703-709. doi:10.4236/wsn.2010.29086
[5] A. Boukerche and S. Samarah, “In-Network Data Reduc-
tion and Coverage-Based Mechanisms for Gene- rating
Association Rules in Wireless Sensor Networks,” IEEE
Transactions on Vehicular Technology, Vol. 58, No. 8,
October 2009, pp. 4426-4438. doi:10.1109/TVT.2009.
2020802
[6] S. Yao, J. Tan and H. Pan, “A Sensing and Robot Navi-
gation of Hybrid Sensor Network,” Wireless Sensor
Network, Vol. 2, No. 4, 2010, pp. 267-273. doi:10.4236/
wsn.2010.24037
[7] Z. Liang and S. Zhu, “Cooperative Distributed Sensors
for Mobile Robot Localization,” Wireless Sensor Network,
Vol. 2, No. 5, 2010, pp. 347-357. doi:10.4236/wsn.2010.
24046
[8] B.-S. Choi, J.-W. Lee and J.-J. Lee, “Target-following
Scheme for Mobile Robots Based on the Virtual
Sub-Goal Points,” Information — An International Inter-
disciplinary Journal, Vol. 13, No. 6, November 2010, pp.
1929-1938.
[9] B.-S. Choi and J.-J. Lee, “Localization of a Mobile Robot
Based on an Ultrasonic Sensor Using Dynamic Obstacles,”
International Journal of Artificial Life and Robotics, Vol.
12, No. 1-2, March 2008, pp. 280-283. doi:10.1007/s100
15-007-0482-4
[10] Z. Zhang, X. Gao, J. Biswas and J. K. Wu, “Moving
Targets Detection and Localization in Passive Infrared
Sensor Networks,” Proceedings of the 10th International
Conference on Information Fusion, Quebec, July 2007,
pp. 1-6. doi:10.1109/ICIF.2007.4408178
[11] T. Lee, J. Shirr and D. D. Cho, “Position Estimation for
Mobile Robot Using In-Plane 3-Axis IMU and Active
Beacon,” Proceedings of IEEE International Symposium
on Industrial Electronics, Seoul, July 2009, pp. 1956-
1961.
[12] B. Kim and J.-S Choi, “Active Beacon System with the
Fast Processing Architecture for Indoor Localization,”
Proceedings of IEEE Conference on Emerging Technolo-
gies and Factory Automation, Patras, September 2007, pp.
892-895.
[13] J. M. Villadangos, J. Urena, M. Mazo, A. Hernandez, F.
Alvarez, J. J. Garcia, C. D. Marziani and D. Alonso,
“Improvement of Ultrasonic Beacon-based Local Position
System Using Multi-Access Techniques,” Proceedings of
IEEE International Workshop on Intelligent Signal Pro-
cessing, Faro, September 2005, pp. 352-357.
[14] S. Han, H. Lim and J. Lee, “An Efficient Localization
Scheme for a Differential-Driving Mobile Robot Based
on RFID System,” IEEE Transaction on Industrial
Electronics, Vol. 6, 2007, pp. 3362-3369. doi:10.1109/
TIE.2007.906134
[15] V. Mehta, W. Sheng, T. Chen and Q. Shi, “Development
and Calibration of a Low Cost Wireless Camera Sensor
Network,” Proceedings of IEEE/RSJ International Con-
ference on Intelligent Robots and Systems, St. Louis,
October 2009, pp. 110-115.
[16] S. Shenoy and J. Tan, “Simultaneous Localization and
Mobile Robot Navigation in a Hybrid Sensor Network,”
Proceedings of IEEE/RSJ International Conference on
Intelligent Robots and Systems, Alberta, Canada, August
2005, pp. 1636-1641. doi:10.1109/IROS.2005.1545194
[17] W. Sheng, G. Tewolde and S. Ci, “Micro Mobile Robots
in Active Sensor Networks: Closing the Loop,” Proceed-
ings of IEEE/RSJ International Conference on Intelligent
Robots and Systems, Beijing, October 2006, pp. 1440-
1445. doi:10.1109/IROS.2006.281968
[18] K. Finkenzeller, “RFID Handbook: Fundamentals and
Applications in Contactless Smart Cards and Identifi-
cation,” Wiley, New York, 2003.
[19] B.-S. Choi and J.-J. Lee, “Mobile Robot Localization
Scheme Based on RFID and Sonar Fusion System,”
Proceedings of IEEE International Symposium on Indus-
trial Electronics, Seoul, July 2009, pp. 1035-1040.
[20] B.-S. Choi and J.-J. Lee, “Sensor Network Based
Localization Algorithm using Fusion Sensor-Agent for
Indoor Service Robot,” IEEE Transaction on Consumer
Electronics, Vol. 56, No. 3, August 2010, pp. 1457-1465.
doi:10.1109/TCE.2010.5606283
[21] S.-G. R, Y. H. Lee and H. R. Choi, “Object Recognition
Using 3D Tag-Based RFID System,” Proceedings of the
IEEE/RSJ International Conference on Intelligent Robots
and Systems, Beijing, October 2006, pp. 5725- 5730.