当前位置：文档库 › RFID-basedmobilerobotguidancetoastationarytarget

RFID-basedmobilerobotguidancetoastationarytarget

RFID-based mobile robot guidance to a stationary target

Myungsik Kim *,Nak Young Chong

Japan Advanced Institute of Science and Technology,School of Information Science,1-1Asahidai,Nomi,Ishikawa 923-1292,Japan

Received 28March 2006;accepted 17January 2007

Abstract

Retrieving accurate location information about an object in real-time,as well as any general information pertinent to the object,is a key to enabling a robot to perform a task in cluttered,dynamically changing environment.In this paper,we address a novel technique for the guidance of mobile robots to help them identify,locate,and approach a target in our daily environments.To this end,we propose a standard for the use of radio-frequency identi?cation (RFID)systems and develop a prototype that can be easily installed in the existing mobile robots.

Speci?cally,when an RF signal is transmitted from an RF transponder,the proposed RFID system reads the transponder-encoded data and simultaneously picks up the direction of the transponder using the received signal strength pattern.Based on the angle of signal arrival,we develop the guidance strategies that enable a robot to ?nd its way to the transponder position.Moreover,to cope with multi-path re?ection and unexpected distortions of the signals that resulted from environmental e?ects,we present several algorithms for reconstructing the signals.We demonstrate that an o?-the-self mobile robot equipped with the proposed system locates and approaches a stationary target object.Experimental results show that the accuracy of the proposed system operating at a frequency of 315MHz falls within a reasonable range in our normal o?ce environment.ó2007Elsevier Ltd.All rights reserved.

Keywords:RFID;Mobile robot;Localization;Loop antenna

1.Introduction

While considerable progress has been made in robots in the last few decades,most of them still have di?culties in the recognition of dynamically changing environments such as our daily environments.Every day,new products are manufactured and brought into our daily life,which requires robots to expand their knowledge accordingly to handle the new objects.With the limited functionalities of the current sensors,however,facilitating robot recogni-tion is quite challenging in the real world.Thus,one of the most important issues in robotics today is how we can make robots develop the competence for the given task autonomously recognizing their environment correctly and e?ciently.

Recent approaches to facilitating robot recognition are structuring the environment and connecting the robot dig-itally with its environment through wireless communica-tion systems.This enables the robot to understand and adapt to changes in the environment without any addi-tional e?orts,as the information about the environment can be directly transmitted to the robot [1–3].To build such an environment,a consistent communication interface should be developed.Recently,RFID has gained increas-ing attention,since the radio-frequency signal can eliminate the need for an optical line of sight and transmit a rela-tively large amount of information [4,5].Thus we can use the RFID tag as an information terminal and embed it into every object in the environment.Then robots read out the tag-encoded data that point to the database using the RFID reader,whereby detailed information about the object can be downloaded from the networked knowledge base.Thus robots can integrate new information seam-lessly into their existing knowledge.Despite all the promise

Corresponding author.Tel.:+81761511248;fax:+81761511149.E-mail address:myungsik@jaist.ac.jp (M.

Kim).

Mechatronics 17(2007)

217–229

surrounding RFID,it is still not possible to locate an RFID tag.We need to develop this extra function for sens-ing the need of the spatial information to the current RFID-based interface to build environmental map and execute tasks in the environment[6–8].

Over the years,for the localization problem in indoor environments,various RF-based techniques have been developed[9].One of the well tried methods was using the known locations of reference stations,where the tran-sponder position was reported by nearby stations[10,11]. However,this approach had the problem of cost and space with the number of stations and the interval between the stations a?ected the accuracy.Likewise,the signal strength-based distance estimation was often used with the stations,but the accuracy deteriorated due to the scat-tering and re?ecting of transmitting signal[12–17].Another method uses the di?erence of arrival time between the RF signal and the ultrasonic pulse[18–20].However,such sys-tems require pre-organization of location of sensors or bea-cons,thus the setup and the running cost might limit their applicability in a large scale space.Also,the optical line of sight is needed for transmitting the ultrasonic pulse.

To deal with the problem of object localization in a real and practical way,we should consider the system’s ease of use and range of applicability.Hence the system employs simple RF transponders and their reader without addi-tional sensors and/or reference stations.In this paper,as a?rst step toward RF-based localization,we propose a simple,low-cost location sensing system incorporating 315MHz active RFID devices.This system is expected to enable the robot to e?ciently gather information and understand the context of the environment using the RFID transponder’s location as well as stored information.

Speci?cally,the reader senses the object location by ana-lyzing the received signal strength pattern with bearings of its loop antenna.To cope with the signal distortions that resulted from environmental e?ects,we present signal reconstructing algorithms that help estimate the angle of arrival correctly.With this system,we make a commercial mobile robot?nd its way to a stationary RFID transponder.

The following section brie?y addresses the fundamental electromagnetic theories required to?nd directions of arri-val of RF signals.Section3introduces the design of the prototype system.Details of direction?nding results are provided in Section4.Several localization modes are pro-posed and evaluated in Section5.Finally,conclusions are drawn in Section4.

2.Direction?nding

2.1.Omni-directional antenna

The location of an object can be speci?ed in a two-dimensional polar coordinate system with two coordinates: the distance of the object from the origin and the direction (or the angle)of the object from the polar semi-axis.Even if both coordinates are not available simultaneously,we can estimate the location by triangulation with either the distance or the direction of the object.Generally,RFID systems use an omni-directional antenna such as a whip antenna,thus the location estimate resorts to the dis-tance-based triangulation.The distance from the signal source can be determined by measuring the strength of the received signal or the time of?ight of the signal[4].

The signal strength is inversely proportional to the dis-tance in the far-?eld,and the sum of the cubic of the dis-tance and the square of distance in the near-?eld, respectively.Thus the distance of the transponder can be calculated from the signal strength and the transponder position is estimated using triangulation technique.How-ever,since the accuracy deteriorates due to surface scatter-ing and re?ecting of transmitting signal,this method is hard to use in an obstacles-cluttered environment.Another way is to calculate the distance of the transponder using the time interval of transmitting signal and back-scattering sig-nal.But it is di?cult to detect the time di?erence correctly, since the time of the RF signal?ies in several meters is very small in the order of pico-seconds.To cope with the above-mentioned problems of the omni-directional antenna used in current RFID systems,we investigate the directional antenna.

2.2.Directional antenna

With the directional antenna such as a loop antenna,the arrival direction of the RF signal can be determined,as the strength of the received signal changes according to the angle between the antenna plane and the transmitted wave plane as shown in Fig.1.If an electromagnetic wave is transmitted to a loop antenna,a voltage V is induced in the antenna coil given as

V/CSB j sinehàuTj;e1Twhere S is the surface area of the antenna,B the magnetic ?ux density of the transmitted signal passing through the antenna,h the rotation angle of the antenna,u the direc-tion of the transponder,and C the environmental e?ect constant[21–24].The pattern of the induced voltage level

218M.Kim,N.Y.Chong/Mechatronics17(2007)217–229

with the rotation of a loop antenna is ideally the Arabic nu-meral eight in polar coordinates,we can then estimate the direction of the transponder using the maximum or the minimum value of the received signal strength.The loca-tion of the transponder can be determined by two methods using the antenna bearings to the transponder,one is using the signal strength.As the distance can be determined using the detected signal strength,the position of the transponder can be determined with the direction and the distance.Since the signal strength of the directional antenna changes according to the antenna bearings,the distance should be calculated using the maximum signal strength of the re-ceived signal pattern.However,the accuracy of the dis-tance estimated from the signal strength is not high in real environments.The other method is the triangulation technique.After sensing the direction to the transponder at two di?erent positions,the location of the transponder can be calculated by two bearings and the interval between two sensing positions.3.Prototype development 3.1.Experimental setup

In order to verify the feasibility of RF-based localiza-tion,we have developed a prototype 315MHz RFID reader for location awareness.The diagram and the photo of the developed location sensing system are shown in Fig.2.The system consists of an electrically small loop antenna and its reader.When an RF signal is transmitted from an RF transponder,the signal received by the loop antenna mounted to the pan and tilt unit is fed to the RF reader module.A signal strength detector in the mod-ule converts the RF signal to the DC voltage and transmits it to a signal analyzer,then the signal analyzer generates the digitalized signal strength and the identi?cation (ID)code encoded in the transponder.The generated signal

strength and ID code together with the antenna facing direction is provided to the application target through a serial communication interface.Then the direction of the transponder signal with an ID code is determined using the signal strength pattern through rotating the antenna facing angle.3.2.Antenna

The loop antenna is designed in the following way:(i)Choose an antenna type that is highly directional.(ii)Using the electromagnetic simulation software

‘‘ACCUFIELD’’developed by Fujitsu limited for electromagnetic compatibility analysis,a number of loop antennas with various shapes and shielding patterns are tested.

(iii)Considering the characteristics such as resonant

frequency,impedance,gain,and radiation pattern,the shape of the antenna is optimized.

(iv)Prototype and manufacture the antenna,and mea-sure its speci?c characteristics.

(v)Match the impedance of the antenna and tune for a

speci?c frequency.Fig.3shows the simulation result of the radiation pat-tern of the antenna design in polar coordinates scaled in dB l V/m.The pattern has apparently two minimum and two maximum levels that display the same characteristics as the signal receiving pattern of the directional antenna.Fig.4a shows the developed loop antenna manufactured by printed circuit board etching,having a width of 78mm and a height of 32mm.The rear side is shielded and perpen-dicularly opened to improve the sensitivity of signal strength sensing about the horizontal plane.The signal receiving pattern of the antenna according to the antenna rotation angle is shown in Fig.4b.The transponder is set

Pan/Tilt Head

Signal Strength Detector

Data Analyzer

RF Transponder

Signal Strength Detector Signal Analyzer

Signal Strength,

ID Code

RS 232C

RF reader Module

Loop Antenna

Loop Antenna a

M.Kim,N.Y.Chong /Mechatronics 17(2007)217–229219

at the direction of0°and the pattern is measured in an elec-tromagnetic dark room.The pattern also has apparently two dominant minimum and maximum levels that can be used to satisfy the direction?nding requirement.If we?nd the antenna rotation angle at which the voltage level is min-imized or maximized,we can estimate the direction of the https://www.wendangku.net/doc/bb10413843.html,paring to the maximum level,the mini-mum level is more evident and less a?ected by environmen-tal conditions.Thus,we can?nd the direction of the RF signal with good accuracy using the minimum level[25]. 3.3.Signal strength detector

Fig.5shows the315MHz RF signal strength detector residing inside the RF reader module.The schematic of the detector is shown in Fig.5a.When an RF signal in the range ofà120dBm toà45dBm is transmitted to the antenna,the signal is?ltered and fed to the power detector, which is then converted and ampli?ed to a DC voltage ranging from0to5V.

Fig.5b shows the picture of the detector that has a width of54mm and a height of24mm.The input and out-put ports include the RX input from the antenna,the5 DCV input from the robot USB port,and the DC output fed to the signal analyzer.Fig5c shows the input–output characteristics,where the input signal within the range of à115dBm toà45dBm is converted linearly to the DC voltage.

3.4.Signal analyzer

Fig.6shows the block diagram of the signal analyzer developed using a micro processor.When an RF signal is

Fig.4.Antenna prototype:(a)loop antenna and(b)signal receiving pattern of the antenna.24 mm

GND

54 mm

220M.Kim,N.Y.Chong/Mechatronics17(2007)217–229

transmitted from a transponder,the induced signal at the antenna is transmitted to the RF reader module and the signal strength detector in the module converts the signal to a DC voltage pulse,which is sent to an A/D converter of the microprocessor.The signal from the transponder used in this experiment is modulated by On–O?Keying (OOK)and the DC voltage pulse from the signal strength detector has the same rectangular shape.Thus,the micro-processor counts ons(1’s)and o?s(0’s)and adds the signal strength when the digitalized pulse signal is higher than the threshold,whereby the ID code and average strength of the signal is obtained.Then the signal analyzer sends an ordered set of ID and strength level to the robot via the RS-232C interface.However,since the o?set level of the voltage is easily varied by the system conditions,the thresh-old is dynamically adjusted by monitoring the minimum level of the transmitted pulse.Automating the above pro-cesses,the transponder direction can easily be determined with an antenna scanning in the horizontal plane[26]. 4.Direction sensing result of the system

4.1.Direction sensing using the prototype

Fig.7shows the received signal pattern measured in the hallway using the developed system.The signal source is a 315MHz active RF transponder manufactured by Circuit

Design Inc.The transponder continuously sends RF signal having a uniform interval with four selectable20byte ID codes through a quarter wavelength whip antenna.The mode of operation requires3DCV.Note that the direction towards which the antenna plane faces the transponder is0°. The strength level becomes minimal when the antenna faces the transponder,and is reduced when the distance increases.

Fig.8shows the statistical results about the estimated direction according to the transponder distance.Directions were measured50times from various distances with respect to the transponder at0°.The?gure indicates that the aver-age directions measured from all the distances are within the range of±1°with a maximum of±4°errors.This accu-racy can approximate the transponder position that will remain within7%of the distance between the communicat-ing pairs of the transponder and the reader.We have thus established a reasonable assurance of accuracy as we expected for various indoor applications.

However,many uncertain factors can exist and contam-inate the original signal in real environments.It is therefore very di?cult to?nd the correct direction if we simply attempt to?nd the minimum strength level in the distorted strength pattern.We present several algorithms to precisely determine the direction to the transponder in the following subsection.

4.2.Signal reconstruction algorithm

In a real environment,RF signals are easily distorted by the environmental e?ect such as re?ection,refraction,and scattering by obstacles.Also,the white noise of the system a?ects the received signal pattern substantially.Fig.9

Threshold

0 1 1 0 1 0 0 1 1 1 0

Strength + ID

The structure of signal analyzer.

M.Kim,N.Y.Chong/Mechatronics17(2007)217–229221

shows the distorted signal pattern by the environmental e?ect.When a metallic chair is positioned between the

robot antenna and the transponder as shown in Fig.9a,the propagation path of the transponder signal may be dis-turbed.In Fig.9b,the solid line is the signal pattern with-out the chair and the dashed line is the distorted pattern by the chair.The signal strength was lowered and abruptly dropped when the chair existed.

Fig.10illustrates the proposed scheme to sequentially reconstruct the distorted patterns.Fig.10a shows the dis-torted signal pattern.The adjacent averaging in Fig.10b eliminates the unexpected peak by replacing the data with the new value by averaging neighboring values around each data point.But the signal pattern is not very symmetrical.As one half is too small comparing another half,the detected minimum level is still incorrect.The minimum level should be found for the data bounded by the in?ec-tion points.If we designate the in?ection points over the received pattern,whereby the pattern can be divided into a valley and two hills,we can ?nd the minimum point in the valley range.Since the signal form should be an abso-lute sine function,we can employ a 4th order polynomial ?tted to the data points as shown in Fig.10c.Then two local maxima and one minimum can be found with the in?ection points from the dashed ?tting line.We can ?nally determine the minimum strength level and its correspond-ing antenna angle between the two inner in?ection points from the actual signal pattern as shown in Fig.10d.Thus,the direction to the transponder can be estimated correctly,even though the direction sensing RFID system is exposed to the environmental e?ect and the white noise [27].

Chair

1.5

2.0

Without Obstacle a

With Chair

e n g t h (D C V )

Mobile Robot

Transponder

222M.Kim,N.Y.Chong /Mechatronics 17(2007)217–229

We believe that this fairly constant level of accuracy of the direction sensing enables to reliably determine the posi-tion of the transponder in indoor environments.The direc-tion-based localization of the transponder will be discussed in the following section.

5.Localization schemes

We test the location sensing accuracy of the proposed RFID system using a mobile robot.The developed direc-tional antenna is attached to the pan and tilt head of the robot that acquires the strength of the signal according to the panning angle of the antenna,as well as the tran-sponder identi?cation data.Three di?erent methods are proposed in this work.

5.1.Navigation through a minimum level direction

The location of the transponder can be determined with two parameters:the direction of the transponder and the distance in the direction calculated by the signal strength. The relationship between the distance and the signal strength measured by the developed system is shown in Fig.11.The strength was measured10times at each posi-tion.Dots indicate the maximum strength of the signal in the logarithm scale and the red line1?ts their average value. It is shown that the strength is reduced rapidly in the near ?eld range,and linearly in the far?eld range.Therefore, the distance to the transponder can be roughly calculated from the signal strength V max as

A1?V b

max

tC1

A2?V maxtC2

(

;e2T

where A and C are the calibration constants re?ecting the environment e?ect and b is the order to?t the slope in the near?eld.We set b as2for simplifying the equation.

However,the distance is apt to be overestimated due to the scattering and the re?ection of the signal in real envi-ronments.As the loss and gain of the strength is varied by the environmental condition,we usually estimate the strength of the arrival signal using the path loss and atten-uation table[28].For example,in the general o?ce envi-ronment,it is known that the error in strength estimation will not exceed about20%.Note that this just gives a rule of thumb for estimating,and if there exist partitions with metallic frame,the error de?nitely increases.The robot thus needs to measure the signal strength intermittently while moving an estimated distance until the pre-speci?ed strength is detected.Note that the curve rises rapidly in the near?eld zone while the robot comes close to the tran-sponder position.Thus we can calculate the distance with enough high accuracy.If the strength passes beyond the pre-speci?ed level,the robot stops moving,since it has approached within close proximity to its target.

Fig.12shows the results of the robot navigation follow-ing the direction of the minimum strength.The inset shows the pattern of the signal strength measured at respective positions.In Fig.12a,the direction of the minimum strength is4°and the maximum signal strength is relatively weak.In Fig.12b,it is shown that the robot then rotates4°clockwise and heads in that direction and travels4m that has a margin of1m from the calculated distance in Eq.

(2).The robot repeated scanning the environment after it traveled4m,so that it could compensate for the errors in the sensing direction as well as the distance calculation. In Fig.12c,it is shown that the robot modi?es its heading angle and travels along the new direction.Finally in Fig.12d,when the maximum signal strength was above the terminating threshold,the red dotted line1in the inset, it has been shown that the robot approaches the transpon-der within50cm of its target.

Fig.13shows the localization results with various tran-sponder distances after calibrating constants A and C,and the terminating threshold strength.Since the shape of the distance to strength curve changes as the sensing range changes from the near?eld to the far?eld,constants A and C were determined by measuring the signal strength at di?erent positions.

However,the transmitted signal is easily a?ected by the environmental e?ect,including humidity and time as well as the system condition.Thus,it is impossible to acquire the correct calibration constant that can be used to deter-mine the exact distance to the transponder.

Using the navigation method,the robot moves directly toward the transponder and corrects the direction and the distance when it terminates the current travel.As this method simultaneously uses the direction and the distance, it is expected to be fast and used in a long narrow space. However,it is di?cult for the ordinary user to use this method which is highly dependent on the?eld calibration.

0123456789101112

Distance (m)

Near

field

Far field

Mixed

area

11.The relation between signal strength and distance.

1For interpretation of the references to color in this?gure,the reader is

referred to the web version of this article.

M.Kim,N.Y.Chong/Mechatronics17(2007)217–229223

5.2.Bearing-based triangulation

Another approach is the triangulation method using the antenna bearings to the transponder measured at two dif-ferent positions as shown in Fig.14.When an electromag-netic signal is transmitted from a transponder,the robot can determine the direction of the transponder h 1at the

current position.The robot then turns h t clockwise or counterclockwise after aligning the heading in the tran-sponder direction and moves to another position to deter-mine the direction of the transponder h 2at that position.The turning angle h t and the moving interval l can arbi-trarily be decided.Then the distance d from the robot to the transponder can be calculated as

Transponder

Robot

5 m

scan

4 m

S i g n a l S t r e n g t h (D C V )

Angle (Degree)

c d

Checks signal strength

S i g n a l S t r e n g t h (D C V )

Angle (Degree)

-20?

S i g n a l S t r e n g t h (D C V )

Angle (Degree)

Travels along the estimated direction

224M.Kim,N.Y.Chong /Mechatronics 17(2007)217–229

d ?l

cos h s tsin h s

tan

h t

;

e3T

where h s is the acute angle of intersection between the tran-sponder direction h 2and the robot heading direction calcu-lated as h s ?p àh 2.

However,the directional antenna prototype includes the direction sensing error as shown in Fig.8.It is also noted that the transponder localization accuracy depends on the turning angle h t and the interval distance l .To deter-mine the turning angle and the interval distance that yield a low localization error,we simulate how the direction sensing error a?ects the localization error.In the right-handed Cartesian coordinate system depicted in Fig.15,the direction sensing system is located at (0,0),denoted in the ?gure as Position 1,and the transponder is located at (0,y 0).The robot ?nds the direction of the transponder

at this position with an error of u .Assume that the interval distance can be written as l =py 0,where p is the constant relating the estimated distance of the transponder and the interval.Thus,the second measurement position is ?àpy 0sin eh t tu T;py 0cos eh t tu T :

e4T

Note that h t and u are negative values,because these are turned clockwise from the y -axis.Thus the coordinates in Eq.(4)indicate Position 2in the ?gure.If we draw Line 1from Position 1to the transponder with the direction sensing error u ,the equation of the line can be written as

x ?àtan u y :e5T

Similarly,Line 2from Position 2extending to the transpon-der direction with the sensing error c can be written as y ?tan ea tc Tx tpy 0

cos ea tc àh t àu T

cos ea tc T

;

e6T

where

a ?tan

à1

1àp cos eh t tu T

p sin eh t tu T

Then,we can determine the location of the transponder as the intersection point between Line 1and Line 2at the following coordinates:àpy 0sin u cos ea tc àh t àu Tcos ea tc àu T;py 0cos u cos ea tc àh t àu T

cos ea tc àu T

e7TFig.16shows the triangulation results of the transpon-der located at (0,1)calculated by Eq.(7)with respect to the interval constant and the turning angle.Considering the accuracy of the prototype,the direction sensing error was assumed to lie within the range of ±4°with a Gaussian dis-tribution.The results are plotted within the range of inter-est,from à0.35to 0.35in x -axis and from 0.65to 1.35in y -axis.

The expected error is considered satisfactorily small with the interval constant 1and the turning angle 15°shown in the ?gure.However,it is sometimes di?cult to have an enough interval distance in spatially constrained real envi-ronments,where obstacles may exist.Also,the transponder distance tends to be overestimated by the environmental e?ect.Therefore,the smaller values of the interval constant might be more appropriate for the real environment.When the interval constant is less than 1,the error becomes small with a 45°turning angle.We therefore determined that the turning angle is kept at 45°and the interval constant remains less than 1in this work.

Fig.17shows how the robot localizes a transponder using the triangulation method with selected parameters.The signal strength pattern measured at each position is also shown.In Fig.17a,it is shown that the robot scans the environment and ?nds the direction of the minimum level to be 0°as expected,since the transponder is located in front of the robot.The robot turns 45°clockwise and moves to the next position as shown in Fig.17b.The mov-

M.Kim,N.Y.Chong /Mechatronics 17(2007)217–229225

ing interval is determined by the maximum level of the sig-nal strength.At the second position,the robot turns 45°counterclockwise and scans again in Fig.17c.Then the robot determines the transponder position using the two bearings measured at each position and the interval between them.Finally,the robot turns 45°counterclock-wise and approaches the transponder.

Fig.18shows the localization results as the transponder position changes.The black circles show the initial robot position and the transponder position,the small dots indi-cate the second measurement point and the ?nal robot positions and the lines are the paths of the robot,and the large gray circle is the 50cm boundary from the transpon-der position.At the start position,the robot measures the direction of the transponder,and turns 45°clockwise to move to the next position.The interval distance constant selected is 0.5.The robot determines the direction at the second position and calculates the distance to the transponder.

When the robot moves toward the transponder located less than 3m away,it can arrive within the 50cm circle

from the transponder as shown in Fig.18a–d.As the dis-tance increases,the localization error increases as shown in Fig.18e,which is resulted from the direction sensing error.Also,if the distance of the transponder increases,the interval distance increases as well.Therefore,the trian-gulation method is not suitable to localize a transponder located a long distance away in a narrow space.5.3.Hybrid

To cope with the problems of both the navigation method and the triangulation method,we simultaneously employ these two methods.As we mentioned before,the triangulation method would not be suitable for a long dis-tance and the navigation method cannot be used without proper calibration.Even though it is di?cult to ?nd the exact position by the navigation method,the robot can move toward the transponder.If the signal strength exceeds the pre-determined strength threshold,the robot ?nds the transponder position using the triangulation method.Fig.19illustrates the basic idea of the

hybrid

Fig.16.Location sensing accuracy with respect to interval constants and angles.

226M.Kim,N.Y.Chong /Mechatronics 17(2007)217–229

method,where the robot navigates toward the transponder in the minimum strength direction and triangulates the transponder position when the robot approaches within a boundary.The boundary was selected as 2m considering the accuracies that the current triangulation and strength-based distance estimation methods can o?er.

It is shown in Fig.20that the robot approaches its docking target.Note that the moving interval is di?erent in Fig.20a and b,because we used two di?erent sets of ter-minating threshold levels and C s in Eq.(2).The robot thus may not reach the position of the transponder or may overrun passing the transponder if only the navigation method is employed.However,using the triangulation method as well as the navigation method,the robot could approach within 30cm from the transponder in both cases.

2 m

c S i g n a l S t r e n g t h (D C V )

Angle (Degree)

turn and move

1 m

scan Transponder

Robot

S i g n a l S t r e n g t h (D C V )

Angle (Degree)

-45?

scan

turn and move

1 m

M.Kim,N.Y.Chong /Mechatronics 17(2007)217–229227

6.Conclusion

The RFID system for location awareness is a very prom-ising and challenging research issue.In this paper,we pro-posed a cost-e?ective location sensing RFID system for mobile robot object location tracking in indoor environ-ments.The prototype RFID system that consisted of a loop antenna,a signal strength detector,and a signal ana-lyzing unit determined the location of the RFID transpon-der analyzing the strength pattern of received signals with rotating the antenna.The antenna was attached onto the pan head of the mobile robot.Because the signal can be distorted in the real environment,three algorithms were

presented to reconstruct the distorted signal and precisely determine the arrival direction of the signal.It was demon-strated that the robot equipped with the proposed system reached the transponder at various distances using three methods:navigation,triangulation,and hybrid.The accu-racy of the overall system fell within a reasonable range as we expected for possible applications in indoor environ-ment.Our future e?ort includes improving the accuracy of the system with the multi-axis array antenna and deploy-ing the mobile robot into networked unknown environ-ments.We are developing implementation scenarios taking into account the localization accuracy.Acknowledgements

This research is conducted as a program for the ‘‘Foster-ing Talent in Emergent Research Fields’’in Special Coor-dination Funds for Promoting Science and Technology by Japan Ministry of Education,Culture,Sports,Science and Technology.This work was also supported in part by Korea MIC and IITA through IT leading R&D Support Project.References

[1]Chong NY,Hongu H,Ohba K,Hirai S,Tanie K.A distributed

knowledge network for real world robot applications.Proc IEEE/RSJ Int Conf Intell Robots Syst 2004:187–92.

[2]Holmquist LE,Gellersen HW,Kortuem G,Antifakos S,Michahelles

F,Schiele B,et al.Building intelligent environments with smart-Its.IEEE Comput Graph Appl 2004;24(1):56–64.

[3]Brumitt B,Meyers B,Krumm J,Kern A,Shafer S.EasyLiving:

technologies for intelligent environments.Handheld Ubiquitous Comput 2000(September):12–29.

[4]Finkenzeller K.RFID handbook:fundamentals and applications in

contactless smart cards and identi?cation.2nd ed.John Wiley &Sons Ltd.;2003.

Transponder

Robot

6 m

scan

S i g n a l S t r e n g t h (D C V )

Angle (Degree)

boundary

turn and move

Angle (Degree)

2 m boundary

-15?

turn and move

scan and determine the position

Angle (Degree)

- 42?

Start pt.

Tag pt.

Navigation

Triangulation

S i g n a l S t r e n g t h (D C V )

228M.Kim,N.Y.Chong /Mechatronics 17(2007)217–229

[5]Microchip.MicroID TM13.56MHz RFID system design guide.

//https://www.wendangku.net/doc/bb10413843.html,/downloads/en/DeviceDoc/21299E.pdf>. [6]Harter A et al.The anatomy of a context-aware application.In:

Proceedings of the5th annual international conference on mobile computing and networking(Mobicom99).New York:ACM Press;

1999.p.59–68.

[7]Brumitt B,Meyers B,Krumm J,Kern A,Shafer S.Ubiquitous

computing and the role of geometry.IEEE Personal Commun 2000:41–3.

[8]Want R,Schilit BN.Expanding the horizons of location-aware

https://www.wendangku.net/doc/bb10413843.html,puter2001;34(8):31–4.

[9]Hightower J,Borriello G.Location systems for ubiquitous comput-

ing.IEEE Comput Mag2001;34(8):57–66.

[10]Want R,Hopper A,Falsco V,Gibbons J.The active badge location

system.ACM Trans Inform Syst1992(January):91–102.

[11]Werb J,Lanzl C.Designing a positioning system for?nding things

and people indoors.Spectrum,IEEE1998;35(9):71–8.

[12]Hightower J,Borriello G,Want R.SpotON:an indoor3D location

sensing technology based on RF signal strength.UW CSE Technical Report;February18,2000.

[13]Bahl P,Padmanabhan Venkata N.RADAR:an in-building RF-based

user location and tracking system.IEEE Infocom2000;2:775–84. [14]Ni LM,Liu Y,Lau YC,Patil https://www.wendangku.net/doc/bb10413843.html,NDMARC:indoor location

sensing using active RFID.ACM Wireless Network2004;10(6): 701–10.

[15]Priyantha NB,Chakraborty A,Balakrishnan H.The cricket location-

support system.In:Proceedings of the6th annual international conference on mobile computing and networking(Mobicom

00).ACM Press;2000.p.32–43.

[16]Priyantha NB,Miu AKL,Balakrishnan H,Teller S.The cricket

compass for context-aware mobile applications.In:Proceedings of the7th annual international conference on mobile computing and networking(Mobicom01).ACM Press;2001.p.1–14.[17]Harder A,Song L,Wang Y.Towards an indoor location system

using RF signal strength.In:Proceedings of the international conference on information technology:coding and computing (ITCC’05),vol.2;2005.p.228–33.

[18]Fukuji Y,Minami M,Morikawa H,Aoyama T.DOLPHIN:an

autonomous indoor positioning system in ubiquitous computing environment.In:Proceedings of the IEEE workshop on software technologies for future embedded systems(WSTFES03);2003.p.53–6.

[19]Martin Abreu JM,Ceres R,Calderon L,Jimenez AR,Gonzalez-de-

Santos P.Measuring the3D-position of a walking vehicle using ultrasonic and electromagnetic waves.Sensor Actuator 1999;75:131–8.

[20]Martin Abreu JM,Jimenez AR,Seso F.Estimating the3d position

form time delay data of us-waves:experimental analysis and new processing algorithm.Sensor Actuator2002;101:311–21.

[21]Pozar DM.Microwave and RF wireless systems.Wiley Text Books;

2000.

[22]Jackson JD.Classical electrodynamics.John Wiley&Sons Ltd.;

1999.

[23]Stutzman WL,Thiele GA.Antenna theory and design.John Wiley&

Sons Ltd.;1999.

[24]Balantis CA.Antenna theory:analysis and design.Wiley Text

Books;1996.

[25]Kim M,Takeuchi T,Chong NY.Object location sensing using

active RFID systems.Proc Int Symp Robot Automat2004: 440–5.

[26]Kim M,Kubo T,Chong NY.RF power detector for location

sensing.Proc Int Conf Control Automat Syst2005:1771–4.

[27]Kim M,Chong NY.RF signal restoration?lter for ubiquitous

location sensing.Int Conf Ubiquitous Robots Ambient Intell 2005:183–7.

[28]Rapaport TS.Wireless communications.Prentice-Hall;2002.

M.Kim,N.Y.Chong/Mechatronics17(2007)217–229229