Indoor Localization by Pattern Matching Method Based On Extended Database

Size: px

Start display at page:

Download "Indoor Localization by Pattern Matching Method Based On Extended Database"

Elfrieda Holt
6 years ago
Views:

1 Vol:8, No:, Idoor Localizatio by atter Matchig Method Based O Exteded Database Gyumi Hwag, Jihog Lee Iteratioal Sciece Idex, Mechaical ad Mechatroics Egieerig Vol:8, No:, waset.org/ublicatio/9997 Abstract This paper studied the CSS-based idoor localizatio system which is easy to implemet, iexpesive to compose the systems, additioally CSS-based idoor localizatio system covers larger area tha other system. However, this system has problem which is affected by reflected distace data. This problem i localizatio is caused by the multi-path effect. Error caused by multi-path is difficult to be corrected because the idoor eviromet caot be described. I this paper, i order to solve the problem by multi-path, we have supplemeted the localizatio system by usig patter matchig method based o exteded database. Thereby, this method improves precisio of estimated. Also this method is verified by experimets i gymasium. Database was costructed by m itervals, ad 6 sample data were collected from radom positio iside the regio of poits. As a result, this paper shows higher accuracy tha existig method through graph ad table. Keywords Chirp Spread Spectrum (CSS), Idoor Localizatio, atter-matchig, Time of Arrival (ToA), Multi-ath, Mahalaobis Distace, eceptio ate, Simultaeous Localizatio ad Mappig (SLAM), Laser age Fider (LF). L I. INTODUCTION OCALIZATION techique, oe of the importat techiques i the field of service idustry, is beig highlighted. The localizatio ca be divided ito outdoor localizatio ad idoor localizatio. I the outdoor localizatio, GS is mostly used, but GS caot be used for the idoor localizatio, as the usually caot commuicate with satellites. For this reaso, may researches to solve this problem are ogoig, ad research about close distace commuicatio is beig focused more tha ever. Localizatio based o the close distace commuicatio used methods such as Time of Arrival (ToA), which measures the distace through radio wave s travel time, ad the SSI method, which measures the distace through the chage of radio wave s itesity []-[]. CSS-Based idoor localizatio system [6], [9], which is suggested i existig research, uses a ToA method to estimate the locatio with the distace [7], [8]. However, there exists a error i distace due to the multi-path caused by the complicated ad arrow idoor, which leads to a large error i localizatio. atter matchig method is beig studied to solve these problems []. This paper proposes the method that This research was fiacially supported byleadig idustry of advaced mechaics ad equipmets of Chugcheog Leadig Idustry Office of the Korea Miistry of Kowledge Ecoomy. Gyumi Hwag is with the Mechatroics group of Chugam Natioal Uiversity, Daejeo, Korea (phoe: ; fax: ; hgmdooly@ate.com). Jihog Lee is with the Mechatroics group of Chugam Natioal Uiversity, Daejeo, Korea ( jihog@cu.ac.kr). ehaces the accuracy of existig patter matchig. II. ATTEN MATCHING A. atter Matchig Method The existig trilateratio method has a very low accuracy due to the error i distace caused by the multi-path we have metioed earlier. Therebythe patter matchig method has bee studied to solve these problems. Also Filterig of algorithm has bee ehaced by usig the exteded (Database), which was to improve the accuracy of existig patter matchig. First at all, patter matchig method localizes by iterpolatio through the similarities betwee data, ad it cosiders the errors caused by the multi-path a characteristic of the data, which is used i the process of localizatio []. For usig the method, we costructed database i (). = m = m m m m [ ] gap () () m [ I ] = () T m [ ] I = () : age data of database poit : age data of gap : Differece betwee with, where m is distace data from achor to tag. meas umber of achor, ad m meas umber of poit from(). Costructio of (previous ) about the certai eviromet is ecessary to develop localizatio method. Also, is rage data received at radom locatio from i (). To calculate gap, we created vector I to modify size of i (). poits with relatively low gap is extracted by comparig the differece i results, as i (). Not oly that, by cosiderig the differece i distace with poits that showed high similarities, weighted values are give for each poits ad iterpolatio is utilized to extract the fial Iteratioal Scholarly ad Scietific esearch & Iovatio 8() scholar.waset.org/7-689/9997

2 Vol:8, No:, Iteratioal Sciece Idex, Mechaical ad Mechatroics Egieerig Vol:8, No:, waset.org/ublicatio/9997 locatio. B. atter Matchig Method Based O Exteded A existig patter matchig method extracted the fial result by piig 6 poits that showed high similarities, ad each differece i the distace to achor has bee summed ad was used as oe represetative value while fidig out the similarity betwee poits. Furthermore, accordig to the circumstace, poits was ot always six, ad which ca be selected agai from to 6 by removig poit far from the distributio about each poit. Although existig database oly had distace data, we tried to improve the accuracy by usig the patter matchig method based o the exteded, which added the receptio rate from achors. C. Exteded Structure ' = m m m m m m m ' : age ad receptio rate of database poit m : eceptio rate of each achors We costructed a ew ', ad exteded from(), by storig the receptio rate of each achors. We stored the receptio rate of each achors i m. The receptio rate is lowered whe a idoors obstacle hiders the propagatio from a certai achor to the tag. Ad it was cosidered to distribute the weighted value, i order to lower the reliability of the achor. D. Filterig by Usig Mahalaobis Distace The existig patter matchig method extracted 6 poits that showed high similarities, i other words, that was i close distace to the real s. However, there were big errors i the fial results because poits that are far from the real locatio has bee chose, which was caused by the structures of idoor eviromet. [ x x x x x x ] T x 6 D () = () = [ y y y y y y ] T (6) y 6 M ( x, y) = T ( x y) S ( x y) D M : Mahalaibis distace S : Covariace matrix of x, y To lesse these errors, 6 poits were basically extracted; the average ad stadard deviatio of the 6 poits were (7) that is threshold to filterig 6 poits. poits that have higher value tha the threshold were removed; ad four to six poits were filtered case by case. (7) E. Weighted Idex Based O eceptio ate [ ρ ρ ρ ] ρ = (8) ρ6 ρ K = (9) ρ ρ ρ ρ [ K K K ] K6 K = () ρ : Differece i distace K : Weighted value I the existig patter matchig method while fidig out the similarity of poits i(8), was calculated to give weighted values o differet poits. The existig method always fixed 6 poits, thereby 6 weighted values are calculated from (9) ad (). K [ ρ ρ ρ ] ρ = () 6 ρ [ ] = () ρ ' = () ρ ρ ρ ρ [ K' K' K' ' ] K = () ' : eceptio rate of each achors K ' : Modified weighted values Not oly ρ of the differece i distace as (), but also was cosidered as (), so that weighted values were lesseed whe receptio rate was low as ().Through filterig by usig Mahalaobis distace, ca be chose from to 6. Thereby cosiderig receptio rate, weighted values were modified as (). K III. EXEIMENTAL ENVIONMENT AND CONSTUCTION OF OBOT SYSTEM A. Experimetal Eviromet Fig. Gymasium Iside the gymasium, data (distace data, real positio data) Iteratioal Scholarly ad Scietific esearch & Iovatio 8() scholar.waset.org/7-689/9997

Vol:8, No:, about poits ( by ) were costructed by m itervals. The 6 sample data (distace data, real positio data) were collected from radom positio iside the regio of poits. B. obot System Fig.

, the with two wheels was made for a comfortable movemet idoors.

receive distace data from each achors. Usig the portable computer based o Liux through OS (obot Operatio System), locatio data of ad distace data from achor to tag could be stored at real-time. Fig.

3 Vol:8, No:, about poits ( by ) were costructed by m itervals. The 6 sample data (distace data, real positio data) were collected from radom positio iside the regio of poits. B. obot System Fig. CSS based module (CorebellNexbee) Iteratioal Sciece Idex, Mechaical ad Mechatroics Egieerig Vol:8, No:, waset.org/ublicatio/9997 Fig. The As i Fig., the with two wheels was made for a comfortable movemet idoors. Uder the was istalled a log distace LF (Laser age Fider) Sesor, which helped get the locatio of the through SLAM(Simultaeous Localizatio ad Mappig), ad Tag was istalled o the upper part, which helped receive distace data from each achors. Usig the portable computer based o Liux through OS (obot Operatio System), locatio data of ad distace data from achor to tag could be stored at real-time. Fig. Laser rage fider (Hokuyo UTM-LX) I order to fid out the real positio of the, which will be used to compare with the estimated positio by the patter matchig method, LF sesor was adhered o the, ad we used positio data through SLAM. CSS based module, which shows the distace data through achor ad tag by ToA method, was used. 8 Achors were adhered to the atea o height of m durig the experimet.i order to receive distace data ad store it, the tag was adhered to the upper part of the. IV. ANALYSIS OF EXEIMENTAL ESULTS A. Trilateratio Fig. Output of trilateratio obot Achor Trilateratio Usig the trilateratio method used i GS, there exists a big error i Fig.. This is because the distace icludes the multi-path that is ot a real distace from achor to tag. Estimated positio shows the error distace about m. As seeig the distributio of the estimated positio, the accuracy ad precisio are also very low. Iteratioal Scholarly ad Scietific esearch & Iovatio 8() scholar.waset.org/7-689/9997

4 Vol:8, No:, B. atter Matchig Method Based O revious Database obot revious Iteratioal Sciece Idex, Mechaical ad Mechatroics Egieerig Vol:8, No:, waset.org/ublicatio/ Fig. 6 Output of previous Usig the existig patter matchig method based o previous, we ca get a locatio value that is very ear to the real positio of the : because, the error is greatly reduced by usig the data that icludes multi-path as the characteristic of poit. However, whe lookig at Fig. 6, we ca easily see that accuracy is slightly low. Because icorrect poits were selected by idoor structure, icorrect positio is estimated. C. atter Matchig Method Based O Exteded - - Fig. 7 Output of exteded Database obot Exteded Usig the patter matchig method based o exteded, we ca get a result with higher accuracy as Fig. 7, because filterig of poits ad receptio rate is cosidered while distributig the weighted values. Also, the precisio ad accuracy of estimated positio is greatly icreased ad compared with the trilateratio or the existig patter matchig method revious Exteded Fig. 8 Compariso of error I Fig. 8, we collected 6 sample data ad compared the errors betwee the existig patter matchig method based o previous ad the patter matchig method based o exteded. There were ot dramatic reductio i the errors i few of the sample data, but i most of them, we could see reductio i the error about sample data. Fig. 9 Error i both axes From 6 sample data, some sample data is chose ad is compared the error i the x-axis ad y-axis. Error rate is lower tha existig method whe it is closer to the X-mark (, ).The errors is reduced by usig patter matchig method based o exteded i Fig. 9. Iteratioal Scholarly ad Scietific esearch & Iovatio 8() scholar.waset.org/7-689/9997

Vol:8, No:, Iteratioal Sciece Idex, Mechaical ad Mechatroics Egieerig Vol:8, No:, waset.org/ublicatio/9997 D. Total Compariso of Error...... 6 7 8 9 6 Trilateratio revious Exteded Fig.

5 Vol:8, No:, Iteratioal Sciece Idex, Mechaical ad Mechatroics Egieerig Vol:8, No:, waset.org/ublicatio/9997 D. Total Compariso of Error Trilateratio revious Exteded Fig. Total compariso of error I Fig., we compared the real positio of the with the estimated positio extracted from trilateratio method, existig patter matchig method, patter matchig method based o exteded. All the sample data showed less error whe usig patter matchig, compared to the trilateratio method. Also, through Table I, we ca see that higher accuracy ad lower error rate was foud whe usig the patter matchig method based o exteded. TABLE I TOTAL COMAISON OF EO Method Average of Trilateratio Method.98 atter Matchig Method Based o revious. atter Matchig Method Based o Exteded.9 V. CONCLUSION Idoor localizatio research is ogoig thaks to ever icreasig attetio o this area. However, there are ot ay stable methods like GS. Ad GS caot be used for idoor eviromet. The existig trilateratio method shows low performace because localizatio is affected by multi-path effect. This paper aimed for a idoor localizatio method with the iexpesive sesor by overcomig the multi-path caused by the barriers of idoor, which was doe by usig the patter matchig method based o exteded. Eve though this method eed database about certai idoor eviromet, this method ca overcome effect by the multi-path, which ca estimate correct locatio tha the existig method. For further research, by applyig it to a movig, a deeper research o movig should be doe based o dead reckoig ad EKF. EFEENCES [] X. Y. Zhou ad M. S. ark, A idoor localizatio mechaism usig active FID tag, i roc. IEEE Iteratioal Coferece o Sesor Networks, Ubiquitous, ad Trustworthy Computig, 6. [] J. Y. Zhou ad J. Shi, FID localizatio algorithms ad applicatios a review, Joural of Itelliget Maufacturig Volume, Number 6 pp , 8. [] M.H. Fracis,.F. Wilso, T. Kos, L.E. Miller, N.. Bryer, D.W. Stroup, ad L. Klei-Berdt, FID-assisted idoor localizatio ad commuicatio for first respoders, IEEE Tras. First Europea Coferece o Ateas ad ropagatio, 6- Nov. 6. [] M. Kuh, B. Merkl, A. E. Fathy, ad M. Mahfouz, Accurate UWB idoor localizatio system utilizig time differece of arrival approach, i roc. IEEE adio ad Wireless Symposium, 6. [] M.A. Stelios, A.D. Nick, M.T. Effie, K.M. Dimitris, ad S.C.A. Thomopoulos, A idoor localizatio platform for ambiet assisted livig usig UWB, i MoMM roceedigs 6 th Iteratioal Coferece o Advaces i Mobile Computig ad Multimedia, 8, pp [6] E.S. Kim, The agig erformace of Chirp Spread Spectrum i Idoor Multipath Eviromet, Seoul Natioal Uiversity, 8. [7] Q. Specer, B. Jeffs, M. Jese, ad A. Swidlehurst, Modelig the Statistical Time ad Agle of Arrival Characteristic of a Idoor Multipath Chael, IEEE J. Select. Areas Commu., vol.8, pp.7-6, March. [8] K.J. Soo, A Study o the Locatio Awareess System usig TOA (Time of Arrival) algorithm, M.S. thesis, Kagwo Uiversity, 7. [9] S.Y. Kim, ad J.H. Lee, CSS based idoor localizatio system developmet utilizig patter matchig method, Chugam Natioal Uiversity,. [] S.Y. Kim, ad J.H. Lee, EKF-based Localizatio of a Mobile obot usig atter Matchig, Joural of Automatio ad Cotrol Egieerig, Vol., No., Jauary. Gyumi Hwag studies i the departmet of Mechatroics Egieerig at Chugam Natioal Uiversity, Daejeo, Korea, sice 7. His research iterests iclude ics, idoor localizatio of mobile. Jihog Lee received the B.S degree i Electroics Egieerig from Seoul Natioal Uiversity, Korea i 98, ad the M.S. ad h.d. degrees from the Korea Advaced Istitute of Sciece ad Techology (KAIST), Daejeo, Korea, i 98 ad 99, respectively, all i Electrical ad Electroics Egieerig, sice. His research iterests iclude ics, idoor localizatio of mobile, itelliget cotrol, multi- localizatio ad path plaig. Iteratioal Scholarly ad Scietific esearch & Iovatio 8() scholar.waset.org/7-689/9997

Radar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1

Radar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1 Advaces i Egieerig Research, volume 8 d Iteratioal Coferece o Automatio, Mechaical Cotrol ad Computatioal Egieerig (AMCCE 7) Radar emitter recogitio method based o AdaBoost ad decisio tree Tag Xiaojig,