Mobile Location System Using Netmonitor and MapPoint server

Transcription

1 oble Locaton System Usng Netmontor and appont server S. Heman,. Oussalah Unversty of rmngham, Electroncs, Electrcal and Computer Engneerng Edgbaston, rmngham 15 2TT Emal: {.Oussalah, Abstract- Due to emergence of GS moble phones, applcatons nvolvng locaton based systems are becomng very relevant to the newly emergng maret. In ths paper a new low cost approach for moble postonng as well as enhanced functonaltes, ncludng, among others, map buldng, s put forward. The proposal reles on the dentfcaton of base statons usng Netmontor software and cell-dentty technque, whch usng Hata-le propagaton model n conjuncton wth extended Kalman flter, allows us to determne accurate coordnate of the moble staton (S) n terms of lattude and longtude coordnates. The resulted target poston wll be evaluated n terms of accuracy both n terms of varancecovarance matrx as well as absolute postonng error wth respect to GPS postonng taen as a reference. Once postonng s acheved, a.net le applcaton s mplemented to access the crosoft appont Web Servce, va the Internet (http protocol), whch vsualses the moble s locaton on a map, fnds nearby ponts of nterest (POI) and computes routes between them. Ths may ncrease substantally the performances of new 3G wreless networ and moble phones. Experments carred out n ths paper nclude detaled nvestgaton n rmngham area where both urban and suburban envronments have been examned. Also expermental comparsons of varous propagatons models have been carred out. An accuracy of less than 20 meters has been acheved, whch shows substantal ncrease over standard Tmng Advance related technques. I. INTRODUCTION Snce ancent tmes, man adopted varous methods to determne hs geographcal locaton: from usng stars, landmars, verbal queres, to compass and maps n the mddle-ages. The emergence of wreless communcaton technology, n the recent years, gave brth to more advanced postonng technques and locaton based servces (LS). The GS Assocaton descrbes LS as servces that use the locaton of the target for addng value to the servce [8]. The nterest n applyng a postonng technque for provdng a locaton-based servce has been ncreasng due manly to the exponental ncrease n moble phone users and the related maret. Indeed, the possblty of locatng a moble phone n a wreless networ then provdng servces based on ths locaton nformaton s of paramount mportance for both commercal and socal benefts. Emergency servces for example represent one of the most mportant applcatons of LS. Accordng to the Integrated Data Communcatons, Inc (IDC), safety and securty ssues led to a demand for a locaton servce [13]. In effect, emergency call centres have seen an ncreasng number of ther calls ssued from cell phones. In the case where the caller gnores hs/her current locaton, a moble staton locaton servce s of a crucal mportance and utlty. For nstance, n the Unted States, the Federal Communcaton Commsson ordered the creaton of the wreless enhanced 911 (E911) emergency servce n 1996 [4, 13]. Ths mandate requred moble operators to provde the Publc Safety Answerng Pont (the emergency call centre) wth the geographc poston of 911 wreless callers, whch forces the moble postonng wthn 400 ft of ts actual locaton 67% of the tme. Typcally, wthout recourse to external sensors le GPS, accurate postonng of moble statons s not possble n cellular networ, as the only nformaton avalable to the regulator s the Cell Identty (CI) of the servng ase Transcever Staton (TS). In the sequel, such accuracy s ultmately lned to sze of CI, whch can be few hundred meters n case of urban le envronment but t may go up to several mles n case of suburban or rural le envronments. Varous methods and technology usng the networ parameters have been developed to accommodate the need for accurate postonng, manly, through a genune combnaton of nformaton ssued from dverse TSs [2,3,5]. Ths nclude, the tmng advance (TA), whch s based on measurng the round-trp propagaton delay of the sgnal transmtted from the base staton to the handset and bac to the TS [12]. The latter allows the user, gven the speed of rado waves, to determne the dstance between the TS and the S, Therefore, usng several TSs, the locaton of the S s determned as the ntersecton of the underlyng range crcles as llustrated n Fgure 1. Fgure 1 S poston estmaton usng three TA values

2 Other related technologes nclude tme dfference of arrval (TDOA) and angle of arrval (AOA) [6,12]. However, not all such technques are accessble to the users as t may requre changes n exstng networ and termnals, or only because of safety regulatons, some data are not suppled to the users by networ provders. In ths paper, a low cost postonng methodology s developed and put forward wth enhanced functonaltes. The proposal maes use of a networ-montorng software referred to Netontor that s nstalled on the Noa 3310, the handset to be used for performng sgnal strength measurements. Ths tool provdes us wth the nformaton related to connectng base staton as well as neghbourng TSs n conjuncton wth ther assocated tmng advances. Next, a propagaton model as well as a Kalman flter based approaches are employed to accurately determne the postonng of the S n terms of lattude/longtude coordnates. The performances of the system are enhanced by mplementng a connexon to the appont Web Servce, whch enables the user to access several addtonal servces, le map constructon, route tracer, nearby ponts of nterest, etc. Ths maes brdge from cellular networ technology to advanced onlne servce, whch wll surely ncrease n next generaton 3G moble phones [10,11,14]. The experments that have been carred out n both urban and suburban areas n rmngham Cty wll then be presented. For each experment, the moble s poston s computed usng our developed postonng system usng fve dfferent propagaton models. A qualtatve / quanttatve comparson of the results wll be carred out. Secton 2 of ths paper provdes a general overvew of the developed system. Secton 3 emphaszes the dfferent propagaton models used n ths paper. Secton 4 detals the Kalman flter approach. Secton 5 detals some expermental results carred out n rmngham Cty area. II. ETHODOLOGY The process of fndng the locaton of the S usng the cellular networ, and thereby enablng further servces through appont server, nvolves several stages as llustrated n the bloc dagram shown n Fgure 2, whch summarses the tass that consttute the entre postonng methodology. - Frst usng the Netontor software, one determnes the Cell Identty (CI) code of each TS communcatng wth the S. The software also dsplays the receved sgnal strength (Rx) transmtted by each prevously ponted out surroundng TS. ore specfcally, Test 1 n Netontor dsplays nformaton related to the servcng TS whch ncludes, among others, CH (carrer channel number) used to communcate to TS. Ths s used to determne the carrer frequency (f) n Hz of the sgnal transmtted by the TS and receved at the handset as shown by the expresson [12]: f = (CH-511) Ths test also dsplays TA (tmng advance) and Rx (receved sgnal strength) from the servcng TS. On the other hand, the use of Test 2 and Test 3 shows nformaton related to neghbour cells n terms of channel numbers CH and receved sgnal strength Rx. The use of Test 17 allows us to force the hand over by communcatng drectly wth neghbourng cells, or equvalently, neghbourng TSs, and therefore, recordng the assocated TAs and Rxs. Fgure 2. loc dagram for the general methodology - In order to determne the locaton of the TSs, one uses the nformaton avalable from the government webste whch updates the locaton of GS and UTS base statons throughout the whole UK. Usng the nformaton on carrer frequency, and GPS postonng, an offlne nvestgaton of matchng the underlyng TS and those ponted out n the webste s carred. Ths yelds the lattude/longtude postonng of each TS nvolved n the experment. The webste also provdes us wth the heght and the transmtter power (Tx) of each TS, whch s very useful for dstance calculaton. - The dstance between each TS and the handset can be determned usng one of the emprcal propagaton models. Ths conssts of Hata-Oumura model, cost 231 extended Hata s model, Walfsh-Iegam model, CCIR model and ECC model [1,6]. These models, nvolvng path loss measurement, use the nformaton about receved sgnal strengths as well as extra nformaton about the TS, whch

3 are retreved from ether Netontor or prevous matchng process. Some of these models also requre addtonal nformaton regardng the structure of the envronment. An alternatve, but less accurate method, conssts of determnng the dstances drectly from the TAs suppled by Netontor. - Once the S-TS dstance s obtaned for the servcng TS, and each of the neghbourng TS, a statc model of extended Kalman Flter s used to determne the postonng of the S locaton n terms of lattude and longtude coordnates. - A connexon wth appont web servce SDK verson 3.5 enables the access to extra servces le map buldng, route tracer, nearest ponts of nterest, among others. For urban areas the path loss s expressed as follows: Loss = log( f ) 13.82log( a( Where a(h ) s gven by: ( log( ).log( 2 ( log(11.75h ) ) 3.2 a( hm ) = 1.1log( f 0.7) h m 4.97, 1.56 log( f ) 0.8, for small/medum cty for large cty For suburban areas the path loss s expressed as follows: Loss = log( f ) log( a( ( log( ).log( 2( log( f / 28) ) 2 Where a(h ) refers to the small/medum moble gan factor. III. PROPAGATION ODELS Ths secton outlnes the fve propagaton models that wll be used to determne the dstance between a base staton (TS) and the moble phone (S). For detaled explanatons of these models, the reader can consult references [1,6]. Each model provdes an expressng of the path loss of the sgnal transmtted by the TS transmtter (Tx) and receved at the S recever (Rx) as a functon of the dstance between the two enttes (TS and S). The followng parameters wll appear on the path loss equaton of each model: Loss: path loss=tx-rx (. d: S-TS dstance (Km). f: carrer frequency (Hz). h : heght of the base staton (m). h: heght of the moble phone (m). The receved power (Rx) and the frequency (f) of the sgnal wll be measured by the handset usng the networ montorng software Netontor (see Secton 2). Whle the transmtted sgnal power (Tx) and the TS heght can be retreved from the base staton database Stefnder. Therefore, wth the path loss beng determned (Tx-Rx), the S-TS dstance ( can be calculated by solvng the path loss equaton havng d as the only unnown. However, some of the path loss equatons, requre addtonal parameters related to the envronment. The value of each of these can smply be estmated dependng on the expermental area or set to ther default values determned by the propagaton model tself. A. Hata-Oumura odel Ths model gves a path loss expresson for each of the urban, suburban and open envronments. It defnes a parameter a(h ) to be the moble heght correcton factor or gan (. As a result Hata subdvded the urban envronment nto large and small/medum ctes Cost 231 extenson of the Hata odel The path loss equaton n d s: Loss = log( f ) log( a( ( log( h )).log( Cm As n the Hata-Oumura model, a(h ) s the moble gan factor. Cm s another correcton factor. For urban envronments: a(h) = 3.2[log(11.75 h)2] 4.97 and Cm = 3 d For suburban envronments: a(h) =1.1 log(f-0.7) h [1.56 log (f)-0.8] & Cm = 0 Although both the Oumura and Hata and the COST231-Hata models are lmted to S antenna heghts greater than 30 m, they can be used for lower S antenna heghts provded that the surroundng buldngs are well below the S antennas as stated by Stuber n [12]. C. CCIR odel Loss = log( f ) log( a( ( log( ).log( Where: a(h) = 1.1 log(f-0.7) h [1.56 log (f)-0.8] =30-25 log 10 (% of area covered by buldngs) The percentage of the area covered by buldng wll be roughly estmated when usng CCIR to estmate the dstance between the TS transmtter and the moble recever. D. The ECC odel Loss = Afs Afm Gb Gr Where: Afs s the free space attenuaton gven by:

4 Afs = log(20 log (f) Afm s the basc medan path loss gven by: Afm = log(7.894 log (f) 9.56 log(f) 2 Gb s the base staton heght gan factor gven by: Gb =log(h/200).[ (log 2 ] Gr s correcton factor gven by: Gr =[ log(f)].[log (h)-0.585] E. The COST231 Walfsh-Iegam odel L fs Lrts Lmds, f Lrts Lmds 0 Loss = L fs, f Lrts Lmds < 0 Wth, free space loss: f r L fs = 32.44d 20.log( ) 20.log( ) hz m Roof loss: w f hroof hr Lrts = 16.9d 10.log( ) 10.log( ) 20log( ) m hz m Over-roof loss: d f b Lmds = lbsh a d.log( ) f.log( ) 9log( ) m hz m h hroof 18.log(1 ), f h > hroof Lbsh = m 0, f h hroof Wth h r : heght of moble antenna (S), h roof : nomnal heght of buldng roof h : heght of TS antenna b a 54, f h > hroof = h hroof, d. h hroof f , 925 = f , 925 f d 0.5 and h / 0.5, f suburban area f urban area h roof f d < 0.5 and h h The parameters a and b are referred to as the dstance factor and frequency factor respectvely. The former accounts for the ncrease n path loss when the S antennas are stuated below the roof tops of adjacent buldngs. oth parameters control the dependency of the mult-screen dffracton loss on the dstance and frequency, respectvely as stated by Stuber n [12]. roof s appled to calculate the poston of the moble phone (S) n lattude and longtude coordnates. The accuracy of the result depends on how far t s from the locaton where sgnal strength measurements have been taen. Ths s determned usng our GPS devce at the exact locaton of the experment. L Let X be the vector representng the poston of the l moble staton (S) at tme, L beng the lattude and l the longtude. As the dynamc of the S s not taen nto account; that s, one assumes when the S s connected to dfferent TSs, the S locaton s constant. In other words, the state model of the target s gven by X 1 = X. (1) L Let be the vector representng the lattude and l longtude of the th base staton (TS). Let D () be the (nose dstance between the th TS and the S poston at tme ; R s the earth radus ( m). D () s defned usng the sphercal law of cosnes as follows: = a cos[ sn( L ) sn( L ) cos( L ) cos( L ) cos( l l )] ε ( ) (2) D ( ) Where ε () s zero-mean Gaussan nose. (2) corresponds to the measurement model of the flter. The resoluton of the Kalman fler equatons (1-2) leads to the followng teratve algorthm, whch s repeated for each S, gven the ntal varance-covarance matrx P and gven a measurement varance R (a choce of R=0.25 was adopte: T T - Flter gan: K = P. H.( H P H R) 1 - State estmaton update: Xˆ ˆ 1 = X K.( d Dˆ ) Where d s the dstance S- th S determned usng the path loss equaton - Update of state covarance: P ( ) 1 = I K H. P Where H () stands for the Jacoban atrx of Dˆ ( ) (.e., D () wthout nose ε () ). E.g., wth Dˆ L = R 1 sn( L ) cos( L ) cos( l Dˆ ( ) Dˆ ( ) H ( ) =, L l l ) cos( L ) sn( L ) 2 [ sn( L ) sn( L ) cos( L ) cos( L ) cos( l l )] IV. KALAN FILTER ODEL Once dstance measurements are obtaned from at least three base statons, the statc model of the Kalman Flter algorthm Dˆ = R l 1 cos( L ) cos( L ) sn( l l ) 2 [ sn( L ) sn( L ) cos( L ) cos( L ) cos( l l )]

5 The followng pseudo-code descrbes the mplementaton of the algorthm Intalze X as the center of gravty of (=1 to n number of TSs-). Intalze P. For each TS do Calculate the value of Dˆ ( ) usng the lattude and longtude of the current base staton and the S poston X ; Construct the Jacoban matrx H Compute the Kalman gan K usng the current P and H matrces Update the S poston X usng the current K, D and the S-TS dstance. Update the error covarance matrx P usng current K and H matrces. END Loop; Fgure 3. Suburban experment map V. EXPERIENTS The developed software s referred to GS oble Locator has been mplemented n C# usng crosoft Vsual Studo.NET Ths secton presents the testng carred out n rmngham Cty area. In order to llustrate the performances of the algorthms n dfferent scenaros, two types of envronments, a urban envronment near cty centre and a suburban envronment as shown n Fgure 3 and 4 where the locaton of the base statons have been reported. Also, n the same graph, the locaton of the S va GPS coordnate, whch only used for reference, s also reported. The power level (Rx) and carrer frequency (f) of each sgnal, pertanng to each TS, have been measured usng Netontor as shown tables 1 and 2. Fgure 4. Urban experment map A comparson of the applcaton of varous propagaton models n terms of locaton accuracy s shown n Fgure 5 for both suburban and urban envronments. Table I Suburban measurements. TS f (hz) Heght (m) Tx ( Rx ( road Street Speedwell Rd Portland Rd Table II urban measurements. TS f (hz) Heght (m) Tx( Rx ( New Street Temple Row ull Street (s) Corporaton St Fgure 5. Comparson of postonng accuracy wth respect to dfferent propagaton models From the precedng t turns out - The Walfsh-Iegam model gves the best results wth an accuracy of 28 metres n all stuatons. - The ECC 33 model performs better n suburban envronments whle the Hata model s much more accurate n urban envronments.

6 - The CCIR and Extended Hata models provde roughly smlar postonng accuraces n both envronments. In suburban experments, the accuracy of the postonng result s dramatcally affected by the choce of propagaton model used for S-TS path dstances. In urban envronments however, the choce of whch model to use slghtly affects the accuracy of the result. to each TS dentfed by Netontor. Next, an extended Kalman flter model was developed to combne the dfferent dstances n order to determne accurate postonng of the S n terms of lattude and longtude coordnate. Fnally, a connecton to appont server allows the user to access several enhanced servces related to map constructon, dentfcaton of nearby ponts and route tracer, etc. An expermental has been carred out n rmngham cty centre n both suburban and urban envronment whle comparng the performances of the dfferent propagaton models (Hata, Extended Hata, CCIR, ECC33 and Walfsh-Iegam). The Walfsh-Iegam acheved the best accuracy wth 28 metres n both experments, whch sounds a very good result gven that usual accuracy usng TA le technque s more than 500 meters. REFERENCES Fgure 6. Example of user s nterface and dsplay of outputs Fgure 6 shows examples of results dsplayed n the PDA handset and user nterface. Ths shows example of result usng Hata-Oumura modal, outcomes of Kalman flter n terms of state estmaton,.e., lattude, longtude coordnate of the S and the assocated varance-covarance matrx. Fnally, n Fgure 6c and 6d, dsplay of nearby places as well as route planner are ponted out. VI. CONCLUSION In ths paper, a new low cost methodology for moble postonng usng GS standard together wth enhanced servces has been put forward. The proposal uses Netontor software avalable n most Noa handsets to dentfy the channel number, cell dentty and receved sgnal strength pertanng to the emttng base transmtter staton as well as each of neghbourng base statons. A matchng wth government database of TS allows us to dentfy the locaton and characterstc of each underlyng TS. The approach uses a wave propagaton model to calculate the dstance from S [1] V.S. Abhayawardhana, I.J. Wassell, D. Crosby,.P. Sellars,.G. rown, Comparson of Emprcal Propagaton Path Loss odels for Fxed Wreless Access Systems, n: Proceedngs of the 61 st IEEE Vehcular Technology Conference, Stocholm, Sweden, 2005, p [2]. Al, Survey of Current Locaton and Postonng Technques, n: Proceedngs of the Frst European Worshop on Locaton ased Servces (LS-2002), Sept., 2002, [3] T. D Roza T and G. lchev, An overvew of locaton-based servces. T Technology Journal 21(1), 2003, [4] C.C. Docet, Revson of the Commssons rules to ensure compatblty wth enhanced 911 emergency callng systems, R- 8143, Rep. No , FCC, 1994 [5] C. Drane,. acnaughtan and C. Scott, Postonng GS Telephones, IEEE Communcaton agaznes, 36(4), 1998, [6] K Fujmoto, J. R. James, oble Antenna Systems Handboo, Artech House, NY, 2 nd Edton, 2001 [7] G.Gunnarsson,. Allen, T. Rantalanen, V. Ruutu and V-. Tettnen, Locaton Tral System for oble Phones, n: Proceedngs of the IEEE Global Telecommuncaton Conference. The rdge to Global Integraton, 1998, [8] A. Kupper, Locaton-based Servces: Fundamentals and Applcatons, John Wley and Sons, NY, [9] K. Raja, Crtcal Analyss and odellng of Locaton Fndng Servces, Justfone plc and Naper Unversty, sc Thess, [10] D. Ralph, S. Searby, Locaton and Personalsaton: Delverng Onlne and oblty Servces, T Communcatons Technology Seres 8, The Insttuton of Electrcal Engneers, London, [11]. Saraya, Geographc Locaton n the Internet, Kluwer Academc Publsher, oston, [12] G. L Stuber, Prncples of oble Communcaton, Kluwer Academc Publshers, NY, 2002 [13] S.C. Swales, J.E. aloney and J.O. Stevenson, Locatng oble Phones and The US Wreless E-911 andate, n: Proceedngs of IEE Colloquum on Novel ethods of Locaton and Tracng of Cellular obles and Ther System Applcatons, 1999, p.2/1-2/6. [14] Tmo Halonen, J. Romero, J. elero, GS, GPRS and EDGE Performance: Evoluton Towards 3G/UTS. John Wley and Sons Publsher, UK, 2003 [15] D. Tse, P. Vswanath, Fundamentals of Wreless Communcaton, Cambrdge Unversty Press, UK, 2005.