Coverage of Hybrd Terrestral-Satellte ocaton n Moble Communcatons Francsco Barceló, Israel Martín-Escalona Dept. d Engnyera Telemàtca de la Unverstat Poltècnca de Catalunya c/ Jord Grona 1-3, Barcelona 08034, Span barcelo@entel.upc.es Abstract: Ths work studes the mprovement n servce coverage obtaned by three dfferent ways of hybrdsng (terrestral and satellte) trangulaton locaton methods for cellular networks. Though the authors assume that terrestral cellular networks use Enhanced Observed Tme Dfference (E-OTD) n 2G or Observed Tme Dfference Of Arrval (OTDOA) n 3G, and that the satellte GNSS uses Asssted Global Postonng System (A-GPS), ther analyss can easly be generalzed to address any other trangulaton method. A smple analytcal model s presented, whch s used for evaluatng the servce coverage of each approach. The numercal results show how hybrdsaton leads to a hgh mprovement and an easy balance between traffc and geographcal coverage. 1. Introducton In the last decade we wtnessed the growth of publc moble telephony at rates much hgher than forecasted. Nevertheless, after ths phase, many experts foresee that conventonal (.e. voce and data) servces may reach saturaton soon. Gven ths stuaton, operators are lookng for new servces. Provdng a wde range of applcatons that nject data traffc nto the network s seen as a promsng approach [1. In ths new scenaro, locaton s a key servce. It s offered as a standalone servce and at the same tme serves as a lower layer for other servces and applcatons (e.g. the user does not need to know hs poston, but the servce he requested needs the poston n order to be provded). In addton, the key role of locaton servces (CS) for publc safety and emergency purposes leads regulators to enhance the requrements for qualty. ocaton s also useful to operators beyond smply what revenue they can get from provdng t; ths s due to the possblty of usng locaton nformaton to optmse the management of network resources [2, 3. 1.1. The need for hybrdsaton The necessary accuracy obtaned from CS can vary from several to hundreds of meters. Several technologes are currently avalable that provde dfferent levels of accuracy and avalablty [4, 5, 6, 7. Methods based on receved power [8 such as Network Measurement Reports (NMR) have a poor accuracy due to the hgh varablty of the rado path, fadng, etc. Cell-ID s always avalable n cellular networks but the accuracy s poor snce the cell radus can be long for certan scenaros. Tmng Advance (TA) and Round Trp Tme (RTT) are easy to use n GSM/GPRS and UMTS respectvely, but they suffer from varable recever chan delays n the MS and the precson of the TA s not very good. Ths paper studes the hybrdsaton of trangulaton methods that n general provde a better accuracy than non-trangulaton ones. However, the latter also have less coverage: note that Cell-ID, TA/RTT and NMR are always avalable as long as the MS s connected to the cellular network, whle there s no guarantee that three or more Base Statons () or satelltes wll be wthn sght of the Moble Staton (MS) n order to trangulate at the specfc moment that a locaton request s launched. Terrestral trangulaton methods such as E-OTD for GSM/GPRS and OTDOA for UMTS work n a smlar manner: 2D locaton s possble f three or more s are n sght. Ths s almost guaranteed n densely populated areas but seldom occurs n rural coverage, as the dstances between s are typcally long. A-GPS s smlar to GPS but t reles on the cellular network to send assstance nformaton to the MS (e.g. almanacs, etc). Ths assstance nformaton greatly mproves the tme to track to satelltes and hence the battery consumpton. The performance of A-GPS s excellent n the open feld, but poor ndoors, n urban canyons, narrow streets and near to tall buldngs; n all these scenaros the vew of the necessary number of satelltes s not guaranteed. Ths suggests that combnng both technques should have the consequence of mprovng coverage: only rural ndoor stuatons should show locaton problems [9. 1.2. Goals The am of ths paper s to compute the coverage of hybrd Terrestral/Satellte locaton methods. The hybrdsaton s assumed to be carred out at the MS (.e., s handset based): the MS computes ts poston wth the avalable nformaton from E-OTD/OTDOA
and/or A-GPS. However, the man conclusons of ths work do not change for other approaches. No further hybrdsaton s consdered n ths work: although some of the methods mentoned above could be added for hybrdsaton, ther accuracy s poor when compared wth that of trangulaton methods. However, the procedure descrbed can be easly generalzed to encompass the hybrdsaton of more than two technques. 2. Assumptons and notaton In ths paper, coverage (C) s defned as the probablty of a locaton technque or a combnaton of technques beng rado-accessble to network resources and termnal at a certan locaton request (.e. traffc coverage) or place (.e. terrtory coverage). In order to make a mathematcal analyss feasble, the defnton of coverage n ths work s slghtly dfferent from the 3GPP defnton of coverage [10. Frstly, ths study evaluates the probablty of the MS beng rado-covered by the servce whle 3GPP also takes nto account possble unavalablty caused by other constrants (e.g. network resources, sgnallng, etc.). Secondly, 3GPP consders only geographc coverage whle ths study extends the concept to traffc coverage. Consequently, there are two fgures to be measured that are related to coverage: Traffc coverage (CT): Proporton of locaton requests correctly answered. Geographc coverage (CG): Proporton of the terrtory covered by the servce. In all cases, as t s assumed that the MS s wthn the coverage range of a network operator, so that the sgnal from at least one s always receved. As both E- OTD/OTDOA and A-GPS need cellular coverage to be provded (.e. both receve assstance through the cellular network), t makes no sense to compute the servce coverage outsde the coverage range of the cellular network. Note that 100% geographcal coverage means that a locaton technque can be provded n all the terrtory covered by the cellular network, assumng that both the network and the termnal are avalable and have suffcent resources. We defne sx envronments: urban/suburban/rural, each of whch s pared wth ndoor/outdoor. Ths classfcaton s suffcently precse to llustrate the procedure and thus acheve the numercal results n Secton 4. For a specfc envronment (), users and the traffc they generate are assumed to be unformly spread along the terrtory. It s assumed that f a user s covered by the locaton servce (.e. there s enough locaton nformaton from s and/or satelltes) hs or her poston can be correctly computed (.e. error-free assumpton). Hence C CT CG, (1) where the ndex ndcates the specfc envronment (e.g. urban ndoor, etc). For the whole network, the unformty assumpton s far from beng realstc: traffc and geographcal coverage fgures are related by the geographcal dstrbuton of traffc n dfferent envronments, as follows: CT t C ; CG g C, (2) network network where t and g stand for the traffc and area shares of envronment respectvely. The number of s from whch the MS s able to receve the sgnal s N, whle the number of satelltes from whch a vald pseudo-range s receved s N. The probablty densty functon (PDF) of recevng the sgnal from a specfc number of s s known for each envronment. The same can be sad for the probablty of seeng a specfc number of GPS satelltes. In all cases, the lowest level of accuracy of A-GPS and E- OTD s assumed to be enough to provde the MS locaton. For smplcty, only the case of 2D postonng s studed n ths paper. Nevertheless, the proposed approach could be easly generalzed to encompass 3D postonng. 3. Hybrdsaton methods: computng coverage 3.1. E-OTD and A-GPS as standalone For 2D postonng, the E-OTD/OTDOA coverage s equal to the probablty of carrer sgnals beng receved from at least 3 s. Hence CE OTD Pr( N > 2). (3) In a smlar manner, the A-GPS coverage can be computed as C GPS ( N ) Pr > 2. (4) For 3D locaton (.e. ncludng heght), the number of necessary sgnal sources changes from 3 to 4 for both technques. 3.2. oose hybrdsaton The smplest possble hybrdsaton conssts n jonng the resultng postons from E-OTD/OTDOA and A-GPS. If both postons are avalable, they can be combned n several ways n order to mprove accuracy (e.g. weghted-averaged, smple selecton of the most accurate, etc); although the combnaton procedure affects the accuracy of the locaton estmate, t does not have an mpact on the coverage results. To determne the poston, t s suffcent f at least one poston as
determned by a standalone technque s provded. Hence [( N < 3) ( N 3) C 1 Pr <, [( N < 3) ( N 3) C 1 Pr <. (5) 3.3. Tght non-synchronzed hybrdsaton Tght hybrdsaton jons tme measurements from ncomng sgnals nstead of from the resultng postons. Non-synchronzed hybrdsaton assumes that there s no tme synchronzaton between the terrestral and satellte networks. Fgure 1 llustrates ths approach, n whch GPS satelltes are all synchronzed and terrestral are also synchronzed wth one another, but there s no synchronzaton between the GPS and the terrestral networks. In ths scenaro, a mnmum of two sgnal generators n each network (.e. 2 satelltes or 2 s) are necessary to determne the poston of an MS. Each par of sgnal transmtters traces an ellpse and the ntersecton between these two ellpses ndcates the poston of the MS. A detaled descrpton of the nonsynchronzed soluton can be found n [11, n whch the authors propose usng Dgtal Audo Broadcast (DAB) statons nstead of GPS satelltes. However, ther proposal can be easly extended to GPS and GSM/UMTS. Agan, the way n whch the measurements are combned or weghted has an mpact on the accuracy of the poston estmate but not on the coverage. Tght non-synchronzed hybrdsaton allows the MS to compute ts poston usng two and two satelltes. Thus, coverage can be computed as follows: [( N 2) ( N 2) C C + Pr, NS [( N 2) ( N 2) C C + Pr. (6) NS 3.4. Tght synchronzed hybrdsaton Ths method assumes that the terrestral and satellte networks are synchronzed. If ths s the case, for trangulaton purposes, satelltes and s can be seen as belongng to the same network. Fgure 2 llustrates ths approach and shows how global synchronzaton allows any combnaton of sgnal generators to be used to trace locaton ellpses. Therefore, the poston can be computed at the MS based on sgnals receved from 3 elements - ether s or satelltes [11, 12. The cost of ths synchronzaton s twofold: more clocks are needed n the CS equpment for synchronzaton purposes (ths cost has a mnor mpact snce t s possble to use clocks that are already avalable n the ) and addtonal sgnallng must be sent from the network to the MS. Several approaches for transmttng ths synchronzaton assstance nformaton between E-OTD and GPS have been presented n [13. There are only three combnatons of the number of receved and satelltes () that lead to a lack of coverage: 1 and 0, 1 and 1, 2 s and 0 (0 s s not consdered snce we assumed that the MS s always covered by the cellular network). In all the remanng cases, the poston of the MS can be provded. The coverage can be computed as follows: S Pr( + > 2) Pr ( N 1 N 2) ( N 2 N 1) ( N 1 N 1 ). C N N > > In a dfferent way, ths equaton can be rewrtten as: Pr( 1) Pr( 2) ( N ) ( N ) C C + N N + S NS + Pr 2 Pr 1 (7) Fgure 2: ocaton process n the Tght Synchronzed hybrdsaton approach 4. Numercal results Fgure 1: ocaton process n the Tght Non-Synchronzed hybrdsaton approach. Ths secton provdes numercal results to llustrate the consequences of the proposed analyss. Table I shows the PDFs of several scenaros of havng a
specfc number of /GPS sources n sght. The data presented n Table I are reasonable hypotheses based on the authors' experence (unfortunately, feld data are scarce n what lterature s avalable). It must be noted that the poor performance of A-GPS n Table I s due to the constrants of the postonng of the MS, whch s often wthn pockets, bags, urban canyons, etc. Some workng hypotheses follow: The terrestral network s GSM/GPRS wth E-OTD. The PDF for E-OTD does not change from outdoors to ndoors, although a slght coverage reducton should be expected n a true network. GPS coverage s strongly reduced from outdoors to ndoors. The average number of s n sght to provde E- OTD coverage decreases from urban to suburban and rural. Ths agrees wth the operators' practce of densely coverng urban areas n whch heavy traffc s expected. Table I. Probablty of sgnal avalable from a gven number of s (for E-OTD) and satelltes (for A-GPS) vs. scenaros. N 1 2 >2 N 0 1 2 >2 Urban Outdoor 0.10 0.20 0.70 0.00 0.10 0.30 0.60 Urban Indoor 0.10 0.20 0.70 0.20 0.30 0.30 0.20 Suburban Outdoor 0.10 0.30 0.60 0.00 0.10 0.20 0.70 Suburban Indoor 0.10 0.30 0.60 0.10 0.30 0.30 0.30 Rural Outdoor 0.30 0.40 0.30 0.00 0.00 0.20 0.80 Rural Indoor 0.30 0.40 0.30 0.10 0.30 0.30 0.30 Table II shows the results of applyng Equatons (3)- (7) to each envronment. Note the mprovement obtaned through hybrdsaton, especally when one or both methods are not able to provde a good coverage as standalone. One must also note the ncrease n coverage for the rural ndoor envronment, for whch both technques exhbt poor performance: coverage reaches 84% whle each method has 30% as standalone. Table II. Computed coverage for each hybrdsaton method vs. scenaros. C E-OTD C GPS C C NS C S Urban Outdoor 0.70 0.60 0.88 0.94 0.99 Urban Indoor 0.70 0.20 0.76 0.82 0.91 Suburban Outdoor 0.60 0.70 0.88 0.94 0.99 Suburban Indoor 0.60 0.30 0.72 0.81 0.93 Rural Outdoor 0.30 0.80 0.86 0.94 1.00 Rural Indoor 0.30 0.30 0.51 0.63 0.84 To compute the overall network coverage, the results n Table II must be weghted-averaged accordng to Equatons (2). The traffc share of each envronment dsplayed n Table III assumes that n all envronments 50% of calls take place ndoors. For the terrtory share, t must be taken nto account that, n urban areas, outdoor coverage (.e. streets) represents a smaller percentage than ndoor coverage (.e. offces, houses). In rural envronments, outdoor coverage (.e. open felds) represents the bgger share. Table III. Traffc (t ) and geographcal (g ) proporton of each envronment n the network. Traffc (t ) Terrtory (g ) Urban Outdoor 0.35 0.02 Urban Indoor 0.35 0.08 Suburban Outdoor 0.10 0.10 Suburban Indoor 0.10 0.10 Rural Outdoor 0.05 0.60 Rural Indoor 0.05 0.10 1.00 1.00 Fgure 3 dsplays the network coverage (.e. for the whole cellular network). An nterestng result n Table IV s that the three hybrdsaton methods studed n ths paper tend to equal the traffc and geographcal coverage for CS n comparson wth the nonhybrdsed case. For conventonal voce and data servces, the geographcal coverage s always lower or much lower than the traffc one, snce deployment starts n the most populated ctes. Nevertheless, f terrestral/satellte hybrdsaton s used, ths msmatch s no longer an ssue for CS. Coverage 1.0 0.8 0.6 0.4 0.2 0.0 Traffc Terrtory 5. Concluson E-OTD A-GPS oose Non- Sync Fgure 3. Computed servce coverage. Sync The hybrdsaton of terrestral and satellte postonng systems greatly mproves servce coverage, wth mprovements occurrng even for the smplest hybrdsaton method (.e. when the resultng postons of both methods are combned). More complex hybrdsatons are made possble by combnng tmng measurements from the terrestral and satellte sources. The latter approaches provde better coverage, as they offer the best performance for synchronzed hybrdsaton. The costs assocated wth optmum
coverage are the hardware (.e. clocks) and sgnallng (.e. further synchronzaton assstance must be sent to the MS). Another remarkable aspect s the smlarty between the traffc and geographcal coverage of the servce, whch contrbutes to farness and to fulfl the regulators requrements. Acknowledgments Ths research was partally funded by the EC through project IST 2000-2604 (EMIY) and by the Spansh Mnstry of Scence through project TIC2003-01748 (RUBI). The authors thank all ther colleagues n EMIY and RUBI for ther valuable comments. REFERENCES [1 C. Drane, M. Macnaughtan, C. Scott, Postonng GSM Telephones, IEEE Communcatons Magazne, Vol. 36, pp. 46-59, Aprl 1998. [2 D.J.Y. ee and W.C.Y. ee, Optmze CDMA System Capacty wth ocaton, IEEE PIMRC 2001, San Dego, USA, pp. 17-21, October 2001. [3 S.S. Wang, M. Green, M. Malkaw, Moble postonng technologes and locaton servces, IEEE Rado and Wreless Conference, pp. 9-12, Boston MA, August 2002. [4 D. Kothrs, M. Beach, B. Allen, P. Karlsson, Performance Assessment of Terrestral and Satellte Based Poston ocaton Systems, IEE Internatonal Conference on 3G Moble Communcatons Technology, March 2001. [5 D. Porcno, ocaton of Thrd Generaton Moble Devces: A Comparson between Terrestral and Satellte Postonng Systems, IEEE Vehcular Technology Conference 2001, May 2001. [6 H. Yn, ocaton Based Servce, T-109.551 Research Semnar on ocaton Busness II, Helsnk Unversty of Technology, 2002. [7 Y. Zhao, Standardzaton of Moble Phone Postonng for 3G Systems, IEEE Communcatons Magazne, pp. 108-116, July 2000. [8 T. Roos, P. Myllymäk, H. Trr, A Statstcal Modelng Approach to ocaton Estmaton, IEEE Trans. on Moble Computng, Vol 1, No 1, pp. 59-69, January 2002. [9 S. Solman, P. Agashe, I. Fernandez, A. Vayanos, P. Gaal and M. Oljaca, GpsOne TM : A hybrd poston locaton system, IEEE 6 th Internatonal Symposum on Spread Spectrum Technques and Applcatons, volume 1, pp. 330 335, 2000. [10 3GPP TS 22.071: Techncal Specfcaton Group Servces and System Aspects; ocaton Servces (CS); Servce descrpton; Stage 1. March 2003. [11 S. Rooney, P. Chppendale, R. Choony, C. e Roux, B. Honary, "Accurate vehcular postonng usng a DAB-GSM hybrd system" IEEE Vehcular Technology Conference 2000., pp. 97 101, Tokyo. May 2000. [12 Y. Zhao, Moble Phone ocaton Determnaton and Its Impact on Intellgent Transportaton Systems, IEEE Trans. on Intellgent Transportaton Systems, Vol 1, No 1, pp. 55-64, March 2000. [13 I. Martn-Escalona, F. Barceló, J. Paradells, "Delvery of non-standardzed assstance data n E-OTD/GNSS hybrd systems", IEEE PIMRC 2002, pp. 2347-2351, September 2002.