A field study on terrestrial and satellite location sources for urban cellular networks

Transcription

1 A field study on terrestrial and satellite location sources for urban cellular networks Israel Martin-Escalona, Francisco Barcelo Dept. d Enginyeria Telemàtica de la Universitat Politècnica de Catalunya (UPC) Barcelona, Spain imartin@entel.upc.edu ; barcelo@entel.upc.edu Cyrille Manente Bouygues Telecom RED, Qualité de service Réseaux et Systèmes d information (QRS) Paris, France cmanente@bouyguestelecom.fr Abstract This work presents a statistical field study of the availability of time sources for location in a true wireless network. Terrestrial (base stations) and satellite (GPS) sources were investigated in three different urban scenarios. The density function of a specific number of sources available for triangulation is presented along with other statistical data in order to assess coverage. Since the fusion of terrestrial and satellite sources to obtain the location in wireless networks has been proposed as a way of improving coverage, the joint density function and cross-correlation between the availability of both types of sources are also presented. This correlation depends on the scenario, leading to the conclusion that the improvement obtained through the fusion of sources depends on both the fusion type and the scenario. Keywords GPS; location; coverage; hybridization; cellular network; field study. I. INTRODUCTION Location services in wireless cellular networks are becoming a key service, since they can be both a standalone service (i.e. the user wishes to know his/her position) and at the same time a lower layer for other services and applications (e.g. it is not important to the user to know his/her position, but the service requested requires the position to be provided) [1]. In addition, the key role of location services (LCS) for public safety and emergency purposes obliges regulators to increase quality requirements. Location is also useful to operators beyond considerations of the revenue they can generate from providing it; this is due to the possibility of using location information to optimize the management of the network resources [2, 3]. Examples of such advantages are intelligent paging and intelligent resource allocation for handoff (i.e. the network knows the location, speed and direction of the mobile station, hence it can reserve a channel for handoff in the next suitable cell). Location methods based on triangulation techniques generally provide greater accuracy than those based on information available in the network such as Cell Identification, Timing Advance, Received Signal Strength, etc. However, there is no guarantee that enough sources will be available for triangulation at a certain position. To lessen the impact of this possible lack of resources, several methods have been proposed for combining information from different triangulation systems in a single triangulation. The information concerning the availability of satellites and base stations is essential for evaluating and comparing the performance of triangulation methods for location in cellular networks. References on this topic are scarce [4] and usually either concern old-fashioned communication systems or study them as standalone systems, without considering a possible fusion of location measurements from different systems such as GPS and terrestrial BS. This study addresses the issue and provides current data concerning the availability of GPS satellites and GSM/GPRS base stations for the same common scenarios and positions. The paper is organized as follows. Section 2 provides a brief description of several methods to combine location data from sources belonging to different triangulation systems. Section 3 presents the general assumptions used in the preparation of this study. Section 4 presents the statistical analysis along with remarks on the performance expectations for different combined approaches and scenarios. In Section 5 the main conclusions are summarized. II. TRIANGULATION OF TERRESTRIAL AND SATELLITE SOURCES The necessary accuracy obtained from LCS depends on the specific service or application and can vary from hundreds to only a few meters. Several technologies are available that provide different levels of accuracy and availability [5, 6, 7, 8]. The accuracy of methods based on the received signal strength [9] such as Network Measurement Reports (NMR) is poor, due to the variability of the radio path, fading and other factors. Cell-ID is always available in cellular networks but the accuracy is very poor except for very small cells. Timing Advance (TA) and Round Trip Time (RTT) are easy to use in GSM/GPRS and UMTS respectively, but they suffer from variable receiver chain delays in the Mobile Station (MS). Angle of Arrival (AOA) can be combined with TA or RTT to obtain the location. Combinations of power and time measurements are also possible [10]. Triangulation methods generally provide greater accuracy than non-triangulation methods but they also have less coverage. Notice that Cell-ID, TA/RTT and NMR are always available as long as the MS is connected to the network, while there is no guarantee that three or more Base Stations (BS) or satellites (SAT) are in line-of-sight of the terminal for triangulation. Terrestrial methods such as E-OTD for

2 GSM/GPRS and OTDOA for UMTS work in a similar way: 2D location is possible if three or more BS are in line-of-sight. A-GPS is similar to GPS but it relies on the cellular network to send assistance information to the MS (e.g. almanacs, etc). This assistance gives A-GPS greater sensitivity than GPS and reduces the Time to First Fix (TTFF). However, the working conditions of A-GPS are in general worse than those of GPS navigation: mobile terminals are often in pockets, bags, indoors, etc, while navigation equipment is usually well located with a clear sky view. Performance of A-GPS is excellent in the open field, but poor indoors, in urban canyons and in narrow streets. In addition to the use of one triangulation system as standalone (satellite or terrestrial) three approaches have been presented in which information from two or more systems is combined when available. These are briefly described below (see [11] for further details and references). A. Loose combination The simplest possible hybridization consists of joining the resulting positions from E-OTD/OTDOA and A-GPS. If both positions are available, they can be combined in several ways in order to improve accuracy (e.g. weighted averaged, simple selection of the most accurate, etc); although the combination procedure affects the accuracy of the location estimate, it does not have an impact on the coverage results. To obtain the position, it is sufficient for the position of at least one technique to be provided as standalone. B. Tight non-synchronized combination This approach assumes that the time sources of both triangulation systems are not synchronized. A detailed description of the non-synchronized solution can be found in [12] in which the authors propose Digital Audio Broadcast (DAB) stations instead of GPS satellites, but their proposal can easily be extended to A-GPS and GSM/UMTS. Again, the way in which the measurements are combined or weighted has an impact on the accuracy of the position estimate but not on the coverage. Tight non-synchronized hybridization allows the terminal to calculate its position with two BS and two satellites. C. Tight synchronized combination This method assumes that the terrestrial and satellite networks are synchronized. If so, for triangulation purposes, SAT and BS can be seen as part of the same network. The cost of this synchronization is twofold: more clocks are needed in the LCS equipment for synchronization purposes and additional signaling must be sent to the terminal. Several approaches to transmit this synchronization assistance information between E-OTD and A-GPS have been presented in [13]. The position can be calculated at the terminal with signals received from 3 elements - either BS or SAT [12, 14]. III. ASSUMPTIONS AND SCENARIOS For simplicity, only comments on 2D positioning are included in the paper, but the proposed approach can easily be generalized to 3D positioning. The study provides data related to coverage only. Accuracy is not considered: it is assumed that every single triangulation is able to provide a satisfactory degree of accuracy. Geometric Dilution of Precision (GDOP) is related to accuracy and has not been taken into account in this paper. It is possible that in some situations the number of sources is theoretically sufficient for triangulation, but the geometry of the sources (e.g. aligned sources) does not provide to the minimum required accuracy. Inclusion of statistics about accuracy and GDOP measurements falls outside the scope of this paper and is left for further research. It will obviously add complexity to the presentation of data. Note that remarks on coverage estimates presented in Section 4 and implicit in Table II are upper bounds: i.e. while not having the necessary number of sources in line-of-sight implies that location cannot be determined, the reverse is not always true. Two working hypotheses are adopted: that the device is under the coverage of the operator and that the users are static (i.e. in line-of-sight of at least one BS). The first is a direct consequence of carrying out a study directed at cellular networks: without at least one BS the terminal is not within a cellular network. Users are assumed to be static since the scope of this study is to evaluate sources for location and not the impact of mobility or tracking. Three scenarios are presented: dense urban outdoor, urban masked outdoor and urban medium indoor. The first represents the commonly named urban-canyon scenario, i.e. an environment composed of high buildings (eight or more floors) and narrow streets where severe multipath is present. masked outdoor describes an environment with a high density of BTS where line-of-sight for satellite reception is masked. This blockage is produced when a heavy concrete roof is placed at the main entrance of a building. The indoor scenario represents the restroom of a building, which is surrounded by several windows. Data are gathered at a distance of 10 meters from windows, which constitutes a medium indoor environment. Note that only constrained environments are considered in this study. Even though these scenarios involve a low percentage of the territory share (e.g. rural areas represent most of the territory covered by mobile networks), they represent most of the traffic share. All the measurements presented in this paper were taken in Paris (France). A terrestrial network implemented E-OTD, while A-GPS was used as a satellitebased technique. Data obtained for each scenario are presented below. All experiments were carried out during a short period (a few hours) so that the propagation conditions could be assumed to be stable. The low standard deviations obtained somewhat corroborate this hypothesis. For each scenario the number of experiments is large enough to guarantee a confidence interval of ±0.05 times the mean with a confidence level of 99%. TABLE I. Method Terrestrial Satellite Loose Tight NS Tight S SOURCE REQUIRED BY EACH TRIANGULATION (2D) Needs 3 BS 3 SAT Terrestrial OR Satellite Loose OR (2 BS AND 2 SAT) 3 sources

3 IV. MEASUREMENTS FROM ACTUAL LOCATIONS A. Statistics presented For each scenario, the following results are drawn from the statistical analysis of the data sets. Frequency histograms of the availability of a number of sources in line-of-sight are presented in Figures 1 to 3 and show the joint frequency histogram in order to allow assessment of the coverage provided by fusion methods. The joint probability mass function can easily be derived by normalizing. 1) Summary of relevant probabilities. Table II displays the probability of the availability of a number of sources in line-ofsight. The number of BS from which the terminal is able to receive the signal is N BS while the number of satellites from which a valid pseudo-range is received is N SAT. The last column in this table records the probability of a number of BS independently of the number or satellites being received. Similarly, the last row after each scenario shows the probability of a number of satellites in line-of-sight without accounting for the BS received. For simplicity and for practical reasons, the probabilities of having more than two sources (i.e. enough to triangulate for 2D) have also been gathered. 2) Mean, standard deviation and correlation figures for satellites and BS are presented in Table III. The correlation checks dependency between the availability of sources from different systems and has been calculated as σ 2 xy C xy =, (1) σ xσ y where σ stands for the standard deviation. TABLE II. Dense Outdoor Masked Outdoor Medium Indoor PROBABILITY FUNCTION USED FOR THE COVERAGE CALCULATION NBS = 1 NBS = 2 NBS > 2 N SAT = N SAT = N SAT = N SAT > N SAT = N SAT = N SAT = N SAT > N SAT = , N SAT = N SAT = N SAT > signal reception: most of the time only one serving BS is available: 56.64%. This means that coverage of the terrestrial triangulation as standalone is low: the probability of receiving more than two BS for 2D positioning is 40.12%. On the other hand, the probability of having three or more satellites in lineof-sight reaches 88.26%: A-GPS can attend most of the requests. TABLE III. MEAN, STANDARD DEVIATION AND CORRELATION OF BS AND SATELLITES Mean Std. deviation Correlation Dense BS Outdoor SAT Masked BS Outdoor SAT Medium BS Indoor SAT Figure 1 and Table III show that BS and satellite availability are correlated, i.e. when the reception conditions improve, the number of BS and satellites available increase together. Fusion of measurements is expected to provide only a slight improvement over standalone techniques. It can be explained by the greater coverage of A-GPS and the positive correlation between the availability of BS and satellites (0.34). This means that an improvement in the availability of satellites will lead to an improvement in the availability of BS and vice versa. Consequently, it is likely that E-OTD information will not assist hybridization under conditions in which A-GPS is not available. C. masked outdoor The urban masked outdoor scenario represents an outdoor space covered by a structure that means the sky is not in lineof-sight. The data for this scenario were collected in the access gate of a building covered by a concrete roof. Data in Figure 2 indicate that coverage in this scenario is highly limited when single triangulation techniques are used. Availability of resources is low if compared with the dense urban outdoor scenario. Thus, E-OTD and A-GPS-based techniques may B. Dense urban outdoor This scenario represents an urban-canyon scenario, i.e. an environment composed of high buildings and narrow streets. Figure 1 shows the restrictions imposed by this scenario on Figure 1. BS-SAT joint frequency histogram in dense urban outdoor

4 of single location techniques. It can also be deduced from Table II that the probability of having more than two BS or more than two satellites in line-ofsight (i.e. determining the position with loose combination, as seen above) is less than 30% which is unacceptable for most services. Due to the low positive correlation, it is expected that tight approaches will improve coverage: from Table II it can easily be calculated that the probability of having three sources (of any kind) in line-of-sight is 83%. Figure 2. BS-SAT joint frequency histogram in urban mask outdoor. attend only 25.34% and 31.87% of requests, respectively, as shown in Table II. Fusion of measurements is expected to improve the coverage regardless of the approach used. Table III shows an inverse correlation between BS and satellites available: a decrease in the number of satellites in line-of-sight leads to an increase in the number of BS being received and vice versa. The combination of location measurements benefits from this behavior: lower performance of one location technique (e.g. A- GPS) will involve some degree of improvement in the other (E- OTD). D. medium indoor This scenario has been defined as an indoor environment with no windows and access-gates close by. It should be noted that the performance of A-GPS techniques is poor in this type of environment due to the lack of direct sky view. Data for this scenario were taken in the restroom of a building, with windows and doors at a distance of approximately 10 meters. Figure 3 shows that the availability of resources is lower than in previous environments, thus seriously limiting the coverage Figure 3. BS-SAT joint frequency histogram in urban medium indoor V. CONCLUSION The study of the number of sources available for the application of triangulation methods in urban cellular scenarios shows that it is difficult to guarantee the service availability of a triangulation system as a standalone service. While outdoors the number of satellites in line-of-sight is sufficient to guarantee the service for a high percentage of requests, this percentage decreases dramatically when the signal is masked or when the terminal is indoors. In such cases, combining measurements from terrestrial and satellite systems could lessen the problem of the low number of sources. The benefit obtained through fusion will depend on the sign and the degree of correlation between the availability of both types of sources: a low negative correlation leading to better coverage. For the constrained masked and indoor scenarios described here, this correlation is essentially neutral (i.e. close to zero), thus providing a benefit when combined measurements are used. ACKNOWLEDGMENT This research was partially funded by the EC through the 6 th FP IST Liaison project and by FEDER and the Spanish Government through project TIC REFERENCES [1] C. Drane, M. Macnaughtan, C. Scott, Positioning GSM Telephones, IEEE Communications Magazine, Vol. 36, pp , April [2] D.J.Y. Lee and W.C.Y. Lee, Optimize CDMA System Capacity with Location, in Proc. of IEEE PIMRC 2001, San Diego, USA, pp , October [3] S.S. Wang, M. Green, M. Malkawi, Mobile positioning technologies and location services, IEEE Radio and Wireless Conference, pp. 9-12, Boston MA, [4] T.E. Melgard, G. Lachapelle, H. Gehue, GPS signal availability in urban area Receiver performance analysis, IEEE Position Location and Navigation Symposium (PLANS 94), April [5] D. Kothris, M. Beach, B. Allen, P. Karlsson, Performance Assessment of Terrestrial and Satellite Based Position Location Systems, Proc. of IEE International Conference on 3G Mobile Communications Technology, March [6] D. Porcino, Location of Third Generation Mobile Devices: A Comparison between Terrestrial and Satellite Positioning Systems, IEEE Vehicular Technology Conference 2001, May [7] H. Yin, Location Based Service, T Research Seminar on Location Business II, Helsinki University of Technology, [8] Y. Zhao, Standardization of Mobile Phone Positioning for 3G Systems, IEEE Communications Magazine, pp , July [9] T. Roos, P. Myllymäki, H. Tirri, A Statistical Modeling Approach to Location Estimation, IEEE Trans. on Mobile Computing, Vol. 1, No. 1, pp , January 2002.

5 [10] M. McGuire, K.N. Plataniotis, A.N. Venetsanopoulos, Data fusion of power and time measurements for mobile terminal location, IEEE Trans. on Mobile Computing, Vol. 4, No. 2, pp , March [11] F. Barcelo, I. Martin-Escalona, Coverage of Hybrid Terrestrial-Satellite Location in Mobile Communications, 5th European Wireless Conference: Mobile and Wireless Systems Beyond 3G, pp , Barcelona (Spain), February [12] S. Rooney, P. Chippendale, R. Choony, C. Le Roux, B. Honary, Accurate vehicular positioning using a DAB-GSM hybrid system IEEE Vehicular Technology Conference 2000, pp , Tokyo [13] I. Martin-Escalona, F. Barceló, J. Paradells, Delivery of nonstandardized assistance data in E-OTD/GNSS hybrid systems, IEEE Proc. of the 13th PIMRC, pp , September [14] Y. Zhao, Mobile Phone Location Determination and Its Impact on Intelligent Transportation Systems, IEEE Trans. on Intelligent Transportation Systems, Vol. 1, No. 1, pp , March 2000.