LOCATING MOBILE PHONES & THE US WIRELESS E-91 I MANDATE

Transcription

1 LOCATING MOBILE PHONES & THE US WIRELESS E-91 I MANDATE S.C. Swales, J.E. Maloney & J.O. Stevenson' Abstract: The first commercial cellular phone systems were deployed in the early 1980s. Since that time there has been an almost global revolution in the development of wireless communications products. By contrast, the means to actually locate wireless transmitters has been available for decades. Dominated by defence applications, the benefits of applying this technology in the cellular world have only just begun to be realised. This is particularly so in the US, where the Federal Communications Commission (FCC) has mandated that cellular and PCS carriers must be able to locate mobile phones making emergency 91 1 calls with an accuracy of better than 125m rms by October KSI, a US based company, have been developing the means to locate mobile phones in an emergency since Based upon years of localisation and tracking experience in defence related work, the basic technology exploits angle-of-arrival (AOA) techniques. Back in 1991, KSI were the first to deploy a system (comprising two sites) that could locate and track an analogue AMPS phone. Currently, KSI have a four-site system that is capable of locating both analogue AMPS and digital TDMA (IS-136) phones. A larger scale field trial is planned for later this year. This contribution describes KSI's technology and provides a sample of the current field trial results. Preliminary data indicates an accuracy of better than 7om rms. I Introduction I.I Background The localisation of mobile phones has been technically feasible for many years, but has yet to experience large-scale introduction and market acceptance. However, with the advent of the FCC mandate for wireless Enhanced 91 1 (E-911) service in 1996, the need to locate wireless callers dialing the emergency 91 1 number is now a requirement for Commercial Mobile Radio Service (CMRS) providers in the US. The specified objective is to provide ubiquitous location coverage for all wireless 91 1 callers within 125m rms by October Since the E-911 mandate, the wireless industry has witnessed a significant growth in the number of companies purporting to provide a wireless location solution [I]. From a purely technological standpoint, there are a variety of options that are on offer today. In a generic sense, these are sometimes referred to as a location determination technology (LDT) or position determining entity (PDE). Each LDT offers certain advantages and disadvantages, especially when considered in the Cellular/F'CS wireless context. When assessing these technologies, the important performance criteria to consider include the accuracy, delay and cost of providing a location. The air interface that must be supported and the propagation characteristics of the operational environment affect all of these parameters. 1.2 Paper Outline With all the hype surrounding the E-911 mandate and the potential for additional location based services (LBS), it is sometimes difficult for the CMRS provider to determine the solution, or solutions, that best meets their location requirements. In the following section (Section 2), a range of LDT options are discussed and compared before focussing in more detail on an infiastructure approach exploiting angle-of-arrival (AOA). One of the most important factors impacting the performance of any LDT is the mobile radio propagation environment. This is addressed in section 3, focussing in particular on the impact of multipath on an AOA solution. Section 4 describes KSI's field trial system and provides a sample of some current results. The final section (Section 5) concludes with a discussion on the state of the wireless location market today. 2 Location Determination Technology 2.1 Mobile Based For the most part, mobile based solutions exploit an existing localisation system, namely the Global Positioning System, or GPS. Deployed by the US government for navigation purposes, the system enables a terrestrial based receiver to determine its location and speed fiom the signals transmitted by a network of satellites. In order for a cellular mobile to exploit this system, the handset must be modified to include an additional antenna and receiver for the GPS signals. KSI Inc., 7630 Little River Turnpike, Annandale, VA 22003, USA; Tel: C ; Fax: ; sswales@ksix.com; Web: The Institution of Electrical Engineers. Printed and published by the IEE, Savoy Place, London WC2R OBL, G

2 The GPS receiver measures the time of arrival (TOA) of the signal from at least three satellites in order to calculate its position. Whilst this offers the potential for extremely accurate location estimates (down to a few metres), there are a number of issues to consider. The first is the fact that the accuracy for commercial applications, as opposed to its military use, is deliberately reduced through something called selective availability. However, this can be mitigated through the introduction of differential GPS (DGPS). This works by having a network of fixed GPS receivers at known locations that provide the mobile receiver with differential corrections through an additional terrestrial wireless communications link. Even with DGPS, this approach suffers from a number of drawbacks. From a RF viewpoint, the receiver accuracy is limited when the satellite signal is blocked, e.g. when operating inside a vehicle, in buildings, under foliage or simply when in an urban downtown city centre. This can be mitigated to some extent by applying even more network assistance, e.g. by aiding handset reception through additional synchronisation cueing and by only having the handset compute the pseudo-ranges to the satellites and passing the results to the network for more complete processing. Another handset based approach requires the phone to measure the time-of-arrival (TOA) of signals transmitted by multiple fvred sites, e.g. the cellular base-stations (BSs). By calculating the difference in the TOAs from three or more sites, the phone is able to determine a location from the intersection of the resulting hyperbolas (two sites producing one hyperbola). This technique is known as observed time difference, or OTD. As with GPS, the need to modify the phone will impact the design from a power consumption and complexity standpoint. This has special significance for the E-91 1 service since it is intended for all phones all of the time. Given that there are over 60 million unmodified handsets in the US today, this presents a significant barrier to the introduction of handset based technology. m N n 1 1 Base Station L Figure 1: TeleSentinel AOA system architecture for E-91 1 support. PSAP 2.2 Network Based The alternative approach is to exploit the existing cellular network of base-stations by placing additional sensors at these sites. By measuring various characteristics of the existing signals, thereby not requiring modifications to the handset, a location can be calculated. One solution is time based and operates in the same way as the handset based OTD except the measurements are taken at the fixed sensor sites. Commonly referred to as time-difference-of-arrival (TDOA), the solution requires three or more sites to produce a 212

3 location. Since the solution is measuring time, the timing resolution and, hence, performance is limited by the bandwidth of the signal. An alternative to TDOA measures the angle-of-arrival (AOA) of the signal. By having two or more sites calculate a line-of-bearing to the phone, the system is able to calculate a location from the intersection of these lines. One of the major advantages of this approach over TDOA is that its performance is not fundamentally limited by the signal bandwidth. The main disadvantage when compared with TDOA is the need to employ a phased-may antenna. A pattern-matching network approach measures the signals received with a phased-array antenna at the BS site and compares the observations with previously measured signal patterns from known locations. Matching the signal patterns produces a location estimate. To date, this approach has only been tested in dense multipath environments that introduce a location-dependant uniqueness to the resulting signals. 2.3 TeleSentineP Solution In order to be able to provide robust and accurate determinations of the location of standard unmodified mobile phones of any type, KSI selected AOA as the most suitable technology. The phones can be transmitting either controvaccess-channel or voicekaffic-channel signals in the format of any air interface. In the order of increasing signal bandwidth, such signals can be of the pure-carrier, NAMPS, ESMR (iden), AMPS, TDMA (IS-136), GSM, CDMA (IS-95) or 3G types. Known as TeleSentinel, KSI s wireless location system can provide the location using registration, call set-up, or other protocol transmissions, and can provide successively refined and updated locations using voicehaffic channel transmissions. Through this flexibility and generality of applicability, the CMRS carrier is provided with a simple, cost-effective means to meet the FCC mandated E-91 1 requirements and to service commercial location-based applications. Avoiding the shortcomings of alternative location technologies, TeleSentinel provides these capabilities without the complexity of maintaining inter-site data-tagging synchronization to within a fraction of a microsecond, without large inter-site backhaul data transfers, and without requiring the modification or replacement of any legacy mobile units. Figure 1 shows the generic TeleSentinel system configuration. The architecture applies a network-ofstars topology interconnecting each of its Central Stations (CSs) with multiple Sensor Stations (SSs). Each SS determines information related to the AOA for one or more wireless phones of interest by exploiting phase relationships in the received signals fi-om these phones. The resultant AOA related data are then transferred to the CS, where the data fkom two or more SSs are combined for an integrated determination of the phone s location. After determining the phone s location, TeleSentinel provides the position information to the appropriate client system, such as a Public Safety Answering Point (PSAP) in the case of a 911 call. KSI has designed the generic TeleSentinel network architecture to parallel that of the existing wireless communications system, and thus the SSs may be deployed with the associated Base Stations (BSs) while the CSs are deployed with the associated Mobile Switching Center (MSC). Depending on the operator s specific requirements, the TeleSentinel configuration may be implemented either as an overlay on the wireless communications system or may be partially or fully integrated with an advanced implementation of the communications system facilities, achieved through cooperative engineering and licensing arrangements. This offers an extremely cost effective means to deploy a wireless location network, with existing or evolving standards interfaces providing the necessary access to KSI s location data. 3 Environmental Considerations 3.1 RF Propagation In any mobile radio system, the transmitted RF signal is subject to a number of propagation effects before reception. These include a strong path loss component and two significant fading components. Path loss is due to the physical separation of the radios whilst the fading is caused by the diffkaction and reflection of the signal energy due to the terrain and any natural or man-made obstructions. The two fading components are generally referred to as fast and slow fading. The fast fading component is the result of multiple reflections of the same signal combining at the receiver, i.e. multipath. This results in rapid fluctuations of the received signal level that generally follow Rayleigh statistics in a typical urban environment. The slow fading, otherwise known as shadowing, is due to the presence of buildings and other obstructions between the transmitter and the receiver. This results in slower 213

4 variations in the signal strength and these tend to observe a log-normal distribution around the mean received signal level. The impact of these propagation effects on a mobile radio system has been the subject of extensive study and research over the past years. In terms of wireless location, the most significant effect due to signal propagation is multipath, with each path arriving at different times and angles. The different arrival times are normally characterised by a parameter know as the delay spread. Clearly, in any time-of-arrival system, e.g. TDOA, this delay spread can be a significant source of location errors. 3.2 Angular Spread The angular spread of the signal will affect an AOA system in a similar manner to delay spread with TDOA, with the uncertainty in angle introducing variations between the measured bearing and the true lineof-sight (LOS) bearing. In fact, in most urban localities, there could be a complete absence of a LOS signal and so the bearing estimate is based upon multipath reflections only. Many studies have been carried out in the past to look at this phenomenon in a number of radio transmission systems. With the advent of smart antennas for base-stations in cellular systems [Z], more recent studies have also considered the effect in a cellular mobile radio environment. In general, with a well sited base-station antenna, the spread in the angle-of-arrival of the multipath signals from a single mobile source is about 1 to 2O with the source located in excess of 1.5 miles from the receiver. This translates into a so called scattering volume around the mobile transmitter, which has a diameter of between 50m and 100m. In these situations, the system can be said to be multipath limited [3,4]. In many locations, multipath will not be the dominant factor affecting the perfmance of an AOA system. For example, in more rural areas the larger separation of BS sites will mean that the level of the received signal limits system performance, i.e. the system can be said to be noise limited. In both cases the geometric relations of the hdamental AOA measurements to the derived positions can be simply visualized as the transverse spread of the angular uncertainty subtended at the radial distance from the sensor position to the estimated position of the signal transmitter. The uncertainty spread in position is the product of the radial distance from the sensor multiplied by the uncertainty in measured AOA. Thus the AOA measurement uncertainty directly determines the accuracy of a position derived from an AOA measurement. The theoretically achievable accuracies for signal analysis and other statistical analysis processes are related to the parametric likelihood function for the distributions involved, and are commonly known as the Cramer-Rao bounds (lower limits). For an implementation of the TeleSentinel technology using AOA for the localization of standard wireless communications signals, the AOA Cramer-Rao accuracies and their relations to the signal analysis conditions and parameters are described in US Patent Number 4,728,959 [SI. In this patent description, it was shown that, with the simplest of antenna element configurations, the derived angle of signal arrival can be related to the phase-difference relation between the paired signals detected with a single antenna element pair. Thus the achievable angular accuracy can be directly related to the accuracy with which the signal phase differences can be determined. Along with other terms, the phase difference accuracy is shown to be approximately proportional to the inverse square root of the signal-to-noise ratio (SNR) as well as to the inverse square root of the time-bandwidth product of the processed signal bandwidth multiplied by the processing integration time. For elucidation of these accuracy relationships, the patent provides a discussion of a specific example with an in-band SNR of 12 db, an AMPS/DAMpS processing bandwidth of at least 20 khz, a processing integration time of at least 50 ms, and an inter-element separation of one half wavelength. Under these conditions, the angular accuracy is seen to be 0.3 degree with the worst case dilution of directional precision by a factor of two. Then, for a distance of 10 km (6.2 statute miles), the patent description relates this angular accuracy to a transverse position accuracy of 50 m. Generally, the forms of discussion included and referenced above indicate that, if communications can succeed, then AOA-based localization can also succeed. 4 Field Trial System 4.1 System Overview The current field trial system is deployed in the Annandale area of Northern Virginia just outside Washington DC. There are four SS sites covering a 2km by 2km area. One of the sites shares a rooftop with 2/4

5 an existing sectored cell site (both local carriers) whilst the other three are located on conveniently placed buildings. The environment is a mixture of suburban dwellings with 10/11 storied office blocks and apartments situated along a busy thoroughfare. As such the system can be said to be multipath-limited rather than noise-limited. This means that the accuracy of the system is dominated by the reflections of the signal from surrounding buildings rather than how far the phone is from the receiver site. The antennas arrays used support omni-directional coverage with three standard omni-directional elements (3dBd gain) in a triangular configuration. The elements are spaced approximately half a wavelength apart which is about 7 for US cellular operation. An alternative antenna configuration would exploit only two elements over a ground plane, thereby providing the equivalent sectored coverage to the cellular BS antennas. Again, the element spacing would be approximately half a wavelength. 4.2 Results Figure 2 represents the results of a recent (typical) test of the latest version of the TeleSentinel system that supports the location of digital TDMA cellular phones. These involve several measures of performance (MOPS), including the root mean square (rms) of all the miss distances (MDs) between the mobile phone location estimates and the truth, the rms of the smallest 90% of the miss distances, the percent of miss distances less than 125 meters, and the maximum miss distances for various percentiles of the miss distances. The figure shows a rms miss distance of 57 metfs. This is well within the 125 meter FCC requirement for Phase II. Since the rms metric can be skewed easily by outliers, an alternative MOP is the rms of the smallest 90% of the miss distances. The figure shows that to be 40 meters in the case shown. Finally, a preferred measure of performance is given by a confidence interval for a given criterion. For example, one could say that there is a 90% probability that the mobile phone is within 50 meters of the estimated location. With this interpretation, the rms value of 125 meters converts mathematically to a statement that the phone has been located to within 125 meters with a probability of from 63% to 68%. The results shown in Figure 2 show that the TeleSentinel system can locate the mobile phone to within 45 meters with a probability of 67%, or to within 125 meters with a probability of 96%. As indicated by any of the above measures of performance, the results clearly show that KSI s TeleSentinel system meets and exceeds the FCC Phase 11 requirement in a moderate, suburbadurban multipath environment. Measurements in the same and different environments are currently ongoing. 5 Discussion At present in the US, there are a number of companies developing and marketing various wireless location technologies. To date, overall field trial results have been inconclusive as to the best approach, although the general opinion seems to favour a combination of technologies. Whilst KSI believes that AOA will provide the most accurate solution for all phones in most environments, there are cases where alternative approaches could be applied. One example would be in moving from the narrowband signal formats to a wider bandwidth signal such as CDMA. This would ensure better timing resolution for a TDOA approach, although the tighter power control requirements could limit the signal reception at multiple sites. In developing its CDMA technology, KSI plans to exploit a single site solution that combines AOA and TOA. Another classic example for combining technologies is in a rural environment where the cell sites are strung in a line along a highway. Whereas the signal may be heard at two sites, the in-line intersection of the resulting AOA bearings reduces the location accuracy. In this case, combining AOA with a single TDOA hyperbola provides a location. In parallel with the development of LDTs, various standards developments are also underway. The Telecommunications Industry Association (TU) under TR45.2 has established an ad hoc group to look at the emergency network and the inclusion of wireless E TlPl.5 have also addressed wireless location for GSM in the US (PCS1900). This work considered various technologies including both mobile and network based. For E-9 11 applications, it was concluded that a network based approach was the most applicable. The FCC mandate has certainly provided the initial impetus for the development of location technologies. With the potential for a wealth of location based services, the market for these services has yet to be filly realised. However, with the continued development and maturing of the technology, it is anticipated that enhanced services, and the market differentiation that this provides, will start to drive the deployment.

6 Acknowledgments The authors would like to acknowledge the dedication and support of the KSI engineers. This also extends to the engineers at Toracomm Ltd in Bristol, UK, whose RF and DSP expertise has made a significant contribution to this effort. References J.H. Reed, K.J. Krmnan, B.D. Woemer & T.S. Rappaport, An Overview of the Challenges & Progress in Meeting the E-911 Requirement for Location Service, IEEE Communications Magazine, April 1998, pp T. Bull et al, TSUNAMI Project Final Report, RACE 11 Project R2108, Feb A.J. Paulraj & C.B. Papadias, [Space-time Processing for Wireless Communications, IEEE Signal Processing Magazine, Vol.14, No.6, Nov W.C.Y. Lee, Mobile Communications Engineering: Theoly & Applications, 2nd Edition, McGraw-Hill, J.E. Maloney & C.N. Katz, Direction Finding Localization System, US Patent Number 4,728,959, March TDMA vs Truth (2/23/99 12:34:38 to 2/23/99 13:25:01) Miss Distance Histogram RMS of Miss Distances = 56.7 meters RMS of min 90 % of Miss Distances = 39.6 meters Miss Distances (meters) Miss Distance Cumulative Distribution 96 % Miss Distances e 125 meters 39 % Miss Distances 27.3 meters 50 % Miss Distances 33.6 meters 67 % Miss Distances 45.2 meters 86 % Miss Distances 73.5 meters 90 % Miss Distances 88.7 meters 99.9 % Miss Distances meters for samples I,,,,,,,,,,,,,,,,,,,,,,,,, I,,,,,,,,,,,,,, Miss Distances (meters) Figure 2: Miss distance distributions for a TDMA phone versus truth (23rd February 1999). 2/6