Indoor/Outdoor Location of Cellular Handsets Based on Received Signal Strength

Show simple item record

dc.contributor.author Zhu, Jian
dc.contributor.author Durgin, Gregory D.
dc.date.accessioned 2010-08-16T21:24:11Z
dc.date.available 2010-08-16T21:24:11Z
dc.date.issued 2004-06-18
dc.identifier.uri http://hdl.handle.net/1853/34418
dc.description.abstract This report documents the results of a ground-breaking set of experiments for mobile handset location within the commercial cellular telephone network. With particular emphasis on the US emergency 911 (E911) location problem, we demonstrate the viability of Received Signal Strength (RSS) techniques to meet the safety requirements set forth by the Federal Communications Commission (FCC) in a semi-urban environment. Furthermore, we conclusively show that RSS location techniques are also accurate for indoor users – a characteristic unique among all currently proposed E911 technologies. Our measurement campaign and test results indicate RSS-based techniques can approach or even surpass the FCC guidelines of 100m accuracy 67% of the time and 300m accuracy 95% for a network with a majority of indoor users. Since most cellular phone calls are now placed from indoor environments, this result has enormous implications for the E911 rollout and public safety. The RSS location technique is a relatively new and controversial method for radiolocation within the cellular network. The principle idea is to solve for users’ xy-coordinates by studying signal strength measurements of nearby cellular sectors made by their handsets. All digital handsets measure the signal strength of neighboring control channels, and report the results back to the serving base station in the form of a network measurement report (NMR). All digital cellular air interfaces include the ability to report NMRs, largely for the purpose of performing mobile-assisted hand-offs (MAHOs). Once this NMR has been received at the base station and routed to the central switching office, its set of signal powers is matched to those in a well-calibrated database of RF maps. The closest match between measured and predicted signals likely occurs at a point near the groundtruth location within the database. This technique is similar to the scheme used to locate WLAN modems in a much smaller-scale location problem [Che02]. The technique has been proposed for use in the cellular network by [Wei03]. To perform this study in radiolocation, we turned the Georgia Tech campus into the world’s first indoor/outdoor cellular location laboratory. The ensuing location tests were performed on an 850 MHz IS-136 cellular network in mid-town Atlanta. The Georgia Tech campus approximates a typical semi-urban or dense suburban area with streets, moderate green space, and many 4-5 story academic and office buildings. Although the potential population density of cellular users is high, there are no skyscrapers or canyons that would be associated with dense urban deployments. A database of RF coverage maps for all nearby serving sectors was created from a combination of propagation modeling and varying degrees of indoor and outdoor measurement calibration using a Comarco IS136 scanner with baseband decoding. Real, pedestrian-style handset measurements were taken with an Ericsson handset connected to an Ericsson TEMs data collection unit. The results in this study show that RSS location techniques can satisfy the FCC E911 requirements for outdoor handsets in semi-urban environments. This result is shown in Section 6.2.6. When a majority of the test handset data originates from indoor locations (as it would in real life), the performance degrades somewhat. For example, the error distance between a location estimate and a handset’s groundtruth position drops from 100m or less 66% of the time to 100m or less 56% of the time (see the indoor analysis in Section 6.2.3). However, this report demonstrates a variety of ways to recover the lost accuracy by modifying the location algorithms, adding indoor calibration measurements, modeling indoor propagation using satellite photogrammetry, and using sequential handset measurements. The most accurate location algorithm is documented in Section 6.2.4; using a sequence of 10 linearly-averaged handset measurements and RF maps calibrated with both outdoor and indoor measurements, the error distance for this case is 100m or less 78% of the time and 300m or less 98% of the time. This upper limit of performance is well above the FCC E911 requirements. This measurement campaign lasted for 4 months (January through April) in the beginning of 2004. All data points were tagged with absolute longitude and latitude coordinates taken from a Global Positioning System (GPS) radio; however, due to the limitations of GPS, many outdoor coordinates and all indoor coordinates had to be painstakingly estimated from geo-referenced maps of campus and manually entered into the database. This is one source of error in our measurements. There are other unique sources of error in our measurements that may make our results somewhat pessimistic. For example, there was a seasonal change in the middle of our data collections where leaves grew back on the campus trees, changing the propagation characteristics by several dB. Also, one of the large buildings within our test area was demolished in the middle of our campaign. We also used a fairly simple location algorithm since we were concentrating on the more complicated question of indoor feasibility. There are many other algorithms that have been proposed which could improve the performance [Aso00][Lai01][PB00]. Several recommendations emerge from this study. Our experimental results suggest that RSS-based techniques may be resilient enough for deployment as a standalone position location technology for satisfying the FCC’s E911 requirements in most populated areas. There are still several questions about this technology that need to be addressed. First and foremost, it is unclear how much cost and effort that is required to maintain the performance in cellular networks that, to one degree or another, are always undergoing buildout, optimization, or modification. Ultimately the ideal solution for the US E911 problem will be a hybrid combination of handset-based Global Positioning System (GPS) technology and an RSS-based location system. These two technologies seem to complement each other so well. GPS works in rural, open-sky environments where all network-based location solution tends to degrade due to the low density of base stations. Conversely, GPS fails whenever satellite links become obstructed. This can happen in any environment, but is particularly accute in urban and indoor areas – precisely the places that RSS radiolocation works best. If public safety is the primary concern, then this long-term tandem of location technologies seems to be most sensible. At Georgia Tech, we are continuing to pursue research in the field of RSS-based position location. Several areas of proposed research are: How well do RSS-based location technologies perform in a wide variety of in-building environments (residences, skyscrapers, retail establishments, etc.)? How do we improve state-of-the-art propagation modeling to build accurate RSS databases in regions devoid of measurement calibration? How can the RSS databases be efficiently calibrated and maintained? There is much work left to be done in development of this late-coming location technology, but initial results are quite promising. en_US
dc.language.iso en_US en_US
dc.publisher Georgia Institute of Technology en_US
dc.relation.ispartofseries Propagation Group ; PG-TR-040618-JZ en_US
dc.subject Cellular network en_US
dc.subject Indoor and outdoor handsets en_US
dc.subject Propagation model en_US
dc.subject Radiolocation en_US
dc.subject Received signal strength en_US
dc.title Indoor/Outdoor Location of Cellular Handsets Based on Received Signal Strength en_US
dc.type Technical Report en_US
dc.contributor.corporatename Georgia Institute of Technology. School of Electrical and Computer Engineering


Files in this item

Files Size Format View
PG_TR_040618_JZ.pdf 1.947Mb PDF View/ Open

This item appears in the following Collection(s)

Show simple item record