Development of Inverse Hyperbolic Positioning Using the GSM Cellular Telephone Network, 16-9222Printer Friendly Version
Inclusive Dates: 10/01/00 - 09/30/02
Background - This internal research program was undertaken to develop a new location technology that operates by exploiting transmissions from the Global System for Mobile Communication (GSM) cellular telephone network. The technology is entirely passive, requiring neither registration into the cellular network nor transmissions to the network base stations. This capability provides an alternative to the Global Positioning System (GPS) for cases in which GPS technology is not appropriate.
Approach - The goals of the program included the establishment of the theoretical background and fundamental algorithms for geolocation using the GSM network, analysis of the potential performance of the system, and demonstration of the technique. The theoretical development included the review, implementation, and adaptation of geolocation techniques from the literature. The algorithm was evaluated using extensive simulation trials. The GSM network of San Antonio, Texas, was studied and characterized. Geolocation experiments were conducted using broadcast network timing information.
Accomplishments - SwRI has developed a two-stage geolocation algorithm that uses a simple, closed-form solution as an initial estimate to an iterative, approximate maximum likelihood algorithm used to compute the final location estimate. This algorithm was evaluated through simulation against the Cramér-Rao bound, a fundamental bound on geolocation performance. The Cramér-Rao bound for this technical scenario was derived, and the experimental results were found to achieve this theoretical limit. Signal-processing techniques were developed to detect synchronization signals transmitted by GSM base stations. Additional techniques were developed to compute high precision time of arrival (TOA) estimates of the messages, which are used for geolocation. The signal-processing techniques and the geolocation algorithm were evaluated using broadcast network timing information from GSM base stations in San Antonio. These trials resulted in a mean location error of 89 meters. Geolocation using a noncoherent average over several seconds resulted in a location error of 7.5 meters.