BRAG: Blind region-agnostic geolocation of LTE mobile users in urban areas

Abstract

Geolocation is the backbone of many novel location-intelligent applications. Additionally, geodata analysis helps model and predict spatiotemporal fluctuations in data traffic, which is important for network optimization, operation cost reduction, and power saving. Furthermore, geodata analysis can be utilized in fields such as transportation, urban planning, tourism, marketing, epidemiology, national statistics, and censuses. Cellular geolocation is advantageous when Global Positioning System (GPS) readings are not available, especially since it does not require altering the network infrastructure or installing expensive equipment. However, cellular geolocation is challenged by the high randomness and dynamics of the environment. In this paper, we propose a blind region-agnostic algorithm to geolocate Long-Term Evolution (LTE) mobile users in urban areas. The algorithm uses timing and signal strength readings, which are readily available at the serving evolved Node B (eNB), to calculate initial estimated positions. Following that, the algorithm uses correlations between the initial estimates along the user’s path to improve its geolocation accuracy. The proposed method does not require training or prior data collection, making it easy to implement in different regions. We tested the method on real data from drive tests in different cities, and the method achieved a mean error of 132 meters and a median error of 91 meters, compared to a mean error of 203 meters and a median error of 125 meters achieved by basic time-advance-based geolocation.