Generation of irregular grid maps for fingerprinting-based mobile radio localization using farthest-first traversal and low-discrepancy sequences

Abstract

Localization techniques play a fundamental role in various applications like in the development of smart cities, making the improvement of such technologies indispensable. In this context, radio frequency (RF) fingerprinting (FP)-based localization methods are attractive due to their lower implementation cost, lower energy consumption, and better performance in non-line-of-sight conditions compared to GPS. With this in mind, we propose an RF FP-based localization method using an irregular grid map generation strategy. The grid map segmentation procedure is based on the assumption that the RF measurements collected in a region carry information on how users are often spatially distributed. The algorithms employed in the irregular grid map generation are the farthest-first traversal and low-discrepancy R-sequence. Different irregular grid map generators are compared with each other and with the regular grid map. We evaluate the generators in two different scenarios: with measurements obtained through simulation with ns-3 and via a driving-test procedure in an urban area. Numerical results indicate that our proposed irregular generator presents a lower localization error in both scenarios and is the only one capable of meeting one of the accuracy requirements stated by the Federal Communications Commission that demands a localization error of up to 50m in 80% of the emergency calls.