WiFi-based urban localisation using CNNs

Abstract

The continuous expanding scale of WiFi deployments in metropolitan areas has made possible to find WiFi access points at almost any place in our cities. Although WiFi has been mainly used for indoor localisation, there is a growing number of research in outdoor WiFi-based localisation. This paper presents a WiFi-based localisation system that takes advantage of the huge deployment of WiFi networks in urban areas. The idea is to complement localisation in zones where the GPS coverage is low, such as urban canyons. The proposed method explores the CNNs ability to handle large amounts of data and their high accuracy with reasonable computational costs. The final objective is to develop a system able to handle the large number of access points present in urban areas while preserving high accuracy and real time requirements. The system was tested in a urban environment, improving the accuracy with respect to the state-of-the-art and being able to work in real time.