View article

Location systems that are based on scanning for nearby radio sources can estimate the position of a mobile device with reasonable accuracy and high coverage. These systems require a calibration step in which a map is built from radio-readings taken on a location-aware device. War driving, for example, calibrates the positions of WiFi access points using a GPS-equipped laptop. In this paper we introduce an algorithm for self-mapping that minimizes or even eliminates explicit calibration by allowing the location system to build this radio map as the system is used. Using nearly 100 days of trace data, we evaluate self-mapping’s accuracy when the map is seeded by three realistic data sources: public war-driving databases, WiFi hotspot finders, and sporadic GPS connectivity. On average, accuracy and coverage are shown to be comparable to those achieved with an explicit war-driven radio map.