On-demand fingerprint selection for 802.11-based positioning systems

Abstract

Fingerprinting is a popular technology for 802.11-based positioning systems: Radio characteristics from different access points are measured at various positions and stored in a database. The database is copied to all mobile devices, and in case that a position estimate is needed, the device compares its currently measured radio characteristics with all the database entries. In this paper, we present two on-demand fingerprint selection algorithms to avoid the cumbersome and time-consuming approach of manually copying all fingerprints. Our algorithms only request those fingerprints from the database that are currently required to compute a position. The two algorithms differ in the way they shape the region for which fingerprints are requested. On-demand selection also allows storage-restricted mobile devices to utilize the positioning system. We carefully evaluate our algorithms in a real-world experiment. The results show that our algorithms do not harm the position accuracy of the positioning system. In addition, we analyze the space requirements of our algorithms and show that the typical constraints of mobile devices are met.