Positioning with Multilevel Coverage Area Models

Abstract

Fingerprinting techniques provide good indoor and urban user location estimates, but using them in large scale requires an enormous radio map (RM) database. To reduce the database size, we build a statistical model of the coverage area (CA) of each wireless communication node (CN) using “fingerprints” (FP), i.e. reception samples. In previous work we modeled each CA as a single ellipse, so only 5 parameters need to be stored in the RM for each CN. In this paper, we investigate the use of multiple CAs for every CN. FPs are grouped based on received signal strength (RSS) criteria and CA models are fitted to different FP groups. Different choices of RSS boundaries are examined with real data. We present a method for positioning using the proposed “multilevel coverage area radio map”. The proposed method is applied on real data sets. The positioning results are compared with those of conventional single level CA positioning and a basic location fingerprint methods. The results show improvement of positioning accuracy compared with positioning with a single level CA. The improvement is due to better use of RSS level information in both the offline phase (constructing the CA radio map) and in the online phase (user positioning). The proposed multilevel CA positioning works with a much smaller RM database than the basic location fingerprint method, without degrading the positioning accuracy in indoor positioning.