Mobile location methods that employ signal fingerprints are becoming increasingly popular in a number of wireless positioning solutions. A fingerprint is a spatial database, created either by recorded measurement or simulation, of the radio environment. It is used to assign signal characteristics such as received signal strength or power delay profiles to an actual location. Measurements made by either the handset or the network, are then matched to those in the fingerprint in order to determine a location. Creation of the fingerprint by an a priori measurement stage is costly and time consuming. Virtual fingerprints, those created by a ray-tracing radio propagation prediction tool, normally require a lengthy off-line simulation mode that needs to be repeated each time changes are made to the network or built environment. An open research question exists of whether a virtual fingerprint could be created dynamically via a ray-trace model embedded on a mobile handset for positioning purposes.The key aim of this thesis is to investigate the trade-off between complexity of the physics required for ray-tracing models and the accuracy of the virtual fingerprints they produce. The most demanding computational phase of a ray-trace simulation is the ray-path finding stage, whereby a distribution of rays cast from a source point, interacting with walls and edges by reflection and diffraction phenomena are traced to a set of receive points. Due to this, we specifically develop a new technique that decreases the computation of the ray-path finding stage. The new technique utilises a modified method of images rather than brute-force ray casting. It leads to the creation of virtual fingerprints requiring significantly less computation effort relative to ray casting techniques, with only small decreases in accuracy.Our new technique for virtual fingerprint creation was then applied to the development of a signal strength fingerprint for a 3G UMTS network covering the Sydney central business district. Our main goal was to determine whether on current mobile handsets, a sub-50m location accuracy could be achieved within a few seconds timescale using our system. The results show that this was in fact achievable. We also show how virtual fingerprinting can lead to more accurate solutions. Based on these results we claim user embedded fingerprinting is now a viable alternative to a priori measurement schemes.
展开▼
