VEHICLE LOCALIZATION IN URBAN CANYONS USING GEO-REFERENCED DATA AND FEW GNSS SATELLITES

Stand-alone Global Navigation Satellites System (GNSS) can provide a very accurate real-time positioning in good visibilty and signals conditions. Unfortunately, for a vehicle evolving in urban areas, those conditions are rarely satisfied, leading to debased computed location or lack of solution if less than four SVs are visible. A way to tackle such a problem can consist in fusing the raw measurement from the GNSS reciever with others data sources like proprioceptives sensors. In this paper, the use for computation fix of a priori charted road network data managed by a Geographical Information System (GIS) is presented. Geo-referenced data are modeled as segments that can be used as constraints or fused with the pseudo-ranges. Using the residuals, a new method for treating the underlying problem of road selection is proposed. Experimental results illustrate the interesting performance of this approach since the map-matched location can be computed using only 3 satellites and a usual 2D map in urban canyons.