부산대학교 도서관

We present in this paper a method for land-vehicle navigation which fuses GPS and odometer (differential odometry) measurements. The positioning is also corrected with road map information. The map is here modelled as a sensor which delivers the node location and the road bearing. This equation allows a direct modelling of the map errors that can be embedded in the filter. The sensor equation is initialized by a matching procedure that uses the Mahalanobis metric as a statistical distance criterium. Hence, the filter that fuses the whole measurements (GPS, odometer, map), has a centralized fusion scheme in which each measurement are sequentially processed. The method is then applied to a transport network scenario in presence of GPS outages in order to study the stability of the method and show the accuracy with the map corrections.