Improving accuracy of GNSS devices in urban canyons

This paper addresses the problem of calculating the ac- curate position of a GNSS device operating in an urban canyon, where lines of sight (LOS) with navigation satel- lites are too few for accurate trilateration calculation. We introduce a post-processing refinement algorithm, which makes use of a 3D map of the city buildings as well as captured signals from all traceable navigation satellites. This includes weak signals originating from satellites with no line of sight (NLOS) with the device. We also address the dual problem - computing a 3D map of the city buildings when the position of the device is given. This is achieved by storing LOS/NLOS rays to all navigation satellites sampled at multiple locations within a region of interest (ROI). These rays are then used to compute the 3D shapes of buildings in the ROI. A series of field experiments confirm that both algo- rithms are applicative. The position refinement algo- rithm significantly improves the device's accuracy and the mapping algorithm allows few users to map a com- plex urban region simply by walking through it.