A propagation modeling approach to urban navigation

The performance of GPS (Global Positioning System) in densely built urban areas is greatly undermined due to multipath propagation. Many attempt alternative means for urban navigation, of which the 'fingerprinting' method has drawn much attention recently. However, the labor of site surveys and a constantly changing environment prevents 'fingerprinting' from being an eificient solution. Therefore, we propose a propagation model which is used to obtain location fingerprints with the aid of modern computers thanks to their growing processing power. At the heart of the technique is the use of a ray-launching model from which a database is generated containing propagation parameters such as received signal strength, time of arrival and angle of arrival as a function of source and receiver location. These are then mapped with physical locations using an artificial neural network. Besides, a sequence based tracking tool is proposed to assist navigation with minimal communication required between mobile device and database. This paper details the generation of a database for real world geometries read from OpenStreetMap and the development of a localisation algorithm.