GPS/INS integration in a severe urban environment

GPS satellite visibility data collected over a ground track in the "deep urban canyons" of lower Manhattan indicated that there were usually too few satellites being tracked to form a position solution. The present study investigates the augmentation of such limited GPS data with a low cost, low accuracy INS. The navigation system is assumed to be self-contained and unobtrusively attached to the host vehicle. Initial covariance matrix analyses produced errors far in excess of that deemed necessary for a remotely sited operator to locate the vehicle on the correct street, when the navigation data is projected onto a map display. A series of software upgrades were then shown to produce significant performance improvements. In the final phase of this effort, a simplified simulation program was used to display navigation data on a grid map of the area in question. It was then found that a revised navigation algorithm produced data that appears to greatly enhance the ability of an operator to correlate the navigation data with the correct streets on the map.