A Practical Approach to Cooperative Localization in GPS-Limited Urban Environments

Existing localization techniques have fundamental limitations which make deployment under commonly encountered conditions difficult. Techniques used to augment GPS generally require specialized infrastructure, tedious calibration tasks, or other requirements that make general purpose application difficult, yet may still be unreliable in practice. In this paper, we examine the utility of cooperatively sharing location data among connected nodes for the purpose of correcting positions with high measurement error, while doing so in a simple manner that can be deployed anywhere without onerous setup tasks or highly specialized hardware. Using a simple data sharing and filtering technique, collaborating users can substantially reduce overall localization error in dead reckoning systems deployed to urban areas where nodes may have a broad spectrum of location quality. We evaluate the effectiveness of this method and examine the system parameters necessary to fully exploit such cooperative localization. We show that mean position error can be reduced by nearly 50 percent for given application scenarios. If distance measurement is available, filtering positions based on estimated error can improve localization accuracy of pedestrian dead reckoning techniques to approximately that of GPS.