A Bayesian Method for Mobile Positioning in Urban Areas

H. Kunczier, A. Friedreich, and R. Zelenka (Austria)


Location based service, Pattern matching, Bayesian network, Motion modelling


Mobile positioning techniques have been under strong de velopment over the last years and the number of location based services is increasing. However, insufficient perfor mance in terms of accuracy and availability as well as cost intensive deployment still prevent location based services from becoming mass market applications. Especially urban areas, strongly suffering from multipath effects, reduce the achievable positioning accuracy. In this paper we concen trate on power level based methods and present a Bayesian approach for positioning. We use Bayesian networks for the characterization of a position and show urban area trial results to verify the method under real life conditions. A motion model is introduced and a 2D filtering technique is proposed to improve the accuracy. Positioning errors below 100 meters in 95% of all test cases have been achieved.

Important Links:

Go Back