Numbers of indoor positioning techniques have been proposed, but none of them is generally deployable in most buildings, in spite of their acceptable accuracy in experiments. The primary reason is that, to work properly, all of them have specific assumptions about context, like infrastructure, environment, and user behaviors. However, assumptions of them are almost mutually independent, and their working context are actually detectable. In this paper, we propose a indoor positioning system framework to combine location estimations from multiple positioning systems. Each system is encouraged to implement its own context-awareness mechanism, and needs to provide a confidence about its estimation. The framework maintains credit history for all participating positioning system entities, and fuses their estimations based on their credits and confidences. Our preliminary deployment shows that it is not hard to provide a highly reliable confidence, and using data fusion techniques like Kalman Filter can further reduce errors in the framework's decision.




Download Full History