Wireless LAN Based Hybrid Positioning System Using Bayesian Inference and Access Point Selection

Abstract

Over the last few years, wireless LAN and location based service has been a hot subject of controversy. In this paper, we propose a Bayesian-based positioning system and a location-based application using existing wireless access points at a college campus. Our system uses signal strength of access points, Bayesian inference, and selection of access points to estimate a user's location. Using this pre-observation model-based system and considering signal strength direction, a user can construct a positining system easily and obtain good location estimation accuracy. We implemented the system and exemplified its effectiveness with experiments at our campus. Fig. 1 shows the result of experiment. The figure shows the location estimation error from the correct location (x-axis) and the cumulative probability parentage of the number of total requests (y-axis). Each line shows the result using different access points. For comparison of another method, we also show the result using the Euclid distance method. In our method, the system selects access points for positioning according to threshold of received signal strength. The access point's threshold settings are listed below. • Selection A : From-50 dBm to-59 dBm • Selection B : From-60 dBm to-69 dBm • Selection C : From-70 dBm to-79 dBm • Selection D : From-80 dBm to-89 dBm Fig. 1. Estimation Accuracy Additional to this, we conducuted two experiments. One is difference of location estimation accuracy to each request period, the other is difference of received signal strength in accordance with time. For example, selection A uses access points of which average received signal strength is from-50dBm to-59dBm. To select appropriate access points, the positioning system itself improve their estimation accuracy. Using this system, we also propose a location-based presence service system. By using the presence system, one can get the location of friends automatically. The application is a kind of a presence service like Instant Messengers. In our system, user's location information is automatically shown with the presence of users. Many people in our laboratory manually turn their instant messenger's name into names with their location, for example [user name @ their location]. This application changes the instant messenger's name, [@lo-cation] in accordance with signal strengths users can observe at that moment. Fig. 2 shows an instant messenger running with our auto name application. This software changes [@ their location] using the location information obtained from our location estimation method. Fig. 2. Auto presence application-3