Distance metrics for location fingerprinting in wireless local area networks

For more than a decade the wireless local area networks (WLAN) market has seen explosive growth that shows no signs of abating as more portable devices, such as smartphones, are becoming available on the market with built-in IEEE 802.11 protocol capabilities. With the technology now ubiquitous in the everyday lives of the general public, there is significant interest in developing new applications and services that leverage the existing IEEE 802.11 infrastructure. One of the most promising areas is location based services (LBS) that utilize the current location of the mobile device to provide additional information. Traditional LBS applications are already widely deployed, relying on GPS technology to provide the location information, but GPS is not always cost effective or useful in indoor environments. To solve this problem, there is significant interest in tracking users in the WLAN environment and location fingerprinting has been identified as a promising approach.;Location fingerprinting is a map based solution that stores past measurements of received signal strengths at known reference/grid points in a database that is later used to localize a user at an unknown location to the closest reference point. To detect the position of a device within WLANs, the intrinsic properties of radio propagation characteristics are used linking the signal strength to the distance. However, this relation is affected by random effects which in turn make a localization a complex problem. The advantage of using location fingerprinting for WLANs is that much of the capital cost for providing the service has already been paid for as this software solution requires no specialized hardware. The signal strength data that is to be captured is provided in the base 802.11 protocol as Received Signal Strength Indicator (RSSI) values, and is readily obtainable from the beacons that are transmitted at regular intervals by the Access Points (APs).;This thesis focuses on processing the RSSI data vectors using the core tools of Euclidean distance and neural networks, to develop distance metrics for the database, and locate the user based on signal strength readings given by a wireless card. The contributions of this thesis are in the development of methods to localize the user, the evaluation of these and other existing techniques in line of sight (LOS) and non line of sight (NLOS) environments using the distance error and percentage of correct classification metrics. In addition, the study of some of the factors affecting the localization of the user provides new insights into design of the overall location fingerprinting system.;The data from the real world test environments was captured using a wireless sniffer and then manipulated in Excel and imported into MatlabRTM for final calculations. Location fingerprinting, when combined with the many techniques used in this thesis, applies statistical classification to separate the grid point signal strength data in the database, with boundaries that will allow for the mapping of a test point from the signal space to a physical location. Within the framework of localizing to the closest reference point with a reasonable accuracy, this thesis demonstrates that location fingerprinting is a powerful approach in WLAN systems.