dc.description.abstract | For over a decade, Indoor Localization has been a crucial topic among researchers. A multitude of localization solutions has been provided so far, including radio frequency based solutions like WiFi, Bluetooth and RFID based localizing and positioning systems. These infrastructure based solutions require a set of additional devices to be installed, which comes with challenges like huge installation costs. These solutions are device dependent. Also, attenuation in signal strength throughout the day leads to a major error in localization. A recent addition to this system is magnetic field based localization techniques. The solution lies in exploiting the ambient magnetic fields present inside buildings and their unique variations caused by the presence of ferromagnetic objects such as pillars, doors, and elevators. Smartphone based built in magnetometers have been the default data sensing platforms. The data collected from smartphones are used by deterministic or probabilistic algorithms for estimating locations. However, the performance of these algorithms depends on the diversity in the sensor models built- in the smartphones, diversity in the users using the phones, and diversity across space and time. There is a dearth of analyses of how these diverse factors affect the performance of magnetic field based solutions. We assess the impact of the four diversity parameters on the dynamic time warping algorithm in estimating the users� location. We discuss our findings from experiments conducted across three different buildings and eight different sensor models with five users. | |