Xu, Y: Autonomous Indoor Localization

Indoor localization is a research domain that aims to locate mobile devices or users in the indoor environments. More and more research has investigated to acquire the location information based upon existing Wi-Fi infrastructure. A technique of using current Wi-Fi data and a fingerprint database containing Wi-Fi fingerprints of desired locations for localization is known as Wi-Fi fingerprinting. Most current approaches for Wi-Fi fingerprinting depend on labor-intensive and time-consuming site surveys by professional staff or users to generate a fingerprint database of desired locations. Moreover, these approaches are not satisfactory for long-term localization of mobile devices in practice due to the costly and continuous update of the fingerprint database.In this thesis, we propose an approach to the indoor localization problem, in which we combine the Wi-Fi fingerprinting technique and the place learning technique to learnand update the Wi-Fi fingerprints of significant locations in an unsupervised manner. Significant locations are locations a user spent at least for a while (e.g., 10 minutes) and are most important and highly frequented in people's daily lives. The conventional approaches use labeled Wi-Fi data intentionally collected by professional staff or users and learn Wi-Fi fingerprints of desired locations. Instead, the proposed approach uses unlabeled Wi-Fi data collected in a user's daily life and learns Wi-Fi fingerprints of significant locations related to user's daily trajectory and activities.We implement an autonomous indoor localization system WHERE based on the proposed approach. The system can automatically learn and update Wi-Fi fingerprints of significant locations, and determine the location of the mobile device when it returns to the learned locations. Moreover, we evaluate various measures of performance, in term of the location accuracy, the computational time, the power consumption, the size of a fingerprint database, and the system