|
About The Talk
Indoor localization is of extreme importance for various pervasive applications, attracting many research attention in the past decades. Various solutions involving WiFi, Bluetooth, Video, and other RF devices have been proposed. Among them, WiFi fingerprint-based is a popular indoor localization technique since it can utilize existing infrastructures (e.g., access points). However, one limitation of this approach is the labor-intensive and time-consuming site survey process, which causes significant difficulties in practice. In addition, providing an accurate fingerprint-based localization system that can work in dense environment with minimal maintenance is challenging due to the practical deployment of the WLAN controller, energy saving, and load balancing policies managing devices association dynamically, which limits the location estimation performance. Besides, the complex and dynamic nature of the indoor environment, which makes fingerprint map maintenance difficult since the signal is easily influenced by the structures, layout, and pedestrians around the examined areas. As a result, solely utilizing fingerprint for indoor localization achieves a much lower accuracy than expected, possibly exceeding more than 10 meters in dense real-world environment.
This thesis (a) explores how to integrate multiple complementary sensors that are already commonly deployed in the real-world environment to improve the indoor localization. Then, (b) based on the enhancement information gathering in step (a), the outdated fingerprint database which might be caused by the ambient changes such as added/removed doors or walls, relocated APs, etc., could be automatically corrected with limited workforce requirements.
|
