Association, Tracking, and Localization of Bluetooth Devices

Sammanfattning

Bluetooth Low Energy (BLE) devices have become integral to modern wireless com munication, facilitating a wide range of applications in consumer electronics, healthcare, security, and beyond. Their capacity to broadcast signals presents significant opportunities for tracking and localization, which could be helpful in domains such as search and rescue operations. However, privacy mechanisms, particularly randomization of Media Access Control (MAC) addresses, pose substantial challenges to persistent device identification and tracking. This thesis explores methodologies for associating BLE measurements with specific devices and estimating their locations, despite the complexities introduced by privacy preserving techniques. By developing robust association and localization frameworks, this study aims to advance BLE-based tracking applications while addressing the inher ent trade-offs between utility and privacy. The methodological framework is empirical and based on BLE signal measurements collected through scanning equipment deployed in both controlled tests and outdoor field conditions. The process begins with identifying unique MAC addresses and extracting rel evant features. These features feed into an association algorithm that attempts to associate individual measurements with specific physical targets despite the address randomization. Alocalization algorithm is then used to estimate the position of each target by combining Received Signal Strength Indicator (RSSI) data with the known locations of the scanners. Both algorithms are validated against ground truth data using reference devices and RSSI f iltering. Association and localization are central to this tracking method: association ensures that fragmented or randomized measurements are correctly grouped per physical device, while localization estimates where each target was at a given time. The output of the methodisaspatiotemporal trace of each tracked device, providing a sequence of estimated positions that can aid in real-time tracking or retrospective analysis. Theassociation algorithm is capable of grouping signal observations from the same de vice, although its accuracy is affected by timing inconsistencies. Localization, based on sig nal strength data, generally performs well when scanners are spatially distributed but can suffer in certain geometric configurations. Adjustments in scanner placement and tempo ral processing show potential to improve association reliability and localization accuracy. While the approach is promising for static or semi-static scenarios, further adaptation is needed for use in dynamic, real-time search and rescue operations.