From Ground to Sky: Localization Framework for Long-Range Wireless Networks

This PhD research addresses the localization problem in emerging long-range wireless networks, proposing power- and cost-efficient localization frameworks to cope with the corresponding network characteristics. These frameworks investigate the localization of both ground and aerial users using ground and aerial anchors, resulting in four localization scenarios: ground-to-ground (G2G), air-to-ground (A2G), air-to-air (A2A), and ground-to-air (G2A). In the context of these four scenarios, the proposed frameworks exploit the advances in machine learning and unmanned aerial vehicles' (UAVs) 3D mobility, leading to four main contributions. The first contribution is a G2G two-step localization framework for longrange Internet of things (IoT) networks. The proposed framework leverages the large-scale metadata collected from a wide-range IoT deployment, enabling received signal strength (RSS) fingerprinting localization with machine learning algorithms. Subsequently, an RSS-based ranging method is explored to further refine the locations estimated in the first step. Our measurement-based results confirmed the effectiveness of the two-step proposed framework, providing localization accuracy sufficient for IoT applications such as assets tracking. However, based on our measurements, the RSS ranging performance significantly degrades with distance due to the increasing shadowing effect. Therefore, limiting the use of the RSS-based ranging to a maximum of 200m is recommended for bounded localization error. Second, an A2G localization framework is proposed, aiming at improving the reliability of RSS ranging methods by exploiting UAVs' 3D mobility. Owing to their swift movement, UAVs can adapt their altitude for a better line-of-sight (LoS) experience. Accordingly, the proposed framework explores UAV trajectory design for minimum localization error while considering the limited energy onboard as a constraint. The conducted simulations show that a UAV anchor can bring the localization error to more than 40% lower compared to its ground counterpart. Following the localization gains obtained for ground nodes using aerial anchors, the third main contribution is an A2A RSS-based localization and detection framework for aerial nodes, e.g., amateur drone. In particular, the proposed framework employs surveillance aerial anchors to detect, and subsequently localize any intruder drone within a no-fly zone. Both analytical and simulation analyses show that there is an optimal altitude for surveillance aerial anchors, minimizing aerial node localization error. Finally, we propose a G2A time difference of arrival (TDoA) localization framework to expand the localization of aerial nodes from an area-limited no-fly zone to a large-scale tracking solution. The proposed framework relies on distributed ground-based crowdsourced wireless networks, providing a continent-wide radio frequency coverage. An autoregressive model and recurrent neural network (RNN) are used, providing time synchronization needed for TDoA. Subsequently, to handle the receiver anchors fixed positions and aerial nodes arbitrary altitudes, the proposed framework employs TDoA, along with a Kalman filter, to track aerial nodes' positions over time. Our measurement-based results proved the effectiveness of the proposed framework, minimizing the localization error by orders of magnitude.

Publication year:2020

Accessibility:Closed