• Citer

Distributed Direct Localization and Tracking of Wireless Transmitters

par Pocoma Copa, Evert Ismael
Président du jury De Doncker, Philippe
Promoteur Horlin, François
Co-Promoteur Quitin, François
Publication Non publié, 2024-02-26

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Localization and tracking play important roles in wireless networks, providing the foundation to a broad spectrum of applications, such as emergency services, smart cities, autonomous driving, and more. In addition, location-based services gain increased attention, as their relevance continues to grow in 5G and 6G networks. Hence, this thesis delves into the localization and tracking of wireless transmitters.Direct approaches estimates the transmitter's position directly from the received signals at the base station (BS), whilst indirect ones rely on derived parameters, leading to less accurate estimates compared to direct ones. Distributed approaches spread the processing across BSs, thereby improving scalability compared to centralized ones. However, distributed localization and tracking approaches encounter various challenges when applied directly on the received signals. Therefore, the focus is narrowed down to the direct and distributed approaches. Additionally, the research is presented on three major axes.The first axis, Direct Localization, starts by refining the centralized iterative positioning estimation (IPE) algorithm. It details issues related to the information loss present in the iterative process of IPE. Two countermeasures are proposed, involving a kernel-based representation of received-signal loglikelihood for improved communication, and maintaining a common processing domain to prevent information losses during domain conversions. Afterwards, it explores Distributed Direct approaches, where a novel grid-based consensus approach, denoted as self-synchronization positioning estimation (SSPE), is introduced. It leverages the self-synchronization mechanism (SSM) to achieve global convergence in a distributed manner, on the posterior distribution of the transmitter's position. SSPE demonstrates its applicability to various network topologies, with localization performance comparable to the direct centralized localization approach.The second axis, Direct Tracking, delves into the integration of SSPE with sampled-Bayesian filtering approaches, such as point mass filter (PMF) and particle filter (PF). Due to the same sampling nature, PMF and SSPE can be integrated straightforwardly. Hence, the focus changes to mitigate the computational burden of PMF, in a rather centralized scenario. This involves a two-fold strategy: first reducing the sampling space, and then employing approximations to avoid complex operations. Subsequently, to integrate SSM and PF, SSPE is adapted to work with stochastic sampling. Additionally, the kernel representation and SSM-related compression of consensus variables are improved, ensuring robust and eficient tracking in both cases.The third axis explores the fusion of active and passive sensors for localization and tracking. This involves combining active sensors, such as frequency modulated continuous wave (FMCW) radars, with passive ones, like time difference of arrival (TDOA), to enhance the overall quality of information. It focuses on medium-level data fusion, specifically considering time-of-arrival estimates and radar detections after the radar processing chain. A combined model is proposed and solved using the constrained weighted least squares (CWLS) method, showing a reduction in localization error compared to the use of a single sensor and facilitating analysis for various measurement combinations.In summary, this thesis makes contributions to the field of localization and tracking in wireless networks by proposing methodologies that promise improved accuracy, reduced complexity, and enhanced fusion capabilities.