Banca de DEFESA: Larissa Aidê Araújo Rocha

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : Larissa Aidê Araújo Rocha
DATE: 29/05/2024
TIME: 09:00
LOCAL: VIDEOCONFERÊNCIA
TITLE:

METHODS FOR ESTIMATING USER POSITION IN THE NEAR-FIELD REGIME

KEY WORDS:

Keywords: Near-Field, DOA, User’s Location, Sub-array, ELAA, 3D Positioning, URA, HOSVD

PAGES: 120
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUMMARY:

Parameter estimation using signal processing methods is a significant topic in various domains such as radar, seismic analysis, sonar, electronic surveillance, and more. This research has focused on two key areas within signal processing: antenna array processing and user position estimation. In the context of antenna array processing, the user’s localization involves estimating the position parameters of sources, often employing subspace methods such as Multiple Signal Classification (MUSIC) and estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT), among others. In far-field scenarios, a source is characterized solely by its direction of arrival (DOA). However, this assumption no longer holds in near-field situations where sources are close to the sensor array. In such scenarios, the wavefront of the signal becomes spherical, necessitating the consideration of two parameters for accurate source localization: the direction of arrival and the distance between the source and the sensor array. The near-field user position estimation is crucial due to several factors, especially when considering advancing networks like sixth-generation (6G), where further improvements are expected in localization and tracking. These improvements will result from the joint use of high frequencies and large arrays. With the significant increase of the antenna number and carrier frequency in future 6G systems, the near-field region of extremely large antenna arrays (ELAAs) will expand by orders of magnitude. As a result, near-field communications will play a critical role in future 6G mobile networks, where the propagation model must account for differences from existing far-field fifth-generation (5G) systems. The spatial density and the capacity of electromagnetic interactions between neighboring elements in ELAAs introduce unique considerations for near-field position estimation. Understanding and addressing these complexities will be the primary focus of this research. This dissertation delves into two interconnected aspects of source localization: two-dimensional (2D) source localization in the near-field and three-dimensional (3D) source localization in the near-field. In our 2D source localization approach, specifically for coordinates [x,y], we propose an innovative method that combines the adaptive subspace estimation with the sub-array architecture to accurately locate users in near-field scenarios. By including the sub-array techniques tailored for ELAAs, we explore the rotational invariance in each sub-array to implement the PAST (Projection Approximation Subspace Tracking) algorithm. This approach is computationally efficient due to its recursive update formula, which negates the necessity for computationally intensive tasks like matrix inversions or eigenvalue decompositions. In terms of precision, our method surpasses existing approaches, as demonstrated in graphs of root-mean-square error (RMSE) and graphs of cumulative distributed function (CDF) evaluations. Furthermore, the accuracy of the PAST algorithm at a distance of 3.5 meters is 0.0250 meters at the 10th percentile, outperforming other source localization methods. Similarly, at 25 meters, PAST achieves an accuracy of 0.3983 (10th percentile). These results highlight PAST’s accuracy and reliability for precise source localization in near-field scenarios, making it a robust choice for such applications. For 3D source localization, specifically for coordinates [x,y, z], we introduce a novel threedimensional position (3D-P) estimation method designed for wireless systems employing Uniform Rectangular Arrays (URAs). This approach virtually partitions the array into subarrays, each is tasked with independently estimating azimuth and elevation angles. To handle the multidimensional data effectively, we employ Higher-Order Singular Value Decomposition (HOSVD), reducing tensor size for a more concise representation of data structure, particularly beneficial in URA applications. Additionally, we utilize Taylor series approximation to address non-linear least square problems, contributing to accurate position estimations, even in intricate scenarios with 8 scatters. Our approach showcases the algorithm’s efficacy in mitigating multipath interference, with noise power exerting minimal influence. The results indicate that sub-meter accuracy is attainable at 30 and 40 dB SNRs for the 2 and 8 scatters in all percentiles, emphasizing the robustness of the technique in favorable conditions. 

COMMITTEE MEMBERS:
Externo à Instituição - FELIX DIETER ANTREICH - ITA
Interno - 1721829 - ADONIRAN JUDSON DE BARROS BRAGA
Presidente - 3157968 - DANIEL COSTA ARAUJO
Interno - 1771918 - UGO SILVA DIAS