Algoritmos de localização em ambientes externos aplicando rede LoRaWan

• Ano de defesa: 2022
• Autor(a) principal: Telles, Guilherme Pazetto
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Brasil; Programa de Pós-Graduação em Sistemas de Energia; UTFPR
• Palavras-chave em Português: Redes remotas (Redes de computadores); Algorítmos; Filtragem de Kalman; Sistema de Posicionamento Global; Wide area networks (Computer networks); Algorithms; Kalman filtering; Global Positioning System; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA::ELETRONICA INDUSTRIAL, SISTEMAS E CONTROLES ELETRONICOS::AUTOMACAO ELETRONICA DE PROCESSOS ELETRICOS E INDUSTRIAIS; Engenharia Elétrica
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/28561

In this paper, we introduce two different algorithms to localization of LoRaWAN devices on large scale outdoor area. The first one is Weighted Centroid (WC), classifying each gateway accordingly to the Received Signal Strength (RSSI) and Signal Noise Ratio (SNR) from the received package of the Target Node (TN). The second one also applies RSSI technique, however, differently of the previous algorithm, combines with to Outlier Detection method named Local Outlier of Probability (LoOP), selecting intersections points from circumferences. As comparing algorithms, we also will apply a Multilateraion (MLT) method, as a Time Difference of Arrival (TdoA) to the same data for results comparison. Looking for better precision estimation, Kalman Filter (KF) was applied to the package series from each TN. To validate the algorithms, they were applied to a database from Antwerp, Belgium, comparing different LoRa behaviors and algorithms characteristics, and also comparing other localization studies developed on the same data. As result, the WC+FK reach a mean error of 566,86 m, while LoOP+FK had 569 m mean. The median from both were 399,04 and 424,38 respectively. These result were better than the compared MLT and TdoAm, which had a mean error applying KF of 1824,94 and 655,03 m respectively. This demonstrates that WC+KF and LoOP improved by 31% and 8.6% the MLT+KF and TdoA+KF ones respectively. The proposals algorithms are also closer to a more complex method as Fingerprinting (FP), with 340 m mean error applied to the same database.