Back Lane Marking Registry: uma abordagem de localização e seguimento de caminho por veículos autônomos em via sinalizada

• Ano de defesa: 2017
• Autor(a) principal: Vivacqua, Rafael Peixoto Derenzi
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Espírito Santo; BR; Doutorado em Engenharia Elétrica; Centro Tecnológico; UFES; Programa de Pós-Graduação em Engenharia Elétrica
• Palavras-chave em Português: Localização e mapeamento simultâneos (SLAM); Veículos autônomos - Direção de automóveis; Visão por computador; Obstáculo - Localização; Dead Reckoning (Navegação); Engenharia Elétrica; 621.3
• Link de acesso: http://repositorio.ufes.br/handle/10/9691

This thesis aimed to contribute with the development of autonomous driving technology. More specifically, the aim of this work was to design, build and test all the hardware and software needed to adapt a commercial vehicle so that it is able to successfully perform a predetermined transport mission repeatedly, as if it was a train traveling on a virtual railway. The issue of dynamic obstacles was not addressed in this research. Three approaches widely used in the literature were studied: (i) direct path tracking by detection of lane markings, (ii) global localization tracking (GNSS), and (iii) path tracking by lane marking maps. A prototype vehicle was built, respecting the low cost philosophy of the project, to carry out the experimental tests. The first two approaches served as the basis for the development of the final approach, based on visual track marker maps. This approach has produced the most positive results of the work and represents the greatest contribution of this thesis: the method called Back Lane Marking Registry or BLMR. The BLMR allows the construction of accurate, reliable and extensive perception of lane markings in the vicinity of the vehicle from low-cost sensors. This extensive perception combined with a fast filtering algorithm and map matching techniques led to localization accuracy which is high enough for autonomous application, as demonstrated in tests conducted in Italy.