Fusão de dados para a localização e navegação de robôs móveis em espaços inteligentes programáveis baseados em visão computacional

• Ano de defesa: 2025
• Autor(a) principal: Oliveira, Matheus Dutra de
• Orientador(a): Vassallo, Raquel Frizera
• Banca de defesa: Cordeiro, Rafael de Angelis , Fernandes, Mariana Rampinelli
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Espírito Santo; Mestrado em Engenharia Elétrica
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Elétrica
• Departamento: Centro Tecnológico
• País: BR
• Palavras-chave em Português: Odometria visual de múltiplas vistas; Localização de Monte Carlo; Fusão de dados; Filtro de Kalman; Multi-View visual odometry; Monte Carlo localization; Data fusion; Kalman filter
• Área do conhecimento CNPq: Engenharia Elétrica
• Link de acesso: http://repositorio.ufes.br/handle/10/20480

The estimation of mobile robot localization in indoor environments is one of the central challenges of autonomous navigation. Among the main techniques used to address this problem are Multi-View Visual Odometry, obtained through a multi-camera network, and Monte Carlo Localization. Both approaches have limitations: areas without camera coverage render navigation unfeasible when relying solely on visual odometry, while symmetric environments hinder convergence in the Monte Carlo method. Aiming to overcome these issues and achieve a more robust and reliable localization estimate, this work proposes the combination of these two global localization techniques through a data fusion approach based on Kalman Filter methods (Extended Kalman Filter and Unscented Kalman Filter). Additionally, the integration of the smart space architecture with the Robot Operating System (ROS) is adopted to implement this fusion. As a result, the fused localization can be integrated into the ROS navigation stack, leading to a complete localization and navigation system, and allowing the system to be triggered by other components of the smart environment. The system was evaluated in critical scenarios and case studies conducted in real environments. The results indicate that the information fusion effectively addresses the inherent limitations of each localization source, while increasing the robot’s global orientation accuracy by up to 12% and improving localization estimates by more than 5.2% when both sources are available.