| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 5.69 MB | Adobe PDF |
Advisor(s)
Abstract(s)
A sequente dissertação resulta do desenvolvimento de um sistema de navegação subaquático para um Remotely Operated Vehicle (ROV). A abordagem proposta consiste de um algoritmo em tempo real baseado no método de Mapeamento e Localização Simultâneo (SLAM) a partir de marcadores em ambientes marinhos não estruturados.
SLAM introduz dois principais desafios: (i) reconhecimento dos marcadores provenientes dos dados raw do sensor, (ii) associação de dados. Na detecção dos marcadores foram aplicadas técnicas de visão artificial baseadas na extracção de pontos e linhas.
Para testar o uso de features no visual SLAM em tempo real nas operações de inspecção subaquáticas foi desenvolvida uma plataforma modicada do RT-SLAM que integra a abordagem EKF SLAM. A plataforma é integrada em ROS framework e permite estimar a trajetória 3D em tempo real do ROV VideoRay Pro 3E até 30 fps. O sistema de navegação subaquático foi caracterizado num tanque instalado no Laboratório
de Sistemas Autónomos através de um sistema stereo visual de ground truth.
Os resultados obtidos permitem validar o sistema de navegação proposto para veículos subaquáticos. A trajetória adquirida pelo VideoRay em ambiente controlado é validada pelo sistema de ground truth. Dados para ambientes não estruturados, como um gasoduto, foram adquiridos e obtida respectiva trajetória realizada pelo robô. Os dados apresentados comprovam uma boa precisão e exatidão para a estimativa da posição.
This thesis presents a subaquatic navigation system for a Remotely Operated Vehicle (ROV). The proposed approach consists in a real time algorithm based in Simultaneous Localisation and Mapping (SLAM) with landmarks in unstructured subaquatic environments. SLAM introduces two main challenges: (i) to recognize the landmarks from raw sensor data, (ii) data association. In the landmark detection process a artificial vision techniques based on point and line extraction was applied. To test the use of environment features in real-time visual slam underwater robotic inspection operations was developed a modified implementation of the RT-SLAM approach that integrates EKF SLAM. The platform is integrated within a ROS framework and allows to estimate an underwater remote operated vehicle trajectory (ROV) in realtime (at 30fps). The subaquatic navigation system was characterized in a teste tank at Autonomous Systems Laboratory. The obtained results allows to validate the proposed navigation system for subaquatic vehicles. The acquired trajectory by the ROV in the controlled environment it is validated by ground truth. Results for unstructured subaquatic environments were also acquired, such as a pipeline, in the sea bottom scenario were the performed robot trajectory is consistently obtained by the system. The presented data validates the accuracy and precision for position estimation.
This thesis presents a subaquatic navigation system for a Remotely Operated Vehicle (ROV). The proposed approach consists in a real time algorithm based in Simultaneous Localisation and Mapping (SLAM) with landmarks in unstructured subaquatic environments. SLAM introduces two main challenges: (i) to recognize the landmarks from raw sensor data, (ii) data association. In the landmark detection process a artificial vision techniques based on point and line extraction was applied. To test the use of environment features in real-time visual slam underwater robotic inspection operations was developed a modified implementation of the RT-SLAM approach that integrates EKF SLAM. The platform is integrated within a ROS framework and allows to estimate an underwater remote operated vehicle trajectory (ROV) in realtime (at 30fps). The subaquatic navigation system was characterized in a teste tank at Autonomous Systems Laboratory. The obtained results allows to validate the proposed navigation system for subaquatic vehicles. The acquired trajectory by the ROV in the controlled environment it is validated by ground truth. Results for unstructured subaquatic environments were also acquired, such as a pipeline, in the sea bottom scenario were the performed robot trajectory is consistently obtained by the system. The presented data validates the accuracy and precision for position estimation.
Description
Keywords
Sistema de Navegação Subaquático SLAM EKF SLAM ROV Ground truth Subaquatic Navigation System