Authors: Sharmin Rahman
Abstract: Exploration of underwater environments with autonomous robots could assist us in a variety of scenarios, ranging from historical studies to health monitoring of coral reef; underwater infrastructure inspection e.g., bridges, hydroelectric dams, water supply systems and oil rigs. Mapping underwater structures is important in several fields, such as, marine archaeology, Search and Rescue (SaR), resource management, hydrogeology, and speleology. However, due to the highly unstructured nature of such environments, navigation by human divers could be extremely dangerous, tedious and labor intensive. Hence, employing an underwater robot is an excellent fit to build the map of the environment while simultaneously localizing itself in the map. The contribution of this thesis is the design and development of a real-time robust Simultaneous Localization and Mapping (SLAM) algorithm for underwater domain. A novel tightly-coupled keyframe-based non-linear optimization framework with loopclosing and relocalization capabilities fusing Sonar, Visual, Inertial and Depth information has been presented. Introducing acoustic range information to aid the visual data in underwater, shows improved reconstruction. The availability of depth information from water pressure enables a robust initialization and refines the scale; as well as assists to reduce the drift due to the tightly-coupled formulation. In addition, we propose to augment the pipeline with magnetometer for a more accurate orientation estimation from the dead reckoning sensor. To address the denser reconstruction of the surroundings in a low lighting conditions, a contour-based reconstruction approach utilizing the well defined edges between the well lit areas and darkness has been developed. Furthermore, we propose a semi-direct sparse approach of reconstruction by jointly minimizing the photometric and reprojection error from direct method and indirect method respectively where indirect method is used for accurate tracking while high-gradient pixels help in reconstruction. Experimental results on datasets collected with a custom-made underwater sensor suite and an autonomous underwater vehicle (AUV) Aqua2 from challenging underwater environments with poor visibility demonstrate performance never achieved before in terms of accuracy and robustness.