• autonomous vehicle
• gaussian process
• mapping
• path planning
• Posidonia Oceanica

Over recent decades, we have witnessed a wide scale deterioration in seagrass habitats in the Mediterranean Sea. The study of this regression requires high quality data that only new methods can provide. This thesis introduces the problem of modelling Posidonia Oceanica meadows distribution in order to plan future missions of autonomous vehicles. Specifically, sparse and dense Gaussian Processes (GP) are used to solve the regression/classification problem and to offer a solid platform to apply Informative Path Planning (IPP). A genetic IPP algorithm is used and the various GP configurations are compared on synthetic data. Finally, the most suitable GP model is selected and applied to real data.