An Artificial Intelligent-based strategy for Sheltering Factor computation

Usage of Unmanned Aerial Systems (UASs) is widely diffused and will involve different aspects of our life in cities in the future. Nowadays UASs are used to accomplish several applications: data collections, surveillance, logistics, search and rescue, to name a few. Despite the rapid growth of this technologies, currently most of the National Aviation Authorities impose strict regulations to the UAS missions pratically hindering autonomous flights. The majority of these operations must be performed in Visual Line-Of-Sight (VLOS) and, to guarantee operations over urban area, they require the use of harmless drones, even if in the recent years the authorities permit to perform operations also Beyond Visual Line-Of-Sight (BVLOS) even over cities. This thesis is based on a solid preliminary work done by the research group in which a risk-aware path planning strategy for UASs has been developed to determine a safe flight mission evaluating the risk caused by the UAS of flying over populated areas. Specifically, the risk is estimated by generating a risk-map and considering several parameters, such as the population density and the sheltering factor.The objective of this thesis is to provide a valid strategy using a Convolutional Neural Network to estimate the Sheltering Factor of a specific area, these values will be used as inputs for the Ris-kMap definition. The results will be then compared with the Sheltering Factor values obtained from the definition of obstacles layer considering the model of the city of Turin from OpenStreetMap (OSM). This research activity is involved within the projects of the Amazon Research Award “From Shortest to Safest Path Navigation: An AI-Powered Framework for Risk-Aware Autonomous Navigation of UASs”