Towards autonomous networks of drones

Unmanned Aerial Vehicles (UAVs) have enormous potential in the public and civilian domain as an emerging technology that allows to carry out missions where human intervention would be impossible, unsafe, or just inefficient. While single remotely controlled drones are used in a wide range of scenarios, an extraordinary improvement could be reached by employing autonomous networks of drones. Current work in this direction highlights significant issues caused by the dynamic topology and the limited energy of the drones, especially in safety-critical missions. In this thesis, we investigate solutions and methodologies for the coordination and communication of UAVs to enable autonomous networks of drones in real applications. We focus on drones' physical constraints and we study algorithms' applicability in real field. We first address the problem of trajectories planning for safety-critical operations and we propose several solutions including approximation and genetic-based algorithms. Then we exploit the drones' communication capabilities to design efficient algorithms in a broader range of scenarios. We consider missions with uncertain target positions, in which drones coordinate by sharing their local observations, and missions in which drones offload the collected data to the depot through a multi-hop connection. To improve drone communication capabilities, we propose an enhanced Routing protocol with Deep Q-Learning. The protocol is designed to work on critical scenarios requiring urgent communication that can tolerate only a bounded delay. Finally, we investigate three novel applications for UAV networks to demonstrate how drones can populate our cities in the next years. We design DRUBER, a distributed parcel delivery system aided by a blockchain framework; we propose DANGER, an emergency network of drones to provide WiFi connection to any smartphone in case of disasters; and we discuss an application of drones for Food Safety and Security.