Drone Swarm System in Firefighting Activities

Defence against forest fires generally takes place through a variety of air/ground interventions, and is based on the coordination of various actors, such as civil protection, fire brigades, police forces, and volunteers. The recent technological development of Unmanned Aerial Vehicles (UAV), also called drones, and their ability to handle situations that are too dangerous for humans can provide a breakthrough in fighting wildland fires. Compared to a single UAV, a cooperative UAV system can perform complex tasks with more safety and efficiency. In this thesis, an innovative forest firefighting system, based on the use of a platform managing a swarm of UAVs, is proposed and investigated. Owing to the various modular subsystems of the platform, such as landing pads as well as cartridge and hive, drones are served, supplied, and housed. Moreover, automatic battery and payload replacement along with extinguishing liquid refilling ensure the continuity of the action. The validity of the approach in Mediterranean scrub fires is illustrated, first computing the critical water flow rate, then simulating the fire spread in a Cellular Automata model both in the absence of firefighting efforts and in the event of an intervention of drones able to generate a continuous flow of extinguishing liquid on the fire front. The fire spread is also simulated by means of a Level Set model, and Model Predictive Control is applied to control the fire front modified by water drop patterns. Moreover, such a drone-platform system could be used for 24-hour monitoring of an area using video and thermal cameras. Again, in this scenario, drones can spray liquids, water and/or retardants, to mitigate any critical situations of drought and low humidity of vegetation, or even extinguish identified outbreaks. However, managing a large number of drones on one or more platforms requires determining the optimal sequence of drone landings so that the gap between target and actual landing times is minimized, and so that drones have their batteries changed and receive the required payloads. Therefore, the drone scheduling issue is studied by formulating linear optimization problems related to drone landings and take-offs on single and multiple platforms. Finally, practical experiments in cooperation with the spin-off Inspire of the University of Genoa have been carried out and, in particular, the issue of precision landing with Real Time Kinematic positioning technology has been tested.