Modelli e algoritmi con applicazioni alla gestione degli incendi

In the last twenty years, wildfires have grown in size and frequency. The effect of a wildfire on the environment can be long-lasting; therefore it is important to identify and quantify the risks of a wildfire in a territory to adopt adequate preventive countermeasures. Choosing between alternative prevention strategies in a such complex scenario, need the integration of science and management. This thesis focuses on the design of models and algorithms useful in the context of fire preparedness measures and addresses the problem of the coordination of multi-robot systems for wildfire monitoring and intervention. We present a graph model able to evaluate the risk of fire in a territory considering the probabilities of fire ignition and propagation in a certain area. To prove the usability of the model, we applied it to a territory in the North of Corsica, an island of the Mediterranean area exposed to a high risk of wildfires due to dry and hot weather during summer. This case study constitutes a proof of concept showing how the model can be applied practically. We also present a prototype of an easy to use web-application designed for fire and risk managers. It includes maps and built-in algorithms to help the planning and the evaluation of fire preventive strategies, like for example the installation of firebreaks. Firebreaks consist of a strip of land in which the fuel is removed. As a result, fire propagation beyond them is blocked or slowed down. We define the firebreak location problem, to address the optimal positioning of firebreaks, study its complexity and present some cases solvable in polynomial time as well as heuristics. We then study algorithmic aspects of the coordination of multi-robotic systems. Robot technology has advanced quickly, assisting humans in increasingly complicated tasks. The development of robots able to operate in forest environments can help in tasks like firefighting and fire prevention by land monitoring. Autonomous, multi-robot systems can cooperate and self-organize providing a robust alternative to application-specific robots. However, the cooperation of multi-robotic systems is a challenging coordination task requiring algorithmic solutions. We address the problem from a theoretical point of view and present algorithms for the coordination of a group of robots. Then, we introduce MOBLOT, a new model for swarm and modular robotics, in which robots can cluster to create bigger computational units called molecular robots. MOBLOT allows us to model a swarm divided into sub-groups of robots that deploy and move in formation.