Project realized by Alessio Zobele and Andrea di Cataldo
Abstract: The rapid proliferation of drone technologies in recent years has enabled a wide range of new applications in cooperative sensing, mapping, and navigation. This report investigates how three drones can exchange information to localize themselves, identify landmarks, construct a shared map of the environment, and ultimately reach consensus despite having access to only a limited number of common reference objects. We examine both the case in which all drones possess equally accurate proprioceptive and exteroceptive sensors, and the scenario in which sensing precision varies across agents. For map recognition and obstacle identification, we employ a modified WLS algorithm tailored to grid-based environments. Additional methods used include Covariance Intersection for information fusion under intermittent communication constraints, the Extended Kalman Filter for self positioning, Metropolis--Hastings consensus and row-stochastic weighting strategies for distributed agreement, centralized fusion for cooperative localization of critical map features, and the trilateration for allow the drones to orientate with respect to a knwon Landmark. Once the target is detected, a centralized fusion strategy refines the global map, and the A* algorithm computes an optimal route for rescue units. Together, these techniques provide a robust framework for collaborative autonomous mapping in environments with sparse shared landmarks.