Distributed Multi-Robot Control for Streets Surveillance from Aerial Images with Neural Networks

We propose a distributed strategy for aerial robot teams to maximize street coverage in monitoring and search-and-rescue missions. A neural network extracts roads from aerial images, generating a probability density via Gaussian Mixture Models to guide robots toward detected streets. Each robot performs a Voronoi-based coverage mission, focusing on areas of interest. The neural network's performance was validated using Google Maps images, while the control architecture was tested in realistic simulations, demonstrating effectiveness.