Evolução de estratégias e controle inteligente em sistemas multi-robóticos robustos

Abstract

This work is related to the application of Artificial Intelligence techniques to develop a Multi-Agent Robotic System applied to the problem of monitoring wild forest fires and to the

execution of fire fighting actions. Our main goal was to evolve strategies (control unit) in order to define the positioning of the fire-fighting autonomous robotic team and to create robust navigation methods used to control robotic agents (combat units). This work was developed based on simulations accomplished using a realistic 3D virtual environment, specially implemented for this purpose, using the software libraries OSG, Demeter and ODE. In the proposed system, a team of autonomous agents work cooperatively in order to successfully perform the identification and fighting of forest fires, without any human intervention. The 3D virtual environment includes several features for realistic simulation of this task, as for example, adoption of irregular terrains, natural processes simulation (e.g. fire propagation), and simulati