Adaptation dynamique d'un drone aux perturbations environnementales par apprentissage

The demonstrator presented introduces a case study of learning by system multi-agent in a simulated environment. The core of this work is the enrichment of a learning experience by demonstration using a mechanism of self-taught learning. This mechanism allows the learning agent to exceed the limits introduced by the demonstration process by allowing him to adapt his behaviour to deal with unforeseen situations. This learning involves a robotic task: learning to follow a trajectory in order to airborne drone. The self-directed learning mechanism provides more robustness to the learning phase.