Titre: Machine learning for road active safety

Abstract: Lane change is one of the most common reasons for road accidents. Thus, designing an efficient collision-less lane change maneuver is of great importance. On the other hand, drones have been deployed in vehicular networks to provide various services.
Under these considerations, we propose a lane change assistance platform, GL-DEAR, with both global control by drones and local control by the ego vehicle, based on deep reinforcement learning. Simulation results prove that GL-DEAR is able to achieve collision-less trips while reducing total travel time, and can adapt to complex scenarios with the presence of random risks and emergency vehicles.