Emergent behaviors and scalability for multi-agent reinforcement learning-based pedestrian models | Publicación