Undesired state-action prediction in multi-agent reinforcement learning for linked multi-component robotic system control | Publicación