Collaborative training of heterogeneous reinforcement learning agents in environments with sparse rewards: what and when to share? | Publicación