Multi-agent reinforcement learning for planning and conflict resolution in a dynamic domain