<scenario_think>
1, All agents coordinate together to cover all landmarks.
2, Need to think how to assign landmarks to each agent to minimal total distance.
3, If the agents' actions follow the llm suggestions, reward should be 1.
</scenario_think>