UNMAS: Multi-Agent Reinforcement Learning for Unshaped Cooperative Scenarios

3m

2s3z

5m vs 6m

8m

3s5z

3s5z vs 3s6z

3 Marines (left is controlled by UNMAS).

2 Stalkers and 3 Zealots (left is controlled by UNMAS).

5 Marines (left is controlled by UNMAS) vs 5 Marines.

8 Marines (left is controlled by UNMAS).

3 Stalkers and 5 Zealots (left is controlled by UNMAS).

3 Stalkers and 5 Zealots (left is controlled by UNMAS) vs 3 Stalkers and 6 Zealots.