This is the codebase of improved EOI on general SMAC tasks. We have also published EOI on sparse-reward so_many_baneling in this repo.
python3 src/main.py --config=eoi with explore_ratio=0.2 --env-config=sc2
Two parameters explore_ratio and episode_ratio control the strength of EOI exploration. The improved version of EOI could be seen as a multi-agent exploration method.
| 2c_vs_64zg | 3s_vs_5z | 5m_vs_6m | |
|---|---|---|---|
| explore_ratio | 0.2 | 0.8 | 0.2 |
The agents are more likely to benefit from individualized behaviors if the trajectory is longer.
3s_vs_5z and 5m_vs_6m are tested on SC2.4.10, and 2c_vs_64zg are tested on SC2.4.16, since the 2c_vs_64zg on SC2.4.10 is too easy.