Dataset Cards - OMIGA

Dataset Cards - OMIGA

2c_vs_64zg

2c_vs_64zg - Download

Metadata

Environment nameVersionAgentsAction typeObservation sizeReward type
SMAC (v1)Modified version of SMAC v1, popularised by MAPPO 2Discrete[478]Dense

Generation procedure for each dataset

Converted from omiga format to a Vault.

Summary statistics

UidEpisode return meanMin returnMax returnTransitionsTrajectoriesJoint SACo
Poor8.91 ± 1.012.5310.00108303481.00
Medium13.00 ± 1.3910.0115.003794010011.00
Good19.94 ± 1.2615.1821.615921510011.00

6h_vs_8z

6h_vs_8z - Download

Metadata

Environment nameVersionAgentsAction typeObservation sizeReward type
SMAC (v1)Modified version of SMAC v1, popularised by MAPPO 6Discrete[172]Dense

Generation procedure for each dataset

Converted from omiga format to a Vault.

Summary statistics

UidEpisode return meanMin returnMax returnTransitionsTrajectoriesJoint SACo
Poor9.12 ± 0.814.809.992425510011.00
Medium11.97 ± 1.2610.0014.992951110011.00
Good17.84 ± 2.1515.0120.023804010011.00

5m_vs_6m

5m_vs_6m - Download

Metadata

Environment nameVersionAgentsAction typeObservation sizeReward type
SMAC (v1)Modified version of SMAC v1, popularised by MAPPO 5Discrete[124]Dense

Generation procedure for each dataset

Converted from omiga format to a Vault.

Summary statistics

UidEpisode return meanMin returnMax returnTransitionsTrajectoriesJoint SACo
Poor8.50 ± 1.191.819.892274710010.96
Medium11.03 ± 0.5810.0811.962771710010.95
Good20.00 ± 0.0020.0020.002773410010.96

corridor

corridor - Download

Metadata

Environment nameVersionAgentsAction typeObservation sizeReward type
SMAC (v1)Modified version of SMAC v1, popularised by MAPPO 6Discrete[346]Dense

Generation procedure for each dataset

Converted from omiga format to a Vault.

Summary statistics

UidEpisode return meanMin returnMax returnTransitionsTrajectoriesJoint SACo
Poor4.93 ± 1.710.009.995126810011.00
Medium13.07 ± 1.2710.0214.9912601210011.00
Good19.88 ± 1.0115.0120.4910017010011.00

6halfcheetah

6halfcheetah - Download

Metadata

Environment nameVersionAgentsAction typeObservation sizeReward type
MAMuJoCoV1.0, Mujoco v2006Continuous[23]Dense

Generation procedure for each dataset

Converted from omiga format to a Vault.

Summary statistics

UidEpisode return meanMin returnMax returnTransitionsTrajectoriesJoint SACo
Medium-Replay655.76 ± 590.40-198.772132.60100100010001.00
Medium-Expert2105.38 ± 1073.24251.943866.09200200020001.00
Medium1425.66 ± 520.12251.942113.52100100010001.00
Expert2785.10 ± 1053.14317.943866.09100100010001.00

2ant

2ant - Download

Metadata

Environment nameVersionAgentsAction typeObservation sizeReward type
MAMuJoCoV1.0, Mujoco v2002Continuous[113]Dense

Generation procedure for each dataset

Converted from omiga format to a Vault.

Summary statistics

UidEpisode return meanMin returnMax returnTransitionsTrajectoriesJoint SACo
Medium-Replay1029.51 ± 141.27895.371517.06175175017500.66
Medium-Expert1736.88 ± 319.64840.772124.15200200020001.00
Medium1418.70 ± 37.04840.771473.86100100010001.00
Expert2055.07 ± 22.071994.032124.15100100010001.00

3hopper

3hopper - Download

Metadata

Environment nameVersionAgentsAction typeObservation sizeReward type
MAMuJoCoV1.0, Mujoco v2003Continuous[14]Dense

Generation procedure for each dataset

Converted from omiga format to a Vault.

Summary statistics

UidEpisode return meanMin returnMax returnTransitionsTrajectoriesJoint SACo
Medium-Replay746.42 ± 671.8970.762801.15131482641601.00
Medium-Expert1190.61 ± 973.4095.273762.69191978254811.00
Medium723.57 ± 211.66128.382776.4991939140001.00
Expert2452.02 ± 1097.8695.273762.69100039114811.00