This repository is the official implementation of Sample-Efficient Multiagent Reinforcement Learning with Reset Replay, which is published in ICML 2024.
Please install the environment as indicated in the subfolder of '/marr-smac'.
To run MARR-QMIX in SMAC on the map of 5m_vs_6m, run this command after changing directory into '/marr-smac':
python src/main.py --config=rrmix --env-config=sc2 with env_args.map_name=5m_vs_6m seed=1 t_max=1010000 > 1.log 2>&1 &
Change the map_name to run on different maps.
The configuration could be revised in /src/config/algs/rrmix.yaml.
Please install the environment as indicated in the subfolder of '/marr-mpe'.
We only need to install MPE, please use the environment files in '/marr-mpe/src/envs/multiagent-particle-envs'.
To run MARR-FACMAC in MPE on the task of 3a, run this command after changing directory into '/marr-mpe':
python src/main.py --config=rrfacmac_pp --env-config=particle with env_args.scenario_name=continuous_pred_prey_3a t_max=2000000 > 1.log 2>&1 &
Change the scenario_name to run on different tasks.
The configuration could be revised in /src/config/algs/rrfacmac_pp.yaml.