Overcooked-AI

We suppose to apply traditional offline reinforcement learning technique to multi-agent algorithm.
In this repository, we implemented behavior cloning(BC), offline MADDPG, MADDPG+REM (MADDPG w/ REM), MADDPG+BCQ (MADDPG w/ BCQ) with pytorch. Now, BCQ is in ' Working In Progress', and it's not implemented completely.

We collected 0.5M multi-agent offline RL dataset and experimented with each comparison methods. We collected this data with online MADDPG agents, and it includes exploration trajectories using OU noise. The experiments are ran on Asymmetric Advantages on the Overcooked environment.

We are looking forward your contribution!

How to Run

Collect Offline Data

python train_online.py agent=maddpg save_replay_buffer=true

While the agents train with 0.5M steps, the trajectory replay buffer will be dumped in your experiment/{date}/{time}_maddpg_{exp_name}/buffer folder.
Please replace the path in config/data/local.yaml to the experiment by-product directory.

Download Dataset

Or, if you want to use our dataset pre-collected, please enjoy this link.
We provide 0.5M trajectories in Asymmetric Advantages layout.
Please download our dataset in your local computer and replace the path in config/data/local.yaml

Train Offline Models

Behavior Cloning

python train_bc.py agent=bc data=local

Offline MADDPG (Vanilla)

python train_offline.py agent=maddpg data=local

Offline MADDPG (w/ REM)

python train_offline.py agent=rem_maddpg data=local

Offline MADDPG (w/ BCQ) (WIP)

python train_offline.py agent=bcq_maddpg data=local

Result

Graph

Online	Offline (0.5M Data)	Offline (0.25M Data)

Video

Online	BC	Offline /w REM

bic4907/Overcooked-AI