All Tools
O
AgentsFreeOpen Source
OVERCOOKED
Cooperative human-AI kitchen benchmark for multi-agent RL
MIT
ABOUT
Fully cooperative human-AI work needs shared layouts where partners specialize, hand off, and adapt without a scripted plan. Overcooked-AI turns kitchen cooking into a MARL benchmark so researchers can measure coordination with other agents and with people.
INSTALL
pip install overcooked-aiINTEGRATION GUIDE
1. Train two cooperative chefs to cook and deliver soups on cramped kitchen layouts
2. Evaluate human-AI coordination with the Overcooked web demo or logged play data
3. Compare MARL algorithms on layouts that require on-the-fly task allocation
4. Collect human-human or human-AI trajectories for imitation and preference research
TAGS
pythonreinforcement-learningmulti-agenthuman-aicoordinationbenchmark