The code is adapted from the source code for the IJCAI-23 paper Multi-Agent Intention Recognition and Progression, by Michael Dann, Yuan Yao, Natasha Alechina, Brian Logan, Felipe Meneguzzi and John Thangarajah, which can be found at https://github.com/mchldann/IJCAI_GR.
To install the Python requirements via Anaconda, use
conda env create -f environment.yml
.
To recreate the results, run:
python python_agent.py scenario_name
where scenario_name is one of {neutral_1, neutral_2, neutral_3, neutral_4}.
Agent scores are automatically logged to results/cooperative_craft_world_dqn/ folder.