RL-multiagent-communication

This is the GitHub repository for our course project, InSPIRe: Interpretable Sharing of Proactive Intention in Reinforcement-learning, for 10-701 Introduction to Machine Learning at Carnegie Mellon University, Fall 2024 semester.

Example.Game.Play.mov

Environment setup

We recommend creating a virtual environment for running this project. Please run the following commands to install the appropriate packages and install the Atari ROM.

$ pip install -r requirements.txt
$ AutoROM --accept-license

Running the training or evaluation experiments

To run a training or evaluation experiment, you should set up your own configuration file. Example training file is available at config_sample.py and an example evaluation file, with human-visible gameplay, is available at config_eval.py

$ python VolleyballPongEnv.py -c <path-to-your-config.py>

Generating the plots from the log files

To generate the final plots from the log specified in the config's output directory:

$ python ParseLog.py -i <path-to-log-file.log> -o <path-to-output-directory> -m "optional string for plot titles"