Deep Deterministic Policy Gradient (DDPG) in a multi-agent particle environment

This python file implements the Deep Deterministic Policy Gradient algorithm in a multiagent tag-game gym environment, slightly modified by Rohan Sawhney. I did it as a final assignment for Move37, a School of AI course.

Getting started

Use the following command with the desired NUMBER_OF_EPISODES to lauch the file and start training :

python episode.py --n_episodes NUMBER_OF_EPISODES

Further hyperparameter tuning can be done by passing arguments listed in episode.py.

Prerequisites

You need to install both gym and PyTorch to run the file.

pip install gym

PyTorch installation

Result

Here we plot the length of episodes as training goes on :

License

This project is under the MIT License - see the LICENSE file for details.

Acknowledgements

Inspired from Rohan Sawhney's implementation.