Pareto Conditioned Networks

This repository contains the code used for:

Reymond, M., Bargiacchi, E., & Nowé, A. (2022, May). Pareto Conditioned Networks. In Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems (pp. 1110-1118).

You can read the paper here.

Dependencies

The code requires Python3.7+ as well as torch for the neural networks, gym for the environments, h5py for logging and opencv-python for preprocessing of image-observations.

How to run

Here is how you run PCN on Deep Sea Treasure:

python main_pcn.py --env dst

This will create a log directory in /tmp/pcn. It also contains checkpoints of the learned policies. You can then execute any of the learned policies as follows:

python eval_pcn.py <logdir>

Optionally, you can add an --interactive flag if you want to manually select policies.