Learning Representations via a Robust Behavioral Metric for Deep Reinforcement Learning

This repository contains the code for the paper "Learning Representations via a Robust Behavioral Metric for Deep Reinforcement Learning" (RAP). If use this code, please cite our paper:

@inproceedings{
chen2022learning,
title={Learning Representations via a Robust Behavioral Metric for Deep Reinforcement Learning},
author={Jianda Chen and Sinno Pan},
booktitle={Advances in Neural Information Processing Systems},
editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},
year={2022},
url={https://openreview.net/forum?id=7YXXt9lRls}
}

Clone This Repo

git clone --recurse-submodules https://github.com/jianda-chen/RAP_distance.git

Or

git clone https://github.com/jianda-chen/RAP_distance.git
cd RAP_distance
git submodule update --init --recursive

Dependencies

• Python >= 3.8
• PyTorch >= 1.7
• DeepMind Control Suite. Please refer to dm_control page to install the dependencies for DeepMind Control Suite
• Other python packages. Install other python dependencies in conda environment by the following command:

  conda env update -f conda_env.yaml

  Getting Started

  To run the Distracting DeepMind Control Suite example, simply run the following command:

  ./run_local.sh

  This script will save the experiment records in ../log directory.

If you need to change the saving directory or run other tasks, please revise the file run_local.sh accordingly.

License

This work is under CC-BY-NC 4.0 license.