Monte Carlo Augmented Actor-Critic for Sparse Reward Deep Reinforcement Learning from Suboptimal Demonstrations

Author implementation of 'Monte Carlo Augmented Actor-Critic for Sparse Reward Deep Reinforcement Learning from Suboptimal Demonstrations'

Read the paper here.

Installation

1. Make a python3.7+ virtualenv: virtualenv venv --python=/path/to/python3.7
2. Activate it: source venv/bin/activate
3. Install pip install -e .. Requires python 3.7+.
4. (Optional) If you want to run robotics experiments, download the mujoco200 binary and licence here. Run pip install mujoco-py==2.0.2.13.
5. (Optional) If you want to run Robosuite experiments (Lift and Door), pip install robosuite

Running

Simply run the commands in commands.sh. The first time you run commands for a particular enviornment you'll need to add the --gen-data flag in order to generate the necessary offline data.

Bibtex

@inproceedings{
    wilcox2022monte,
    title={Monte Carlo Augmented Actor-Critic for Sparse Reward Deep Reinforcement Learning from Suboptimal Demonstrations},
    author={Albert Wilcox and Ashwin Balakrishna and Jules Dedieu and Wyame Benslimane and Daniel S. Brown and Ken Goldberg},
    booktitle={Thirty-Sixth Conference on Neural Information Processing Systems},
    year={2022},
    url={https://openreview.net/forum?id=FLzTj4ia8BN}
}