Create assistive agents to help humans with real-time control tasks.
Install Python dependencies with the pip package manager using
pip install -r requirements.txt
Install gym and replace gym/envs/box2d/lunar_lander.py with deepassist/lunar_lander_for_{bots,humans}.py.
Install baselines and replace baselines/baselines/deepq/simple.py with deepassist/simple.py.
Extract this zip file into deepassist/.
For the quadrotor experiments, install ardrone_autonomy and vicon_bridge.
Please contact the author at sgr [at] berkeley [dot] edu if you have questions or find bugs.
If you find this software useful in your work, we kindly request that you cite the following paper:
@InProceedings{Reddy/etal/18a,
title={Shared Autonomy via Deep Reinforcement Learning},
author={Reddy, Siddharth and Dragan, Anca D. and Levine, Sergey},
booktitle={Arxiv 1802.01744},
year={2018},
url={https://arxiv.org/abs/1802.01744}
}