https://github.com/marcometer/action-space-compositions-in-deep-reinforcement-learning

Application of concurrent discrete and continuous actions on two novel DRL environments to mimic human input devices.
https://github.com/marcometer/action-space-compositions-in-deep-reinforcement-learning

Last synced: 5 days ago
JSON representation

Application of concurrent discrete and continuous actions on two novel DRL environments to mimic human input devices.

• Host: GitHub
• URL: https://github.com/marcometer/action-space-compositions-in-deep-reinforcement-learning
• Owner: MarcoMeter
• Created: 2018-11-06T13:26:13.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2020-01-16T12:02:25.000Z (almost 5 years ago)
• Last Synced: 2024-10-11T05:51:00.969Z (about 1 month ago)
• Language: Jupyter Notebook
• Homepage:
• Size: 115 MB
• Stars: 3
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Action Spaces in Deep Reinforcement Learning to Mimic Human Input Devices

This repository contributes the environments Shooting Birds and Beastly Rivals Onslaught, which were examined in the underlying paper (headline name).

Hyperparameters can be retrieved from the [config](https://github.com/MarcoMeter/Action-Space-Compositions-in-Deep-Reinforcement-Learning/blob/master/config/trainer_config.yaml).

A video showing the results can be found on [Youtube](https://www.youtube.com/watch?v=Pb14i3srRWc&feature=youtu.be).

# Dependencies

This project was created with [Unity's ML-Agents Toolkit](https://github.com/Unity-Technologies/ml-agents) v0.7 along with the Unity Engine version 2018.2.20f.

To install the required python packages run:

```

pip install -e /ml-agents/.

pip install tensorflow==1.7.*

```

Checkout the "update" branch if you are looking for project files using more recent dependencies (ml-agents & Unity).

# Citing this paper

Link to the paper:

https://ieeexplore.ieee.org/document/8848080

```

@inproceedings{Pleines2019,

  author    = {Marco Pleines and

               Frank Zimmer and

               Vincent{-}Pierre Berges},

  title     = {Action Spaces in Deep Reinforcement Learning to Mimic Human Input

               Devices},

  booktitle = {{IEEE} Conference on Games, CoG 2019, London, United Kingdom, August

               20-23, 2019},

  pages     = {1--8},

  year      = {2019},

  url       = {https://doi.org/10.1109/CIG.2019.8848080},

  doi       = {10.1109/CIG.2019.8848080}

}

```