large-scale safety-critical control benchmarks for reinforcement learning algorithms
Project description
controlgym
Description
Controlgym
provides 36 safety-critical industrial control environments and 10 infinite-dimensional PDE-based control problems with continuous, unbounded action and observation spaces that are inspired by real-world applications. This project supports the Learning for Dynamics & Control (L4DC) community, focusing on vital issues: convergence of reinforcement learning (RL) algorithms in policy development, stability, safety, and robustness of learning-based controllers, and the scalability of RL algorithms to high and potentially infinite-dimensional systems. We provide a detailed description of controlgym
in this paper.
Installation
Clone the Repository
To get started, clone the controlgym
repository and navigate to folder:
git clone https://github.com/xiangyuan-zhang/controlgym.git
cd controlgym
Windows Installation
# Step 1: create and activate a virtual environment (Optional)
# Example a: using venv
py -3.10 -m venv controlgym-env
.\controlgym-env\Scripts\activate.bat
# Example b: using conda
conda create -n controlgym-env python=3.10
conda activate controlgym-env
# Step 2: install pytorch with cuda (optional)
pip3 install --upgrade pip
pip3 install torch --index-url https://download.pytorch.org/whl/cu121Install the repository
# Step 3: install the controlgym repository
# Example a: using pip
pip3 install -e .
# Example b: using poetry
poetry install
# Step 4: deactivate the virtual environment (Optional)
# For venv
.\controlgym-env\Scripts\deactivate.bat
# For conda
conda deactivate
Linux/MacOS Installation
# Step 1: Create and activate a virtual environment (Optional)
# Example a: using venv
python3.10 -m venv controlgym-env
source controlgym-env/bin/activate
# Example b: using conda
conda create -n controlgym-env python=3.10
conda activate controlgym-env
# Step 2: Install the controlgym repository
# Example a: using pip
pip3 install -e .
# Example b: using poetry
poetry install
# Step 3: Deactivate the virtual environment (Optional)
# For venv
deactivate
# For conda
conda deactivate
Getting Started
Check out our code examples in this Jupyter notebook file.
Reference
- Zhang, X., Mao, W., Mowlavi, S., Benosman, M., & Başar, T. (2023). Controlgym: Large-Scale Safety-Critical Control Environments for Benchmarking Reinforcement Learning Algorithms. arXiv preprint arXiv:2311.18736.
@article{zhang2023controlgym,
title = {Controlgym: Large-Scale Safety-Critical Control Environments for Benchmarking Reinforcement Learning Algorithms},
author = {Zhang, Xiangyuan and Mao, Weichao and Mowlavi, Saviz and Benosman, Mouhacine and Ba{\c{s}}ar, Tamer},
journal = {arXiv preprint arXiv:2311.18736},
year = {2023}
}
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for controlgym-0.1.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | f0ec6e79878eedb93264d741800ee7243a12c61b86ce9caa3f506d74b966b252 |
|
MD5 | 75a06d28db32b61bc1c47af3b3386e83 |
|
BLAKE2b-256 | 95886292fdcb4022050732dc27149617403b06573793d81abec9b4ec23fb877d |