RlControl

Designing Controllers via Deep Reinforcement Learning

📝 About The Project

The project aims to control linear/nonlinear dynamical systems in model-free and model-based setting with SOTA deep reinforcement learning algorithms

Try App Live with Streamlit!

Installation Steps

Follow steps below,

1. Create Conda environment:
    - 'conda env create --name rlcontrol --file=environment.yml'

2. Install OpenAI Gym Environment
    - Install gym_control environment: `pip install -e .`

- Export conda env yaml: `conda env export > environment.yml --no-builds`
- Export conda env yaml as requirements.txt: `conda list -e > requirements.txt`

Run App in Local

1. Activate rlcontrol environment
  `conda activate rlcontrol`
2. Run streamlit app and watch app at `http://localhost:8502`
  `streamlit run app.py` 

Build for docker based deployment

• docker build -t streamlit .

🍴 Prerequisites

The following open source packages are used in this project:

• streamlit
• torch
• ray
• rllib
• control
• gymnasium
• tensorboardx
• tqdm
• numpy
• matplotlib
• plotly
• pandas
• pickle5
• zipfile36

Future Work

• Track experiments from db (postgresql)
• Fix seeds