This repository showcases the use of the pyflyt
module's rocket landing environment and various reinforcement learning (RL) algorithms to solve it. The TQC (Truncated Quantile Critics) algorithm from Stable Baselines has shown the best performance.
- Introduction
- Installation
- Usage
- Simulation & Inference with Trained Model
- Screenshots
- Video Demonstration
- Training Metrics
- References
This project demonstrates the application of RL algorithms in the rocket landing environment provided by pyflyt
. The environment is a challenging testbed for algorithms to learn and perfect the landing of a rocket.
- Python 3.8+
- Poetry
-
Clone the repository
git clone https://github.com/artemi8/RL-Rocket-Landing.git cd RL-Rocket-Landing
-
Install Poetry
Follow the instructions from the official Poetry documentation.
-
Install dependencies
poetry install
To use the environment and run your experiments, follow the steps below.
To test the best performing TQC model, follow the instructions provided in the Jupyter notebook located at sample_runs/rocket_landing_experiments-testing.ipynb
.
-
Launch Jupyter Notebook
poetry shell
jupyter notebook
-
Open and run the notebook
Navigate to
sample_runs/rocket_landing_experiments-testing.ipynb
in the Jupyter interface and follow the instructions to run the inference with the trained TQC model.
Watch the rocket landing demonstration:
TQC_rocket_landing.mp4
Go to training metrics.