Not that this project is not maintained, but should still be functional
This is a reinforcement Learning based control solution for the Quanser 2DoF Inverted Pendulum.
The implementation mainly focus on Stable Baselines implementation of Proximal Policy Optimization (PPO) and Keras-RL implementation of Double Deep Q Network. It also includes a generic wrapper for OpenAi GYM, Stable Baselines and Keras-RL.
This project was part of the course Design project in Systems, Control and Mechatronics at Chalmers University of Technology, Sweden, during autumn of 2018. An article about the findings from the project is avaliable in the repository: 2Dof_Inverted_Pendulum_RL.pdf
Jonathan Jansson
Bernardo T. Barata
Christian Garcia
Fredrik J. Tornéus
(Apart from OpenAi GYM + stable-baselines/keras-rl and respective dependencies)
- openai
- gym
- stable-baselines
- numpy
- PyQt5
- pyqtgraph