A simple project showcasing the use of Q-learning for balancing a double pendulum.
This repository contains the code necessary to run a Q-Learning algorithm ruling the behaviour of a pendulum. It focuses on the adaptation of the work in the paper Human-level control through deep reinforcement learning to a simpler environment, along with many suggestion from Implementing the Deep Q-Network for implementing the details. The resulting deep neural network is able to accurately estimate the Q-function of the pendulum along with a greedy policy which together allow it to produce episodes with low cost, managing to balance the pendulum starting from any state.
- Python3
- Tensorflow
- Keras
- Numpy
- Orca
It is sufficient to execute the DQN.py script
This project is licensed under the MIT License - see the LICENSE.md file for details