Skip to content

This project focuses on balancing a pole in OpenAI's Gym Cartpole environment using various deep learning techniques. It aims to apply and compare different approaches to solve the balancing challenge effectively.

Notifications You must be signed in to change notification settings

khaledabdrabo98/gym-deep-rl

Repository files navigation

Deep Reinforcement Learning

Important info

  • Dependencies are in requirements.txt
  • Best performances (episodes recordings) are in episodes/
  • Pretrained model weights are in pretrained_models/

How to use it ?

  1. Create new virtual environment
python3 -m venv ./ venv
  1. Install all dependencies
pip3 install -r requirements.txt
  1. Activate the newly created env
source venv/bin/activate
  1. Test demo (with saved pretrained model weights and does not trigger episode recordings)
python3 demo.py
  1. (Extra) Choose which agent you want to test (Notice: these commands will trigger the recording of episodes and will save the newly trained model weights)

    1. Random Agent (episode recordings are in episodes/random_agent)
      python3 random_agent.py
      
    2. Deep Q-Network CartPole agent (episode recordings are in episodes/dqn_agent)
      python3 dqn_agent.py
      
    3. Deep Q-Convolutional Network CartPole agent (episode recordings are in episodes/conv_dqn_agent)
      python3 conv_dqn_agent.py
      
    4. Deep Q-Convolutional Network MineRL agent
      python3 minerl_agent.py
      

Finally, don't forget to deactivate the virtual environment after testing

deactivate

Developed by

  • Khaled ABDRABO p1713323
  • Jean BRIGNONE p1709655

About

This project focuses on balancing a pole in OpenAI's Gym Cartpole environment using various deep learning techniques. It aims to apply and compare different approaches to solve the balancing challenge effectively.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages