search

dfki-ric-underactuated-lab / double_pendulum

Dual purpose Acrobot and Pendubot Platform
BSD 3-Clause "New" or "Revised" License
33 stars 22 forks source link

DQN #6

Closed theovincent closed 1 year ago

theovincent commented 1 year ago

Hi,

I would like to add the possibility of learning a DQN agent to the repo. I have added:

Feel free to tell me if anything is missing :)

fwiebe commented 1 year ago

Hi @theovincent thanks a lot for your controller contribution!

For the merge request, I have the following remarks:

  1. src/python/setup.py Could you also add the added dependencies in the extras "RL" to the "all" section?
  2. src/python/double_pendulum/controller/DQN/__init__.py this __init__.py file is missing
  3. I would suggest to store the policy and the parameters at data/policies/design_A.0/model_2.0/pendubot/DQN/
  4. examples/reinforcement_learning/DQN/train_dqn.py I think there is a "sys.argv" missing as argument for the train function when this script is executed.
theovincent commented 1 year ago

Hi @fwiebe,

Thank you for reviewing my PR. I have done the points you asked for. Please tell me if anything else is missing :)

fwiebe commented 1 year ago

Thanks @theovincent I merged your controller. The controller is now listed on our simulation leaderboards here (simulation) and here (robustness) :)