Skip to content

A custom open ai gym environment for solo experimentation.

License

Notifications You must be signed in to change notification settings

WPI-MMR/gym_solo

Repository files navigation

Solo Gym
WPI Logo
Build Status Coverage Status

A custom open ai gym environment for Solo experimentation.

screenshot


TL;DR

  • An OpenAI Gym-conformant environment featuring ODRI's Solo 8
  • OpenAI Gym API offers out-of-the box compatibility with most RL frameworks
  • Fully encapsulated and multiprocessing-ready
  • Easily extendable environments, rewards, and observations
  • Automatically normalize observation and action spaces
  • Realtime simulations supported for non-discrete timesteps

Note that this project is a heavy work-in-progress. While we may not be able to work on your issue, we appreciate any feedback. Additionally, if you have any interesting ideas that you'd like to implement, we'll happily take contributions!

🌟 We appreciate your star, it helps!

Installation

The recommended way to set up this environment is to use virtualenv's for its sandboxing. With that:

  1. Clone and navigate to this repository from your terminal
  2. python -m venv venv-gym-solo
  3. source venv-gym-solo/bin/activate
  4. pip install --upgrade wheel pip
  5. pip install -e .

And the package should be installed! Note that this is in development mode so any local changes will be reflected in the package as well.

Running Tests

All of our tests can be run by using python's built in unittest. However,due to how PyBullet works, we require some extra dependencies. To run the tests, use the following procedure:

  1. Install xvfb (on Ubuntu, you can do sudo apt install xvfb)
  2. Follow the instructions above for creating the virtual environment, but use pip install -e .[test] instead of pip install -e .
  3. Navigate to the root of the repository
  4. python -m unittest discover -v .

Releases

No releases published

Languages