marcsto / rocketlander

Easy way to run OpenAi Gym Rocketlander (aka Falcon Rocket) in your browser

Ever wanted to write code that lands a SpaceX-like rocket on a platform in the ocean? Now you can try it directly from your browser by using the following colab notebook:

https://colab.research.google.com/drive/1gFdJueFbLYBPuI8ss_0WHG0VOXj2SNdT#scrollTo=phVnskT-RKXr

At the time of writing, Google provides a free virtual machine with a K80 GPU to train on.

Credit

This is a copy of https://github.com/EmbersArc/gym that only has the rocket_lander.py file to make it easy to git pull into an existing gym environment.

Local Usage:

If you want to run the environment on your own machine instead of in the colab notebook, follow these steps:

Install openai gym with box2d support

pip install 'gym[box2d]'
pip install box2d box2d-kengz

Clone this repo

git clone https://github.com/marcsto/rocketlander.git

Add the following to your python code to register the environment

from gym.envs.registration import registry, register, make, spec
register(
    id='RocketLander-v0',
    entry_point='rocketlander.rocket_lander:RocketLander',
    max_episode_steps=1000,
    reward_threshold=0,
)

This assumes that this repo (the rocketlander dir) is a sub directory of where your python code is.

i.e.:
      run_python.py
      rocketlander/rocket_lander.py

Please see EmbersArc's excellent repository for more details on the rocket lander gym environment