A simple two-player environment for openai/gym
Go to file
2017-04-03 21:47:10 +02:00
examples Add a usage example 2017-04-01 12:01:16 +02:00
gym_random_walk Random starting state 2017-04-01 10:53:41 +02:00
.gitignore Initial commit 2017-03-31 00:36:55 +02:00
LICENSE Initial commit 2017-03-31 00:36:55 +02:00
README.md Update README.md 2017-04-03 21:47:10 +02:00
setup.py This is necessary for any actual installation taking place 2017-03-31 00:48:43 +02:00

gym-random-walk

A minimal example of a custom environment for https://github.com/openai/gym.

What the environment provides is not that important; this is meant to show how what you need to do to create your own environments for openai/gym.

For concreteness I used an example in the recordings of David Silver's lectures on Reinforcement Learning at UCL.

(0) - A - B - C - D - E - (+1)

You start off at one of the positions A to E, you can move right or left, reaching the "+1" terminal state gives you a reward of +1, and going all the way "to the left" will give you a terminal reward of 0.

Instead of calling them the above, I just made them the states 0, 1, ...6.