OpenAI gym Gridworlds

Implementation of three gridworlds environments from book Reinforcement Learning: An Introduction <http://incompleteideas.net/book/the-book-2nd.html>_ compatible with OpenAI gym <https://github.com/openai/gym>_.

Usage

.. code::

    $ import gym
    $ import gym_gridworlds
    $ env = gym.make('Gridworld-v0')  # substitute environment's name

Gridworld-v0

Gridworld is simple 4 times 4 gridworld from example 4.1 in the [book]. There are four action in each state (up, down, right, left) which deterministically cause the corresponding state transitions but actions that would take an agent of the grid leave a state unchanged. The reward is -1 for all tranistion until the terminal state is reached. The terminal state is in top left and bottom right coners.

WindyGridworld-v0

Windy gridworld is from example 6.5 in the book_. Windy gridworld is a standard gridworld as described above but there is a crosswind upward through the middle of the grid. Action are standard but in the middle region the resultant states are shifted upward by a wind which strength varies between columns.

.. _book: http://incompleteideas.net/book/the-book-2nd.html

Cliff-v0

Cliff walking is a gridworld example 6.6 from the book_. Again reward is -1 on all transition except those into region that is cliff. Stepping into this region incurs a reward of -100 and sends the agent instantly back to the start.