Env2048

OpenAI Gym environment for the game 2048!

• Free software: 3-clause BSD license

Features

• Any size game board!

• Rendering support with PyGame!

  You can also get an array of pixels with mode='rgb_array' for training image-based agents.

• Different reward types! Set the reward_type parameter on initialization.

  'score' uses the normal score update from the game.

  This is the sum of all tiles created by merging that occurred after an action.

  So if two 8 tiles are merged after an action, then the reward is +16.

  'survival' gives a reward of +1.0 on every timestep unless:

  The action did not produce a change in state; reward is -0.1.

  The game is over; reward is 0.0.

  'milestone' gives a reward of +10.0 every time a new tile is reached.

Installation

To try it out yourself:

$ git clone https://github.com/thomashopkins32/Env2048.git
$ cd Env2048
$ pip install .

Example Usage

Here is an example using random actions:

from env2048.env import Env2048
env = Env2048()
observation = env.reset()
env.render()
done = False
while not done:
     action = env.action_space.sample()
     observation, reward, done, info = env.step(action)
     env.render()