This is a wrapper to use ViZDoom, a "Doom based AI Research Platform for Reinforcement Learning from Raw Visual Information" together with OpenAI Gym.

There is a branch with an alternative reward system for the Health Gathering scenario (each collected health pack yields +1 reward)

In the docker branch, rendering is excluded, because there must not be a display in docker containers.

sudo apt-get install cmake libboost-all-dev libgtk2.0-dev libsdl2-dev python-numpy
git clone https://github.com/simontudo/vizdoomgym.git
cd vizdoomgym
pip install -e .

Use one of the environments (see list below for all available envs):

import gym
import vizdoomgym
env = gym.make('VizdoomBasic-v0')

List of available environments:

VizdoomBasic-v0
VizdoomCorridor-v0
VizdoomDefendCenter-v0
VizdoomDefendLine-v0
VizdoomHealthGathering-v0
VizdoomMyWayHome-v0
VizdoomPredictPosition-v0
VizdoomTakeCover-v0
VizdoomDeathmatch-v0
VizdoomHealthGatheringSupreme-v0

Detailed information about the environments