jarvisisfriday / gym_pomdp Goto Github PK

This package is an extensions of OpenAI Gym, for Partially Observable Markov Decision Process.

• Python
• OpenAI Gym
• PyGame

Check out the latest code git clone https://github.com/d3sm0/gym_pomdp.git

To use it as a package simply run: python setup.py install

Import the library and gym as and call the environment:

import gym
import gym_pomdp
env = gym.make("Tag-v0")

All environments are parametrized as in the original papers. In order to get larger state space or more enemies, it's easy to change the board_size in the specific environment.

• Tag
• Tiger
• BattleShip
• Network
• RockSample

General overview POMCP solver Point-based value iteration Similar work

David Silver and Joel Veness made this possible by releasing the code POMCP open source. And @manuel for proof test.

• Add pocman environment [WIP]