sohamghosh121 / pacmangym

Open AI Gym version of Berkeley AI Pacman with images as states

OpenAI Gym Berkeley-Pacman

Port of Berkeley AI Pacman to an OpenAI Gym environment.

The Berkeley Pacman AI engine is provided at http://ai.berkeley.edu/project_overview.html

Info

This repository provides two environments:

• BerkeleyPacman-v0: which provides the Berkeley Pacman AI interface as is, with a randomly chosen layout. Layouts are in layouts/ folder
• BerkeleyPacmanPO-v0: which provides the Berkeley Pacman AI interface with only a partially observable portion of the map (immediate nearby locations in a 3x3 grid)

Usage

To use these environments, do:

import gym
import gym_pacman
env = gym.make('BerkeleyPacman-v0')

Note: This was created for the 10-703 Project at Carnegie Mellon University. This is still under active development

pacmangym's Issues

How to choose between fully observable and partially observable tasks?

DType Error

I just downloaded this file and tried to run it and I am getting this error:

File "/Users/Deydey/Desktop/PacmanGym/gym_pacman/envs/pacman_env.py", line 54, in PacmanEnv
shape=(84, 84, 3), dtype=np.uint8)
TypeError: init() got an unexpected keyword argument 'dtype'

How do you fix this ?
I am trying to use your PacmanGym environment to run a reinforcement learning algorithm.