PacmanEnv

Modified Pac-man environment for dynamic navigation task

Quickstart

Run following command on kernel with default parameters below

• Pacman Agent : None -> keyboard Input
• Ghost Agent : Random movement ghost
• Layout : 16 x 16 open space with three reward and one ghost

$ python3 pacman.py

Environment Structure

Game

All rules, valid movements, score information are store in three files.

• game.py : rules
• pacman.py : main function
• utils.py : helper functions

Agents

There are two agent classes in the environment all inherited from the Agent(game.py) class.

GhostAgent : All ghost agent is defind in the ghostAgents.py file and must inherit GhoastAgent(ghostAgent.py) class.

Ghost agent is indexed by the class name and this name can be used by inserting '-g' option upon calling the pacman.py. (ex. python3 pacman.py -g PredatorGhost)

PacmanAgent : Pacman agent is defind in the pacmanAgents.py file.

Default agent is KeyboardAgent were it receives user input for movement selection.

See pacmanDQN_Agents.py for self-behaving agent example.

Configuration

Configuration holds the agent's coordinate and moving direction. This value can be obstained by two getter functions.

• getPosition()
• getDirection()

AgentState

Hold the state of an agent (configuration)

• getPosition: Get agent position from Configuration
• getDirection: Get agent travelling direction from Configuration

Grid

A 2D nested array object. Return positions of Pacman map as [x][y]. x is horizontal and y is vertical.

GameStateData

Get new state data from previous event.

Game

Manage all aspects of game play

• run : Start the main loop of game. Initialize the environment, check time, solicit action form agent, and result of the game.

Layout

Layout file is a plane text file located in in the ./layouts/. All layout file should have the .lay file extension. Following text is interpreted to corresponding component of the environment, and other characters are converted into an empty space. For the visual convention, we used underbar for the empty space.

• . : reward
• % : wall
• G : ghost
• P : pacman
• _ : empty

Graphics

Define graphic components of the environment

Citation

This repository is folked from tychovdo/PacmanDQN and original work is done by UC Berkeley AI class project (http://ai.berkeley.edu).

Citation below is from the forked repo.

@article{van2016deep,
  title={Deep Reinforcement Learning in Pac-man},
  subtitle={Bachelor Thesis},
  author={van der Ouderaa, Tycho},
  year={2016},
  school={University of Amsterdam},
  type={Bachelor Thesis},
  pdf={https://esc.fnwi.uva.nl/thesis/centraal/files/f323981448.pdf},
}

• van der Ouderaa, Tycho (2016). Deep Reinforcement Learning in Pac-man.

Acknowledgements

DQN Framework by (made for ATARI / Arcade Learning Environment)

• deepQN_tensorflow (https://github.com/mrkulk/deepQN_tensorflow)

Pac-man implementation by UC Berkeley:

• The Pac-man Projects - UC Berkeley (http://ai.berkeley.edu/project_overview.html)