Skip to content

duguyue100/minesweeper

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

54 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MineSweeper

PyPI PyPI

Build Status Build status Codacy Badge license

A python Minesweeper with interfaces for Reinforcement Learning.

This is a simple game I wrote for learning Deep Reinforcement Learning.

How to Install

In order to install this package, you need to have at least

  • numpy
  • pyqt

installed. If you don't want to bother with details of packages installation, you can use Anaconda as your Python distribution.

And then install the package by

From PyPI (the latest stable version)

pip install minesweeper

or

From GitHub (the bleeding edge version)

pip install git+git://github.com/duguyue100/minesweeper.git \
-r https://github.com/duguyue100/minesweeper/blob/master/requirements.txt

Once you installed, you can start the Mine Sweeper GUI by typing following under terminal

ms-gui.py # yeah, with .py, otherwise Windows CMD may not be able to find it!

Objectives

  • A command line mine sweeper game.
  • A GUI interface.
  • Interfaces for receiving move info and send board status.
  • Capability of playing multiple games using different UDP ports.

Todo List

  • Basic structure setup
  • update rules for msboard (click)
  • update rules for msboard (flag)
  • update rules for msboard (unflag)
  • update rules for msboard (question)
  • TCP interface for the game
  • console interface for the game
  • GUI interface for the game
  • Complete reset button and end game condition
  • Control GUI interface through remote commands (another thread?)
  • Write Tests, collect user feedback (if possible)
  • Write more TCP connection related functions (e.g. dump array)
  • Design, train and test a DQN controller (learning on the way :)).
  • Demo for the trained controller.
  • Handling arbitrary size of board.
  • Play game only with pixel inputs?

Screen Shot

Game Shot

The GUI Player

usage: ms-gui.py [-h] [--board-width BOARD_WIDTH]
                 [--board-height BOARD_HEIGHT] [--num-mines NUM_MINES]
                 [--port PORT] [--ip-add IP_ADD]

Mine Sweeper Minesweeper with interfaces for Reinforcement Learning by Yuhuang
Hu

optional arguments:
  -h, --help            show this help message and exit
  --board-width BOARD_WIDTH
                        width of the board.
  --board-height BOARD_HEIGHT
                        height of the board.
  --num-mines NUM_MINES
                        number of mines.
  --port PORT           The port for TCP connection.
  --ip-add IP_ADD       The IP address for TCP connection.

The default value of the game is

board_width = 20
board_height = 20
num_mines = 40
port = 5678
ip_add = "127.0.0.1"

Control The Game

As shown in test-board.py, you can control the game through the APIs. This provides a way of accessing and generating large number of games by designing your own flow.

Alternatively, you can also access and control a game through TCP connections. The examples is given in test-tcp-com.py. The TCP connection is initiate by given IP address and port number. You can easily connect the connection using netcat (for Mac OS X, nc), e.g.

nc 127.0.0.1 5678

Once you connected, type help for viewing command line.

Welcome to Mine Sweeper!
You have 5 types of moves to use:
(1) Click	: click: X, Y
(2) Flag	: flag: X, Y
(3) Question	: question: X, Y
(4) Unflag	: unflag: X, Y
(5) Print board: print

Of you can still use other programs for controlling the game. Please check related APIs from msgame.py

Contacts

Yuhuang Hu
Email: duguyue100@gmail.com

About

A python Minesweeper with interfaces for RL.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages