Skip to content

MattChanTK/gym-maze

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

23 Commits
 
 
 
 
 
 
 
 

Repository files navigation

gym-maze

A simple 2D maze environment where an agent (blue dot) finds its way from the top left corner (blue square) to the goal at the bottom right corner (red square). The objective is to find the shortest path from the start to the goal.

Simple 2D maze environment

Action space

The agent may only choose to go up, down, right, or left ("N", "S", "E", "W"). If the way is blocked, it will remain at the same the location.

Observation space

The observation space is the (x, y) coordinate of the agent. The top left cell is (0, 0).

Reward

A reward of 1 is given when the agent reaches the goal. For every step in the maze, the agent recieves a reward of -0.1/(number of cells).

End condition

The maze is reset when the agent reaches the goal.

Maze Versions

Pre-generated mazes

  • 3 cells x 3 cells: MazeEnvSample3x3
  • 5 cells x 5 cells: MazeEnvSample5x5
  • 10 cells x 10 cells: MazeEnvSample10x10
  • 100 cells x 100 cells: MazeEnvSample100x100

Randomly generated mazes (same maze every epoch)

  • 3 cells x 3 cells: MazeEnvRandom3x3
  • 5 cells x 5 cells: MazeEnvRandom5x5
  • 10 cells x 10 cells: MazeEnvRandom10x10
  • 100 cells x 100 cells: MazeEnvRandom100x100

Randomly generated mazes with portals and loops

With loops, it means that there will be more than one possible path. The agent can also teleport from a portal to another portal of the same colour.

  • 10 cells x 10 cells: MazeEnvRandom10x10Plus
  • 20 cells x 20 cells: MazeEnvRandom20x20Plus
  • 30 cells x 30 cells: MazeEnvRandom30x30Plus

Installation

It should work on both Python 2.7+ and 3.4+. It requires pygame and numpy.

cd gym-maze
python setup.py install

Examples

An example of finding the shortest path through the maze using Q-learning can be found here: https://github.com/tuzzer/ai-gym/blob/master/maze_2d/maze_2d_q_learning.py

Solving 20x20 maze with loops and portals using Q-Learning

About

A basic 2D maze environment where an agent start from the top left corner and try to find its way to the bottom left corner.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages