koulanurag / maze-world

Random maze environments with different size and complexity for reinforcement learning research.

maze-world

Random maze environments with different size and complexity for reinforcement learning and planning research. This is in-particular to investigate generalization and planning ability in dynamically changing environment.

Disclaimer: This project is largely a amalgam of references mentioned here.

Installation

• Using PyPI:

  pip install maze-world
  

• Directly from source (recommended):

  git clone https://github.com/koulanurag/maze-world.git
  cd maze-world
  pip install -e .

Environments Zoo!

RandomMaze-11x11-v0	RandomMaze-21x21-v0	RandomMaze-31x31-v0	RandomMaze-101x101-v0

See all here.

Quick-Start:

import gymnasium as gym

env = gym.make("maze_world:RandomMaze-11x11-v0", render_mode="human")
terminated, truncated = False, False
observation, info = env.reset(seed=0, options={})
episode_score = 0.

while not (terminated or truncated):
    action = env.action_space.sample()
    observation, reward, terminated, truncated, info = env.step(action)
    episode_score += reward

env.close()

See entire quick-start guide here.

Testing:

• Install: pip install -e ".[test]"
• Run: pytest -v --xdoc --cov=./ --cov-report=xml

Development:

If you would like to develop it further; begin by installing following:

pip install -e ".[develop]"

References:

1. Gym-Maze
2. Mazelab
3. Custom Gym environment based out of gymnasium
4. Wilson Maze Generator