MinigridSafe

A fork of Farama's Minigrid extended for safe reinforcement learning research.

Extensions

• Dual Action Modes: Omnidirectional (rotate + forward) and Cardinal (absolute N/S/E/W directions)
• Slippery Tiles: Directional and generic surfaces with probabilistic movement
• Adversaries: NPC agents with configurable behaviors that interact with the player
• PRISM Export: Convert environments to PRISM model format for formal verification
• ASCII Environments: Build custom environments from ASCII layouts

Installation

pip install -e ".[testing]"

Upstream Attribution

Built on top of Minigrid by the Farama Foundation.

@inproceedings{MinigridMiniworld23,
  author       = {Maxime Chevalier{-}Boisvert and Bolun Dai and Mark Towers and Rodrigo Perez{-}Vicente and Lucas Willems and Salem Lahlou and Suman Pal and Pablo Samuel Castro and Jordan Terry},
  title        = {Minigrid {\&} Miniworld: Modular {\&} Customizable Reinforcement Learning Environments for Goal-Oriented Tasks},
  booktitle    = {Advances in Neural Information Processing Systems 36, New Orleans, LA, USA},
  month        = {December},
  year         = {2023},
}