/Goal-conditioned-Reinforcement-Learning-with-environmental-and-policy-priors

A novel Goal-conditioned Reinforcement Learning Algorithm with environmental and policy priors

Primary LanguagePython

Requirements:

  • Python 3.7 to 3.10
  • OpenAI Gym v0.22 to v0.25
  • NumPy 1.18+
  • Matplotlib (optional, only needed for display) - 3.0+

Environment Reference File

Chevalier-Boisvert, Maxime, Lucas Willems, and Suman Pal. "Minimalistic gridworld environment for openai gym." (2018): 13-20.

Installation

After installing requirement:

cd gym-minigrid_1216
pip3 install -e .

then, you can control the agent by keyboard running "manual_control.py"

Environment Version

#hard environment register( id="MiniGrid-twoarmy-17x17-v4", entry_point="gym_minigrid.envs:Twoarmy_v4", kwargs={"size": 17}, )

#easy environment register( id="MiniGrid-twoarmy-17x17-v6", entry_point="gym_minigrid.envs:Twoarmy_v6", kwargs={"size": 17}, )