/rl-gridworld

Gridworld sandbox for reinforcement learning

Primary LanguagePython

RL-GridWorld

Graphical gridworld sandbox for reinforcement learning. Uses Q-learning to learn utility function and greedy policy w.r.t. learned utility. The GUI is implemented using PyQt5.

Run with:

python3 main.py

Features

  • Dynamic world size
  • Customizable learning rate, discount factor, rewards
  • Support for episodic and non-episodic settings
  • Different policies
    • Random
    • Greedy w.r.t. learned utility
    • ε-Greedy

Screenshot

Dependencies

  • PyQt5
  • NumPy