florianvazelle / unity-rl

Markov Decision Process and Temporal Difference algorithms

Unity RL

Unity RL is a Unity C# application with multiple implementations of reinforcement learning algorithms on three mini-games.

Setup

This project work on Unity 2019.4.14f1

Table of Contents

Algorithms

• Markov Decision Process
  • Policy Iteration
  • Value Iteration
• Monte Carlo
  • Exploring Starts
  • On Policy (first/evry visit)
  • Off Policy (first/evry visit)
• Temporal Difference
  • SARSA
  • QLearning

Games

• Grid World
• Tic Tac Toe
• Sokoban

Demos