Q-learning_Concentration

The project consists of a memory game in which the agent, using the Q-learning algorithm, learns how to provide assistance to the user. Once it learns the optimal policy it uses Theory of Mind to provide more targeted and useful assistance.

Project structure

• concentration_1_suggest is the folder which contains a first approach to the problem. In this case the agent can help only on the second flip.
• concentration_2_suggest is the folder in which the agent can suggest before and after the first flip. It's the main folder, which contains the Theory of Mind agent.

Installation instructions

pip install -r requirements.txt

How to train the agent

N.B.: the agent is already trained!

# in the robot folder
py main.py

How to play with agent's suggestions

If you want to play with agent's suggestions you can find a tutorial here: https://drive.google.com/file/d/1irXKeHzRwjW5KeJAQs-9mCeZEr_bzscU/view?usp=sharing

Demo

A demo on how to start the program and how to play is available at this link: https://drive.google.com/file/d/1tr41x4EwkZHiFOsofJ6gxm_XytnNoaQO/view?usp=sharing P.s it's a little be slow because of OBS.

Robot folder

The agent will provide suggestions using the Furhat SDK. You can see the repo here: https://github.com/GiovanniFalcone/Concentration_furhat

UI

credit: https://github.com/yunkii/animal-memory-game