This repository contains the code for the project "The Legend of Zelda: Reinforcement Learning", it's purpose is to train an agent to play the game "The Legend of Zelda: Link's Awakening" for GameBoy using Reinforcement Learning.

Clone the repository

git clone https://github.com/msosav/zelda-reinforcement-learning
cd zelda-reinforcement-learning

Create a virtual environment and activate it

python3.11 -m venv venv
source venv/bin/activate

Install pytorch

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

Install the requirements

pip install -r requirements.txt

The program is divided into two main parts: the training and the testing.

To train the agent, run the following command:

python main.py train

To test the agent, run the following command:

python main.py test ./checkpoints/<checkpoint>.zip

Example:

python main.py test ./checkpoints/best_model_5000.zip

The reward system is based on the following rules:

• +1 for each item in the inventory