2048 AI

This repository contains a collection of AI agents designed to play the game 2048 using various strategies, including Monte Carlo methods and reinforcement learning techniques. The project is intended to serve as an open-source resource for those interested in studying and developing AI for complex decision-making problems.

Table of Contents

• Introduction
• Installation
• Usage
• Agents
  • Monte Carlo Agent
  • Enhanced Monte Carlo Agent
  • Advanced Monte Carlo Agent
  • QLearning Agent
• Contributing
• License

Introduction

2048 is a popular single-player sliding block puzzle game. The objective is to slide numbered tiles on a grid to combine them and create a tile with the number 2048. This project implements several AI agents that play the game, using different techniques to make decisions and optimize their gameplay.

Installation

To get started with this project, clone the repository and install the required dependencies:

git clone https://github.com/AlexFigas/2048-ai.git
cd 2048-ai
pip install -r requirements.txt

Usage

You can run any of the AI agents to see how they perform in playing 2048. Each agent has a main script that can be executed directly. For example, to run the AdvancedMonteCarloAgent:

python src/AdvancedMonteCarloAgent.py

Example Output

As the AI agent plays the game, it will display the current state of the board and the chosen moves in the console and in a GUI (in the case of the AdvancedMonteCarloAgent.py).

Best move: up
Final board:
[[   4    2    0    0]
 [   8    4    2    0]
 [  32   16    8    4]
 [ 256  128   64   32]]

Agents

Monte Carlo Agent

The MonteCarloAgent uses a basic Monte Carlo simulation to evaluate potential moves. It performs a series of random simulations to estimate the best possible move based on the resulting scores.

• File: src/MonteCarloAgent.py
• Strategy: Random simulations

Enhanced Monte Carlo Agent

The EnhancedMonteCarloAgent further enhances the Monte Carlo method with a simplified heuristic scoring system, balancing exploration and exploitation to find better moves.

• File: src/EnhancedMonteCarloAgent.py
• Strategy: Monte Carlo with simplified heuristics

Advanced Monte Carlo Agent

The AdvancedMonteCarloAgent builds upon the enhanced Monte Carlo approach by incorporating heuristics like board smoothness, monotonicity, and weighted tiles to guide its decision-making process.

• File: src/AdvancedMonteCarloAgent.py
• Strategy: Monte Carlo with advanced heuristics

QLearning Agent

The QLearningAgent applies reinforcement learning, specifically Q-learning, to develop a strategy by learning from previous moves and their outcomes.

• File: src/QLearningAgent.py
• Strategy: Q-learning (Reinforcement Learning)

Results

In the end I ran a 500 simulation testing with the AdvancedMonteCarloAgent which I believed it was the best approach, took over 8 hours to complete and here are the results:

Contributing

Contributions are welcome! If you'd like to contribute to this project, please follow these steps:

• Fork the repository
• Create a new branch (git checkout -b feature-branch)
• Commit your changes (git commit -m 'Add new feature')
• Push to the branch (git push origin feature-branch)
• Open a Pull Request

License

This project is licensed under the GNU General Public License v3.0 License - see the LICENSE file for details.

TODO List:

• Create a Agent interface to serve to all agent implementations
• Refactor code to have better organization and readability
• Improve weighted tiles utilization in the heuristic function of AdvancedMonteCarloAgent
• Create new agents