What is this?

This is a personal project to optimize an algorithm for playing the game 2048. So far, I've implemented a Monte Carlo search algorithm that utilizes the GPU of the user's computer to parallelize the simulations and thus run much faster.

Directory Structure

• 2048AI/game_files/MonteCarloAI.py: Contains the main algorithm for the Monte Carlo search.
• 2048AI/game_files/modules: Contains modules for running the aforementioned algorithm.
• 2048AI/game_files/original: Contains original files from the course, and files for profiling the performance of an algorithm that doesn't utilize the GPU.

Requirements

• Nvidia GPU and CUDA toolkit installed.
• Python 3.x

Necessary Packages

To run this project, you need to install the following Python packages:

• pycuda: Access to Nvidia's CUDA parallel computation API from within Python.
• numpy: Package for scientific computing with Python.
• pygame: Set of Python modules designed for writing video games.

You can install the necessary Python packages using pip:

pip install pycuda numpy pygame

Make sure the version of PyCUDA matches the version of your CUDA toolkit.

Credit

The code for the game itself Game2048.py and for playing it manually Play_Game2048.py was developed by DTU lecturers for the course "Introduction to Intelligent Systems" 02461, and has only been slightly modified by me. The rest is developed by me.