Scribd
Upload
73 views

Delhi Technological University Department of Computer Science

1) The document proposes creating an AI game player for the 2048 game using two algorithms: Minimax and Monte Carlo Tree Search. 2) The motivation is that 2048 is a challenging game for humans, and the project aims to create an AI that can play with a high success rate. 3) The AI will be evaluated by testing it over 1000 iterations with randomly generated boards, and the probability of the AI winning will be used as the success metric to compare the two algorithms.

Uploaded by

Nameet Jain
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
73 views

Delhi Technological University Department of Computer Science

1) The document proposes creating an AI game player for the 2048 game using two algorithms: Minimax and Monte Carlo Tree Search. 2) The motivation is that 2048 is a challenging game for humans, and the project aims to create an AI that can play with a high success rate. 3) The AI will be evaluated by testing it over 1000 iterations with randomly generated boards, and the probability of the AI winning will be used as the success metric to compare the two algorithms.

Uploaded by

Nameet Jain
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

DELHI TECHNOLOGICAL UNIVERSITY

DEPARTMENT OF COMPUTER SCIENCE

CO 304: Artificial Intelligence


Project proposal

Submitted To: Submitted By:


Mr. Anurag Goel Nameet Kumar Jain
2K18/CO/221
Pranjal Choudhary
2K18/CO/253
1

Overview
Here we propose to create an AI game player for the famous game 2048 using two
different Artificial Intelligence Algorithms Minimax and Monte Carlo Tree Search and
compare success percentage in both cases.

Motivation
1. 2048 Game is a very interesting yet not easy game for humans. There are thousands
of tutorials over the web to improve game play.
2. The complexity associated with the gameplay and a very low success chance
motivated us to create an AI which could play the game with a high success rate.
3. We also are not sure on which algorithm should be employed to create this AI. Thus,
we will be creating our AI gameplayer using two different algorithms and compare
performance in both cases.

Proposed Models/Techniques
Algorithms/Techniques Used

To create this AI we will be using the Monte Carlo Tree search method. Monte Carlo
Method is the idea of using a large number of random simulations of an experiment to
gain insights into the experiment’s end results. Random Simulations of an experiment are
frequently referred to as Monte Carlo simulations.

Also, in contrast with Monte Carlo we will also be using the Minimax algorithm to see
which one performs better on our selected scenarios. Minimax is a kind of backtracking
algorithm that is used in decision making and game theory to find the optimal move for
a player. Two players that are used are called minimizer(tries to get the lowest score
possible) and maximizer(tries to get the highest score possible).
2

Libraries Used

● Random
● Sys
● OS
● Tkinter
● Numpy

Using the above prerequisites we will be able to create a functional AI along with the Game
UI , in order to beat the 2048 game, which will be able to predict and beat the game in the
fastest way possible.

Success Evaluation Metric


1. For evaluating the AI, we will be testing it on around 1000 different iterations.
2. In each iteration the board will be generated randomly and we will check the AI’s
ability to perform.
3. The probability of AI winning the game will be considered as the Evaluation metric.
3

Deliverables
A high functioning AI which can perform better than humans can perform, also takes less
time in order to beat the 2048 game will be delivered. Also we will be able to compare the
two algorithms, i.e. Minimax and the Monte carlo method and we will be able to tell
which performs better.

You might also like