Creating an Aftificial Intelligence for a Strategy Game

OData support
Dr. Juhász Sándor
Department of Automation and Applied Informatics

The popular strategy games played by humans usually have simple rules however finding winning strategies is a complex task. Creating computer algorithms that can play against humans is really hard especially when it is an incomplete information game where the usage of brute force methods are impossible due to the huge size and complexity of the event space. While is it important to have a decent artificial intelligence in these games most of the time their behavior is quite simple and it's performance lacks which often compensated with "cheats".

The goal of this thesis is to learn about the game "Starcraft: Brood War" and the tools which makes it possible to create artificial intelligence and to create a robot which is capable of playing a whole one on one game on its own.

In this thesis I am examining the artificial intelligence creating tools, analyzing the game and identifying the subtasks then making algorithms to solve them, measuring these algorithms and comparing them to other solutions. Finally I am creating a working robot from these parts and measure its performance against the built-in artificial intelligence.

My robot is better than the built-in artificial intelligence in every examined aspects and is easily extensible with new behaviors thanks to its architecture.


Please sign in to download the files of this thesis.