Map generation with genetic algorithm

OData support
Rajacsics Tamás
Department of Automation and Applied Informatics

One of tasks of game development, which have the greatest impact on sales of computer games, is the diverse building of game space. This task is both a design and artistic challenge, and it may take a lot of time and money.

In my thesis I review the level making methods tried in the game industry so far, I analyze their properties, and I collect the tools which can make the task partly automated.

I write about genetic algorithm in detail, touch on its origins, I talk about its structure, how it works and where it is used. I compare its properties with other algorithms, mostly the ones int the previous chapter.

In the second part of my thesis I examine how the genetic algorithm can be used for the purpose of level generation, and how the adaptable and exchangeable parts of the algorithm can be linked to the concepts of level generation. I will give a detailed explanation of the strictly implementation dependent parts.

I present the software I made. This program is able to create levels according to expectations, it works as discussed int the previous chapters and does not need human input during the generation of a level. Additionally I explain my design choices and decisions, I talk about the challenges of implementation, problems that emerged during work and their solutions.

In the last part I analyze the results of testing the algorithm with a self-made test program. I demonstrate the effects of design choices and parameter settings on memory usage and time consumption.

I survey the possibilities of the expansion of the software, the necessity and possible methods of optimization, and how it can used in the future both in real-time and during game design.


Please sign in to download the files of this thesis.