Zeldagen is a procedural Zelda-like dungeon generation project, utilizing genetic algorithms. The name "Zeldagen" combines "gen" from both generative and genetic.
It was built as a project for the course "Fundamentals of Artificial Intelligence" of the Computer Science faculty, Università degli Studi di Salerno.
The idea behind Zeldagen stems from the challenge of creating high-quality gameplay experiences like those found in The Legend of Zelda series. These games have captivated millions due to their well-balanced and rewarding dungeon designs. However, achieving such balance and challenge in procedural generation is difficult without human design.
In most games that use procedural generation, there is a compromise between variety vs. quality. To provide fresh and non-repetitive experiences, the generated content often sacrifices quality. Zeldagen aims to solve this issue by generating dungeons that are not only diverse but also satisfying for players.
The goal of Zeldagen is to generate Zelda-like dungeon topologies that are procedurally created yet still feel rewarding and challenging for human players. The goal is to produce them quickly, without compromising quality.
To achieve this, we decided to use an intelligent agent in the form of a genetic algorithm. Genetic algorithms are ideal for optimization, allowing for iterative improvements and refinements to the generated dungeons.
More information about Zeldagen and how it works can be found in the documentation, in the docs/zeldagen.pdf file, in italian language.
The src folder contains the executable source code for three things:
- the genetic algorithm itself (executable through main.py);
- the benchmarks for the genetic algorithm (executable through benchmark.py);
- the demo (executable through demo.py).
The constants.py file contains different constants used throghout the algorithm to easily change them, such as the number of rooms of a dungeon and the room types enum.
The dungeon.py file is part of the genetic algorithm and represents an individual, containing the code for initialization, fitness calculation and printing the graph.
The demo folder also contains a demo_styles.tcss, used by the Textual Python library to apply style to the demo TUI application.
The docs folder contains the documentation report as a PDF file zeldagen.pdf, with its relative LaTeX document zeldagen.tex.
There's also the bibliography.bib file which is used by BibTeX to create the bibliography paragraph.
The docs/assets folder contains multiple images used in the documentation report.
To run the project, you'll have to install Python 3.10.6 and Graphviz 12.2.1, then install all packages in requirements.txt through pip.
If you only want to run the genetic algorithm or the benchmarks, you don't need to install anything else and can simply run them through the virtual environment with the Python interpreter.
If you want to run the demo, you'll have to make sure that your machine can handle Ollama flawlessly.
If this is the case, proceed to install Ollama. Then, you can simply run the demo's terminal user interface by running the demo.py file with your Python interpreter, after activating the virtual environment.
Note: you'll have to run the file from inside the src/demo folder. The working directory path is important.
If you have issues running the project or want more informations, you can send me a mail at [email protected], or open an issue in this repository.
This project is licensed under the GNU License - see the LICENSE file for details.