This project investigates multiple automation techniques for level design through the implementation of various prototypes. The types of techniques are influenced by art, narrative structure and other non-game forms of media, as well as taking inspiration from game related level design theory and academic research. The project finds that these techniques, when used to structure levels, are a powerful tool for learning and searched based level automation.
Prior to developing Prototypes in Unity I developed console based applications to test ideas researched in the early stage of this project. This provided a fast way to understand and visualise concepts befofe managing assets in the Unity environment.
The Sentence Generator aims to investigate how parse tree can be used for level automation. This prototype proved that parse tree's could be successfully used to procedurally generate a sentence where each string has the potencial to replaced with a more complex data type.
The Markov Number Generator was developed when first learning about Markov Chain's. The intention of this prototype was to help visualise how the underlying system would function in a Unity 3D environment to help with the development of the Markov Chain Tile-Based Level Generator.
This repository contains the Unity Project in which all prototypes were developed. These projects are discussed and showcased in the Final Video and include investigations into the techniques below.
This prototype aims to automate the level design process by using the parse tree technique and the narrative structure The Hero’s Journey.
This prototype looks at level design automation for exploration based game genres such as Rogue-likes along with world overviews and layouts such as the New Mario Bros (Nintendo EAD) world lobby view.
Song structure typically used in Pop-Music, takes advantage of a Bridge to break the pattern of a song. Likewise, Iambic Pentameter, found in poetic writing, uses off beats to break the pattern. Breaking patterns is a technique currently adopted in Game Level Design to create memorable experiences. The Rhythm Based Platformer prototype was implemented to test this research.
Inspired by Piet Mondrian’s piece, Composition II in Red, Blue and Yellow and the main issues with the Markov Chain Tile-Based Level Generator prototype. This prototype aims to learn the individual features in a level instead of individual tiles.
This prototype aims to automate the level design process by learning from levels in which each
This prototype uses a Markov Chain to learn the probability between tiles in pre-made, 3D levels. As seen in the video linked below these learned probabilities are used to generate new permutations of the learned level.
- This YouTube Video explains the technique used to automate the generation of a platformer leveland learns from more than one level.