Tile Embedding: A General Representation for Procedural Level Generation via Machine Learning

TL;DR

This paper introduces tile embeddings, a unified representation for 2D game tiles, enabling procedural level generation with less reliance on annotated datasets, and demonstrates its effectiveness in predicting tile affordances and generating levels.

Contribution

The paper proposes tile embeddings learned via autoencoders, reducing the need for extensive human annotations in procedural level generation for 2D games.

Findings

Tile embeddings effectively predict affordances for unseen tiles.

The representation supports procedural level generation for both annotated and unannotated games.

Autoencoder-based learning captures visual and semantic tile information.

Abstract

In recent years, Procedural Level Generation via Machine Learning (PLGML) techniques have been applied to generate game levels with machine learning. These approaches rely on human-annotated representations of game levels. Creating annotated datasets for games requires domain knowledge and is time-consuming. Hence, though a large number of video games exist, annotated datasets are curated only for a small handful. Thus current PLGML techniques have been explored in limited domains, with Super Mario Bros. as the most common example. To address this problem, we present tile embeddings, a unified, affordance-rich representation for tile-based 2D games. To learn this embedding, we employ autoencoders trained on the visual and semantic information of tiles from a set of existing, human-annotated games. We evaluate this representation on its ability to predict affordances for unseen tiles, and to serve as a PLGML representation for annotated and unannotated games.