Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Procedural Content Generation in Games: A Survey with Insights on Emerging LLM Integration

Mahdi Farrokhi Maleki, Richard Zhao

TL;DR

This survey addresses the need for a current, comprehensive review of Procedural Content Generation (PCG) in games, including the rise of deep learning and Large Language Models (LLMs). It categorizes PCG into search-based, ML-based, other, LLMs, and combined methods, and analyzes five content types. It provides a 2019–2023 trend analysis across major conferences, identifies gaps (notably 3D PCG and industry transfer), and highlights the prominence of LLMs and hybrid approaches. The work guides researchers and practitioners toward leveraging LLMs and integrated methods to accelerate content generation, while emphasizing evaluation and ethical considerations.

Abstract

Procedural Content Generation (PCG) is defined as the automatic creation of game content using algorithms. PCG has a long history in both the game industry and the academic world. It can increase player engagement and ease the work of game designers. While recent advances in deep learning approaches in PCG have enabled researchers and practitioners to create more sophisticated content, it is the arrival of Large Language Models (LLMs) that truly disrupted the trajectory of PCG advancement. This survey explores the differences between various algorithms used for PCG, including search-based methods, machine learning-based methods, other frequently used methods (e.g., noise functions), and the newcomer, LLMs. We also provide a detailed discussion on combined methods. Furthermore, we compare these methods based on the type of content they generate and the publication dates of their respective papers. Finally, we identify gaps in the existing academic work and suggest possible directions for future research.

Procedural Content Generation in Games: A Survey with Insights on Emerging LLM Integration

TL;DR

This survey addresses the need for a current, comprehensive review of Procedural Content Generation (PCG) in games, including the rise of deep learning and Large Language Models (LLMs). It categorizes PCG into search-based, ML-based, other, LLMs, and combined methods, and analyzes five content types. It provides a 2019–2023 trend analysis across major conferences, identifies gaps (notably 3D PCG and industry transfer), and highlights the prominence of LLMs and hybrid approaches. The work guides researchers and practitioners toward leveraging LLMs and integrated methods to accelerate content generation, while emphasizing evaluation and ethical considerations.

Abstract

Procedural Content Generation (PCG) is defined as the automatic creation of game content using algorithms. PCG has a long history in both the game industry and the academic world. It can increase player engagement and ease the work of game designers. While recent advances in deep learning approaches in PCG have enabled researchers and practitioners to create more sophisticated content, it is the arrival of Large Language Models (LLMs) that truly disrupted the trajectory of PCG advancement. This survey explores the differences between various algorithms used for PCG, including search-based methods, machine learning-based methods, other frequently used methods (e.g., noise functions), and the newcomer, LLMs. We also provide a detailed discussion on combined methods. Furthermore, we compare these methods based on the type of content they generate and the publication dates of their respective papers. Finally, we identify gaps in the existing academic work and suggest possible directions for future research.

Paper Structure

This paper contains 16 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Related Works
  3. Review Structure
  4. Novelty of the Work
  5. Content Types
  6. PCG Algorithms
  7. Search-Based Methods
  8. Machine Learning-Based Methods
  9. Large Language Models
  10. Combined Methods
  11. Analysis
  12. Trends
  13. Gaps in Research
  14. Possible Research Directions
  15. Conclusion
  16. ...and 1 more sections

Figures (3)

  • Figure 1: A timeline showing the types of algorithms appeared in PCG-related research papers during the most recent five years.
  • Figure 2: A timeline breaking down research published 2019-2023 in select game-related conferences on the topic of PCG, sorted by targeted content type. A few note-worthy LLM-based works are pointed out.
  • Figure 3: A word cloud of all authors in the published papers at the selected conferences during the 5 years. The size is proportionate to the number of co-authored papers.