Procedural Generation of Dungeons

Roland van der Linden, Ricardo Lopes, Rafael Bidarra

IEEE Transactions on Computational Intelligence and AI in Games · 2013 · 126 citations · 29 references

Concepts

TL;DR

Procedural content generation has largely been limited to specific game elements, rarely applied to full levels, yet dungeon levels—characterized by unique pacing and spatial design—offer an ideal domain for PCG, though current methods lack powerful, intuitive control and remain in early stages. This paper surveys existing procedural techniques for generating dungeon game levels. The authors review common practices, evaluate their strengths and weaknesses, and outline promising future challenges. Recent studies suggest that incorporating gameplay-related criteria offers high-level control, and the emerging foundations for gameplay-based dungeon generation show promise for mainstream PCG adoption.

Abstract

The use of procedural content generation (PCG) techniques in game development has been mostly restricted to very specific types of game elements. PCG has seldom been deployed for generating entire game levels, a notable exception to this being dungeons: a specific type of game level often encountered in adventure and role playing games. Due to their peculiar combination of pace, gameplay, and game spaces, dungeon levels are among the most suited to showcase the benefits of PCG. This paper surveys research on procedural methods to generate dungeon game levels. We summarize common practices, discuss pros and cons of different approaches, and identify a few promising challenges ahead. In general, what current procedural dungeon generation methods are missing is not performance, but more powerful, accurate, and richer control over the generation process. Recent research results seem to indicate that gameplay-related criteria can provide this high-level control. However, this area is still in its infancy, and many research challenges still lie ahead, e.g., improving the intuitiveness and accessibility of such methods for designers. We also observe that more research is needed into generic mechanisms for automating the generation of the actual dungeon-geometric models. We conclude that the foundations for enabling gameplay-based control of dungeon-level generation are worth being researched, and that its promising results may be instrumental in bringing PCG into mainstream game development.

References

29