Abstract

Procedural Content Generation for 3D game levels faces challenges in balancing spatial coherence, navigational functionality, and adaptable gameplay progression across multi-floor environments. We introduce a framework centered on the offline, LLM-assisted construction of reusable databases for architectural components (facilities and room templates) and gameplay mechanic elements. Our multi-phase pipeline:

  1. Selects and arranges instances from a Room Database to form a multi-floor global structure with inherent topological order;
  2. Optimizes internal facility layouts in each room using predefined constraints from a Facility Database;
  3. Integrates progression-based mechanics by placing components from a Mechanics Database according to topological and spatial rules.

A subsequent two-phase repair system guarantees navigability. Experiments on diverse scenarios show the framework generates varied, navigable levels and simulates distinct pacing strategies via simple parameterization, advancing scalable PCG for complex 3D environments.

Citation

Xu, K., & Verbrugge, C. (2025, November). A Database-Driven Framework for 3D Level Generation with LLMs. In Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE 2025).

@inproceedings{xu2025dbLLM,
  author    = {Kaijie Xu and Clark Verbrugge},
  title     = {A Database-Driven Framework for 3D Level Generation with LLMs},
  booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE)},
  year      = {2025},
  month     = {Nov}
}