Procedural Generation of 3D Maps with Wave Function Collapse: Optimization and Advanced Constraints

The Wave Function Collapse algorithm is a widely used Procedural Content Generation technique for creating structured scenarios using local neighborhood constraints. This work presents an extension of the algorithm to generate three-dimensional scenarios, incorporating non-local constraints and key optimizations. The proposed improvements include assigning weights to tiles, layer-based generation, specific appearance constraints for unique or ranged tiles, and an automated neighbor creation and assignment method using connectivity rules. These modifications facilitates the generation of coherent and structured 3D environments, providing greater control and adaptability to the process. Finally, some optimizations are proposed and the approach's effectiveness is evaluated analyzing the impact of constraints on the algorithm's coherence, diversity, and runtime.