Instant Architecture

✨ Summary

Paper summary

The paper introduces a procedural framework for automatically generating detailed architectural models from a shared database of design rules. Its central formalism, the split grammar, is a restricted three-dimensional set grammar operating on attributed, parameterized geometric shapes. Split rules subdivide a shape while preserving its occupied volume, whereas conversion rules replace one basic shape with another contained within the original volume. These restrictions make the grammar suitable for automatic derivation while preventing the uncontrolled growth and geometric overlap associated with some growth-oriented grammars.

The framework separates spatial structure from final appearance. The split grammar recursively derives a building layout, including façades and architectural elements such as windows, doors, cornices, and wall sections. Attributes attached to shapes and rules subsequently determine material and geometry. Attributes also express high-level goals, such as style, complexity, symmetry, shop presence, or vertical and horizontal emphasis.

Two mechanisms provide control over automatic rule selection. The attribute-matching system compares intervals, containment requirements, priorities, and statistical preferences associated with a current shape and candidate rules. Deterministic matching first eliminates incompatible rules; stochastic selection then chooses among compatible alternatives while using per-building random values to preserve coherence. The control grammar is a separate attributed context-free grammar that distributes design decisions spatially—for example, consistently across a floor, façade column, or selected region—rather than applying them independently and randomly.

The authors implemented a database of approximately 250 rules and 40 attributes using ten basic shapes. Generated buildings generally contained between 1,000 and 100,000 polygons and required roughly one to three seconds to create on the reported hardware. The examples demonstrate façade variation, coherent ornamentation, nontrivial three-dimensional floor plans, cylindrical structures, and iterative refinement as additional design goals are specified. The principal limitation is that highly complex architectural details and configurations must be supplied as terminal shapes or handled by separate modeling tools. The authors also note that practical urban applications would require substantially larger rule databases and more user-friendly authoring tools.

Influence on subsequent research

The paper established split grammars as an important foundation for procedural architectural modeling. A direct successor, Procedural Modeling of Buildings (2006), developed the CGA shape grammar approach for producing detailed building shells and is explicitly part of the same research lineage. (cg.tuwien.ac.at) Later work extended split-grammar methods to curved and conic architectural geometry, addressing limitations of the original box-oriented formulation. (sciencedirect.com) Subsequent research and applications have also reused or adapted split-grammar concepts for architectural reconstruction, user-facing procedural modeling, and alternative environments such as Minecraft. (sciencedirect.com) These references document a continuing research influence on grammar-based procedural modeling; the sources reviewed here do not by themselves establish a specific commercial deployment of this paper’s original implementation. The publication metadata and date are corroborated by the TU Wien research record. (cg.tuwien.ac.at)