Physics engine improves verification of building designs for safety codes
Vibe Building
Computational Engineering, Finance, and Science
Summary
Designing buildings that meet strict safety rules for earthquakes and wind is hard. Most computer programs create designs that look right but haven't been thoroughly tested against these rules. This paper presents a new system where a physics engine is used directly to check if building designs will pass safety tests. The language model only suggests changes, but the physics engine truly decides if designs are safe or not. This method results in more reliably safe building designs compared to relying on language models alone.
What this means in practice
- •For architectural design teams: Use physics-engine evaluations to generate building designs that reliably meet seismic and wind safety codes during early planning.
- •For civil engineering firms: Integrate physics-adjudicated design checking into workflows to reduce costly errors in code compliance before construction begins.
- •For computer-aided design software developers: Develop software tools embedding physics engines to automatically verify and suggest safe building designs compliant with multiple safety codes.$Commercial implications: Enables commercial design software to offer automated, verified code-compliant building design solutions to construction professionals.
Authors
Yongqing Jiang, Haoran Luo, Jianze Wang, Xin Zhou, Kaoshan Dai, Zhiqi Shen
Abstract
Automated building design must comply with seismic and wind codes and satisfy structural mechanics constraints, yet most existing agents produce visually plausible models without verification grounded in mechanical analysis and code compliance. We introduce the Vibe Building task and propose PE-Loop (Physics-Engine-in-the-Loop), an agent in which a deterministic physics engine is the sole source of evaluation signals, mapping code constraints to a physics process reward, while the language model is confined to proposing discrete revisions (section menu, topology, and lateral system). Designs are verified by held-out seismic and wind time-history checks and a constructability gate. On VB-Bench, 3,577 physics-adjudicated building instances across six code families, PE-Loop achieves the highest verified success rate under three of four backbone LLMs, the highest held-out seismic pass rate under all four, and the highest held-out wind pass rate under three. Replacing the physics verdict with a language-model judge, all else fixed, leaves 58.43% of accepted designs noncompliant. These results suggest that reliable structural design rests less on a stronger LLM proposer than on an adjudicator the proposer cannot influence, a division of labor for agents whose outputs must hold up in the physical world.