This is the authors' abstract. We don't add key points for this paper.
Accurate urban geometric layouts are critical for urban planning, simulation, and sustainability management; however, no ideal solution currently exists for large-scale urban geometric layout creation. We propose a generative AI-based urban layout model capable of encoding arbitrary 3D city blocks to a unified latent representation and generating urban layouts for the 330 cities in North America that have over 100,000 inhabitants. Given only a few percentage of the city blocks, our approach is able to generate an entire realistic 3D city, to support street-scale physical simulations, to predict social-economic metrics, and to enable “what-if” scenarios for policy-making. Compared to data-driven approaches, ours provides a unified representation suitable for large-scale urban simulation and policy-making as part of a digital twin framework, which would otherwise require a time- and resource-consuming process of obtaining the position, geometry, and height of all buildings individually.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
He et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: