U-Net-Accelerated Quality-Diversity Optimization for Climate-Adaptive Urban Layouts
2026-06-03 • Neural and Evolutionary Computing
Neural and Evolutionary ComputingMachine Learning
AI summaryⓘ
The authors developed a faster way to evaluate city layouts for better climate adaptation by using a special deep-learning model (U-Net) instead of slow physics simulations. They compared their method to a traditional approach that uses Gaussian processes and found that their U-Net model learned the patterns much better, even with random training data. This improvement lets them quickly generate many building layout options that consider airflow and building density. Their method is included in an open-source tool called OpenSKIZZE, which can produce thousands of climate-friendly designs in under ten minutes.
urban layout optimizationclimate adaptationcold-air ventilationsurrogate modelsU-NetGaussian processMAP-Elitesquasi-random samplingdeep learningdesign space exploration
Authors
Alexander Hagg, Tania Guerrero, Dirk Reith
Abstract
Optimizing urban layouts for climate adaptation requires balancing building density with cold-air ventilation. Because physics-based climate simulations are computationally expensive, planners typically evaluate fewer than ten manual designs. \gls{qd} algorithms offer a way to systematically illuminate the design space, but they require surrogate models to be practical. In this paper, we replace a slow, regulatory physics simulator with a spatial deep-learning surrogate (U-Net) inside an offline MAP-Elites loop. We systematically compare this spatial approach with a traditional \gls{gp} surrogate across different training-data strategies (quasi-random Sobol sampling vs.\ active \gls{qd} bootstrapping). Our results reveal that scalar \gls{gp} surrogates fail catastrophically when trained on random samples, requiring expensive, actively generated \gls{qd} archives to generalize. In contrast, the spatial inductive bias of the U-Net allows it to learn the underlying physics mapping robustly ($R^2 = 0.996$), completely independent of the training data source. This allows offline \gls{qd} optimization to achieve highly accurate fitness rankings ($ρ= 0.994$) using only a one-time batch of random training samples. The resulting pipeline, deployed in the open-source OpenSKIZZE tool, generates thousands of diverse, climate-evaluated building layouts in under ten minutes.