Urban AI

We develop and evaluate AI methods for analyzing urban dynamics, predicting socio-economic outcomes, and supporting planning decisions — with explicit attention to model limitations, interpretability, and real-world validation.

About

Urban planning decisions are increasingly shaped by data, models, and digital platforms. Our Urban AI research examines how these technologies can be designed, evaluated, and governed so that they support planning practice rather than obscure it.

We develop machine-learning models, decision-support systems, and urban digital twin methods for problems such as urban renewal, socio-economic impact assessment, parcel aggregation, and multi-domain urban simulation. These projects combine spatial data, planning knowledge, and computational modeling to help planners compare alternatives before interventions are implemented.

We also study the known limitations of urban AI: fragmented and heterogeneous data, model bias, poor transferability across cities, and the difficulty of translating technical outputs into actionable planning evidence. Our work therefore emphasizes interpretability, rigorous benchmarking, and validation in real planning contexts alongside partner institutions.

The aim is to establish urban AI as a reliable, auditable research and planning infrastructure — one that is useful not only under ideal conditions, but in the messy, data-scarce environments where most planning actually happens.