Behavioral Urban AI focuses on understanding individual decisions — how people choose travel modes, residential locations, and daily activities — that together shape urban life.

Focus: We integrate discrete choice models with deep learning to capture behavioral heterogeneity, substitution patterns, and elasticities at scale. By designing interpretable neural architectures grounded in microeconomic theory, we recover behavioral insights while rigorously benchmarking machine learning against classical demand models.

Objective: The goal of Behavioral Urban AI is to make individual behavior both predictable and explainable, so that predictions remain accurate without becoming black boxes. By bridging econometrics and AI, we aim to inform equitable transportation and land-use policy grounded in how people actually decide.

Complexity Urban AI treats the city as an interconnected complex system — networks of people, places, and infrastructure that evolve across space and time.

Focus: We develop spatiotemporal deep learning and network models that capture urban dynamics, with a distinctive emphasis on uncertainty quantification, calibration, and robustness. By making deep learning trustworthy under data sparsity, disruption, and distribution shift, we model travel demand, congestion, transit operations, and infrastructure systems as they truly behave.

Objective: The goal of Complexity Urban AI is to design resilient and equitable urban systems that operate reliably under uncertainty and adapt to shocks — from natural disasters to behavioral change. By quantifying what our models do not know, we build systems that fail safely and recover gracefully.

Generative Urban AI enhances urban design, planning, and engineering by integrating generative and foundation models into the creative process.

Focus: We leverage diffusion models, vision-language models, and large language models to generate land use, mobility, and building patterns guided by sustainability goals. Rather than image-to-image mapping alone, we pair goal-conditioned generation with evaluation models that score results against explicit targets — greenery, safety, accessibility — while keeping humans in the loop.

Objective: The goal of Generative Urban AI is to automate and augment the process of urban design, exploring optimized and creative solutions that traditional methods may not uncover. By combining generation with rigorous evaluation, we push toward more adaptive, sustainable, and livable cities.