Generative AI for Data-Driven Co-Design of Urban Infrastructures

Team: Gioele Zardini (Lead PI), Rudge (1948) and Nancy Allen Assistant Professor in the Department of Civil and Environmental Engineering and Jinhua Zhou (co-PI), Director of The Urban Mobility Lab and Professor of Cities and Transportation; and Marco Pavone, Associate Professor Stanford University Department of Aeronautics and Astronautics. 

Summary: We will establish generative AI principles for urban co-design, identifying interfaces, architectures, essential data, and state-of-the-art tools; will integrate the principles in the existing urban co-design framework, enhancing evaluation metrics to account for extreme scenarios, and to find optimal solutions in terms of various trade-offs (e.g., designing fleet sizes and operations with varying demands); and deploy the developed framework on an exhaustive list of real-world case studies, by allowing users to co-design an urban ecosystem in simulation via an interactive graphic interface   

“This approach paves the way to structure such urban design problems in a modular and compositional way, at the interface of intellectual and computational tractability, while leveraging the emerging strength of generative AI to produce critical, unforeseen scenarios to test. In particular, it promises to provide an interface for various stakeholders of the mobility ecosystem to reason about such complex co-design problems, blending advances in optimization and generative AI.”