R05 → Urban Form-flow Reciprocal Generative Design
Keywords:#Urban Form-flow#Generative design#Machine Learning# Urban Heat Radiation # Wind Simulation#Python# Year:2024
This is part of the Digital Futures 2024 workshop "Urban Form-Flow Reciprocal Generative Design".
FORM-FLOW RECIPROCAL GENERATIVE DESIGN
This project explores how various urban flows—such as traffic, people, resources, and information—interact with spatial morphology in dynamic and reciprocal ways. By viewing the city as a living organism, the workshop emphasizes designing adaptable urban forms that evolve alongside internal and external flows, aiming toward a sustainable and responsive urban future.
My Contribution
In this collaborative project, I was primarily responsible for urban heat radiation and wind simulation, treating them as key environmental flows. I also led the training and deployment of a Pix2Pix generative model, integrating 12 types of urban flow diagrams to explore the bidirectional influence between form and flow. My work contributed to the environmental performance assessment of different morphological prototypes and guided the generation of spatially responsive design solutions.
REsearch process
The design workflow involves four major stages:
(1) collecting multi-source spatial and flow-related datasets of the target area;
(2) analyzing the interaction mechanisms between urban form and flow using statistical and ML-based models;
(3) constructing a generative model to propose optimized urban layouts responding to flow conditions;
(4) visualizing results through multi-scale diagrams and simulations.
GENerative Strategy
By embedding flow-related indicators into the generative modeling process, the project explores spatial configurations that are not only morphologically efficient but also improve mobility, accessibility, and energy exchange within the urban system. The iterative design process enables responsive optimization across scales, from block to district.
Tools and Technologies
The project relies on a cross-platform workflow combining Python, Rhino & Grasshopper, and machine learning libraries. It supports dynamic data parsing, model training, and visual rendering, enabling the smooth transition from analytical insight to design generation.
Key Outcomes
The final outcomes include an interactive “form-flow” generative prototype, multi-scenario simulations of spatial interventions, and a set of optimized urban morphologies that balance density, connectivity, and flow efficiency. These results demonstrate the potential of data-informed generative design in future urban transformation.