Modeling the Appearance and Behavior of Urban Spaces.

The complexity of urban spaces and of the phenomena that take place in them calls for the use of customized computational tools within the city planning and design workflow. We develop a framework that benefits different parts of this workflow, including the design of 3D representations of planned developments, the estimated prediction of the effects of these developments on city behavior, and the visualization of these effects for further analysis. Our approach infers values of urban model variables from user-specified geometric and behavioral constraints and high-level design goals, enforces the procedural generation process to produce geometric assets that behaviorally and geometrically resemble plausible real-world cities, and automatically creates 3D urban models for visualizing behavioral phenomena occurring in urban spaces. Our framework allows users to experiment with alternative spatial and functional configurations of a city, while interactively receiving visual and quantitative feedback that facilitates assessing and understanding the effects of their choices.