Spectral (Re)construction of Urban Street Networks: Generative Design Using Global Information from Structure

Abstract

Modeling and analysis of urban form is typically performed using local generative design techniques, (e.g., shape grammars), with closed sets of local rules operating on elements. While this approach is powerful, the open variety of possible non-unique choices over the element and rule sets does not answer an important closure question: How much information, i.e., how many elements and rules, exhaustively capture all the information on structure? This paper investigates the inverted principle: using global system information to reconstruct a design. We show that orthogonal eigenmodes of a street network’s adjacency matrix capture global system information, and can be used to exactly reconstruct these networks. Further, by randomly perturbing the eigenmodes, new street networks of similar typology are generated. Thus, eigenmodes are global generators of structure. Outcomes provide new mechanisms for measuring and describing typology, morphology, and urban structure, and new future directions for generative design using global system information.