Evaluating Symmetry and Order in Urban Design: A Computational Approach to Predicting Perception of Order Based on Analysis of Design Geometry

Abstract

Symmetry and order are among the fundamental principles in architecture and urban design. Although symmetry and order are often associated with each other, the ways we think about the two are different. Symmetry has a precise mathematical definition; thus, in principle, it is possible to compute it. Order, in contrast, is a very vague and subjective notion. The goal of this work is to establish a link between computable symmetry and perception of order in the context of urban design. The full complexity of interaction between symmetry, order, and perception in real-world urban design is beyond the scope of the presented thesis. Therefore, I have put my best efforts to narrow down the scope of the research topic and to rule out insignificant or controversial details. As a result, the presented work consists of four parts. First, I limit the analysis of urban design to two-dimensional top-down views. I present Quick Urban Analysis Kit - a tool that has been used as an online urban design exercise. It has allowed me to collect design data in a machine readable format specific to the needs of the study. Second, I develop an efficient symmetry detection approach that processes edges in a design drawing to find potential symmetry patterns and measure their significance. This part of the work serves as a computational foundation of the analysis. Third, I conduct an online survey experiment to asses the perception of order in design drawings. Using the results of the experiment, I create a statistical model of the perceived order measure. Last, I build a regression model connecting computed symmetry and perceived order and conclude the work by analyzing the results. I show that the model agrees with the perception of order in the collected set of design drawings. I estimate the uncertainty of choices in the conducted experiment and assess the accuracy of the model predictions. Finally, I provide the evidence explaining the structure of dependence between different types of symmetry and the perceived order measure.