(Georgia Institute of Technology, 2019-05-22)
Feng, Chen
The design of urban street networks is critical to how a city looks, feels, and functions. Moreover, the arrangement of streets inside the “superblocks”, which are the large urban areas divided up by the primary street network of the city, gives cities unique characters. This dissertation studies the street network designs at the scale of a square superblock that measures half a mile, or 800 m, on each side—a particularly common dimension for the spacing of arterial streets in the U.S., China, and many other countries. The contemporary urban landscape has been significantly shaped by two distinctive traditions for organizing streets at the scale of a superblock. At one extreme is the deployment of a uniform grid, differentiated only by street widths or intensity of development along the streets. At the other extreme is the “tree-like” pattern in which most separate branches or disjoined enclaves or loops are attached to the main streets, imposing a segregating hierarchy defined by mobility and access. This study explores street network designs that fall between these extremes; the designs in question can be described as differentiated grids. More specifically, we ask: (a) How to create differentiated grids by progressively deforming a square grid? (b) What different kinds of differentiated grids are there? (c) What is the relationship between the different rules that can be applied to creating differentiated grids and the emerging types of differentiation? To study those questions, eight different “syntactic operators” have been developed to progressively deform a street network. For each type of operation, a generative rule/algorithm was created to sequentially apply the operation on a uniform grid up to a specified number of times. An additional generative algorithm was also created to allow operations to be mixed in random sequences. Each generative algorithm was applied to generate a total of 600 differentiated street grids. This resulted in a “design universe” consisting of 5400 differentiated street grids that could be analyzed comparatively and queried for the presence of properties of interest. Such properties include graph connectivity, street density, block size and shape, intersection density, geometric regularity, directional reach, directional distance, and the diversity in syntactic conditions. In addition, the centrality structure of designs was studied. The aim was to formulate and test alternative definitions of “integration cores” and to develop relevant typologies. Consistent with space syntax literature, an integration core is defined as comprising the streets that are closer to all parts of the street network in terms of directional distance. Query algorithms were developed to select designs based on the definitions of alternative types of integration cores. Four main conclusions were reached. First, different types of operations have different capacities to influence the properties of a street network. Second, there are multiple dimensions of differentiation (e.g., differentiation in geometric alignment of streets, differentiation in configurational properties such as DDL, differentiation in block shapes, etc.). In many cases, measures along the different dimensions of differentiation are related. Their predictable relationship can be quantified. Third, while the relationship between different dimensions of differentiation usually has a consistent direction, its slope can vary, depending on the type of operation used to create the differentiation. The variation in slope suggests that properties that may be desirable (for example the creation of a diversified street grid) can be achieved with varying costs regarding properties that may be undesirable (for example the creation of less accessible streets). Fourth, the (local) generative rules used to generate designs do not necessarily lead to specific emergent global properties of the street network of the superblock. Although we cannot predict the specific syntactic type we get by applying a specific generative rule, we know that by applying certain generative rules, we are more likely to generate designs of a specific syntactic type. Thus, the thesis makes two significant contributions to the field of space syntax studies. First, it demonstrates how the systematic generation and querying of universes of designs can be used to rigorously define and enrich key syntactic ideas that have hitherto remained intuitive, such as the ideas of “deformed grid” and the “shape of the integration core”. Second, it demonstrates that in principle, the design of street networks at superblock scale can be studied according to the typologies of interface between local and global integration and according to the typologies of differentiation of the street grid.