Topological Distance Between Nonplanar Transportation Networks

Cite as: Abdelkader, A., G. Boeing, B. T. Fasy, D. L. Millman. 2018. “Topological Distance Between Nonplanar Transportation Networks.” Fall Workshop on Computational Geometry. Queens, New York.

Download PDF

Abstract

In recent years, street network analysis has received substantial attention in scholarly and professional urban planning and transportation engineering. Street networks are typically modeled as a graph with intersections as nodes connected to one another by street segments as edges. This graph model enables the computation of various metric and topological measures by efficient algorithms, including accessibility, connectivity, centrality, eccentricity, betweenness, clustering, block lengths, and average circuity. One drawback of most existing research in street network analysis is the common assumption of planarity or approximate planarity of the network. In a planar model, the street network is represented in two dimensions such that grade-separated edge crossings such as bridges and tunnels create artificial nodes in the graph. While planar simplifications can be useful for computational tractability, they misrepresent many real-world urban street networks. The failure of this assumption can lead to substantial errors in analytical results. For the present study, we focus in particular on map comparison, which can serve the field of urban morphology and the evaluation of map reconstruction algorithms. A recent line of work showed the advantage of topology-based measures for map comparison. Using the common representation of streets by their centerlines, streets are split into a sequence of line segments or street segments in 2D. Leveraging insights from the nascent field of topological data analysis, the shape of the network can be described by tracing how the connectivity of street segments evolves as segments are gradually thickened. These methods of topological map comparison can be applied in transportation engineering and urban planning research/practice to quantify the difference between two graphs — accounting for non-planarity — to measure street network evolution over time, to compare the structural and functional differences among proposed urban design alternatives, or to match trip trajectories to pre-existing infrastructure models. We extend and enhance the topology-based measure presented in to accommodate grade-separated nonplanar street networks. Our work is similar in spirit to the study of multi-layered environments in motion planning. With the aim of extending known algorithms and data structures for 2D spaces to accommodate surfaces embedded in 3D, the surface is partitioned into layers. Each layer is required to project onto the ground plane without self-intersection. The connections between layers can be thought of as staircases and ramps; ignoring heights, paths across layers are measured by the projected distance.