Using Ensemble Analysis to study the effects of Network Topology on Performance in Urban Road Networks

TL;DR

This paper introduces an experimental framework for analyzing how urban road network performance depends on structural properties, using synthetic network ensembles that are plausible and comparable, to better understand the impact of demand and infrastructure characteristics.

Contribution

It proposes a novel framework for generating realistic synthetic network ensembles and systematically studying their performance variations under different structural parameters.

Findings

Performance varies with network density and connectivity.

Demand levels significantly influence link volume-to-capacity ratios.

Network structure impacts the Price of Anarchy in urban road networks.

Abstract

Studies of the effect of network structure on performance have, thus far, been restricted to examining ensembles of synthetic networks generated by canonical models from the Network Science literature, which do not plausibly represent real road networks. Furthermore disparate ranges of parameter settings for demand and supply structure are often used, making it difficult to compare one study with another. To address these deficiencies, this paper proposes an experimental framework for the design of numerical investigations to study how network performance varies with respect to the structural properties of urban road networks. The key step within this experimental framework is the generation of a spectrum of ensembles of synthetic networks, which vary with respect to a single aspect of network structure and which adhere to plausibility constraints that ensure relevance to real urban road networks. This paper then demonstrates the application of this experimental framework to an investigation of how two performance indicators; the average link Volume-to-Capacity ratio and the Price of Anarchy, vary with respect to the density of travel demand, and the size, density and connectivity of network infrastructure. In so doing, several key challenges of such investigations are identified, with respect to the computational burden of the experiments and the need for empirical data on some important aspects of network structure such as the distribution of capacity.