A model for the generation of road networks

TL;DR

This paper introduces a flexible network generation model based on a parameter alpha, capable of producing realistic road networks and analyzing their properties related to efficiency, robustness, and cost.

Contribution

The authors propose a novel alpha-model for road network generation that captures diverse network types and incorporates local and weighted centrality measures.

Findings

The alpha-model can replicate real-world road network structures.

Optimal alpha values relate to network robustness and efficiency.

The model provides insights into urban network self-organization.

Abstract

As part of the effort undertaken to understand urban environments and their generation, we need to explore models that produce statistically valid configurations of roads. These sort of models will help us to derive plausible mechanisms for the spatial location of population. This task is of fundamental importance, as we need to create an experimental environment that allows us to disentangle the specificities of a spatial configuration from the ideal system. Creating statistically valid models of road networks along with models of city generation allows to average the effects of geometry bringing us one step closer to understanding urban environments. To completely understand road networks we need to be able to grasp what principles of economy do their growth entail. It is therefore of interest to explore the possible shape that a performance function would have for transportation systems. In this work, we tackle this issue by proposing a network generation model based on a single parameter $\alpha$ which is capable of creating any type of network from trees to quasi-surfaces and which is shown to generate networks close to the real road networks under study. This is obtained through the definition of local and weighted versions of centrality measures. These centrality measures deal with distance-decay effects and nodes having different masses. We set ourselves to determine the properties and different regimes of this $\alpha$-model and we lay out a definition for the performance of a network taking into account factors such as robustness, construction cost, congestion and distance. We obtain the optimal alpha from the analysis of the space of possible performance functions, giving an intuition on the self-organisational properties of the original network.