The Loss of Efficiency Caused by Agents’ Uncoordinated Routing in Transport Networks

Zhongzhi Xu,Junjie Wang,Pu Wang,Li Sun

doi:10.1371/journal.pone.0111088

Abstract

Large-scale daily commuting data were combined with detailed geographical information system (GIS) data to analyze the loss of transport efficiency caused by drivers’ uncoordinated routing in urban road networks. We used Price of Anarchy (POA) to quantify the loss of transport efficiency and found that both volume and distribution of human mobility demand determine the POA. In order to reduce POA, a small number of highways require considerable decreases in traffic, and their neighboring arterial roads need to attract more traffic. The magnitude of the adjustment in traffic flow can be estimated using the fundamental measure traffic flow only, which is widely available and easy to collect. Surprisingly, the most congested roads or the roads with largest traffic flow were not those requiring the most reduction of traffic. This study can offer guidance for the optimal control of urban traffic and facilitate improvements in the efficiency of transport networks.

Highlights

In this era of unprecedented global urbanization, the fast growth of human mobility has put immense pressure on urban roads [1,2,3], which has manifested in the form of severe traffic congestion and traffic-related air pollution [4,5,6,7]
The different patterns observed for Price of Anarchy (POA) versus R in these three counties confirm that the distribution of travel demand needs to be considered in estimating the price of anarchy
We generate morning-peak commute ODs for three Bay Area counties and study the price of anarchy using actual travel demand in large-scale road networks

Summary

Introduction

In this era of unprecedented global urbanization, the fast growth of human mobility has put immense pressure on urban roads [1,2,3], which has manifested in the form of severe traffic congestion and traffic-related air pollution [4,5,6,7]. The number of trips between a pair of origin and destination N can be approximated by a power-law distribution P(N)~509:2N{3:32 for all the three counties R2w0:99, showing that travel demand between most pairs of locations was small, but there was high volume between a few origins and destinations (Figure 3d). The equilibrium flow fUE was measured using the random OD of San Francisco and observed to follow an exponential distribution P(fUE)*e{fUE=1,162:8 (Figure 3e).

Results

Conclusion