Investigating the Empirical Existence of Static User Equilibrium

TL;DR

This study empirically tests the existence of static user equilibrium in real traffic data from Paris, revealing significant heterogeneity and chaotic dynamics that challenge the assumption of stationarity in traffic modeling.

Contribution

It provides the first empirical validation of the static user equilibrium assumption using real-world traffic data, highlighting its limitations due to observed heterogeneity and chaos.

Findings

High spatial and temporal variability in shortest paths.

Invalidation of local stationarity assumption.

Chaotic evolution of congestion patterns during peak hours.

Abstract

The Static User Equilibrium is a powerful framework for the theoretical study of traffic. Despite the restricting assumption of stationary flows that intuitively limit its application to real traffic systems, many operational models implementing it are still used without an empirical validation of the existence of the equilibrium. We investigate its existence on a traffic dataset of three months for the region of Paris, FR. The implementation of an application for interactive spatio-temporal data exploration allows to hypothesize a high spatial and temporal heterogeneity, and to guide further quantitative work. The assumption of locally stationary flows is invalidated in a first approximation by empirical results, as shown by a strong spatial and temporal variability in shortest paths and in network topological measures such as betweenness centrality. Furthermore, the behavior of spatial autocorrelation index of congestion patterns at different spatial ranges suggest a chaotic evolution at the local scale, especially during peak hours. We finally discuss the implications of these empirical findings and describe further possible developments based on the estimation of Lyapunov dynamical stability of traffic flows.