Modeling a vehicular traffic network. Part I

TL;DR

This paper introduces three mathematical models for vehicular traffic networks using graph theory and stochastic matrices, enabling traffic forecasting and analysis of different system configurations.

Contribution

It presents novel models incorporating both closed and open traffic systems, including nonlinear dynamics with multiple stationary states, applied to real city traffic data.

Findings

Models successfully forecast traffic flow over multiple time steps.

Stationary states are identified for different network configurations.

Application to Tigre city traffic demonstrates practical utility.

Abstract

We propose three models for the traffic of vehicles within a network formed by sites (cities, car-rental agencies, parking lots, etc.) and connected by two-way arteries (roads, highways), that allow forecasting the vehicular flux in a sequence of $n$ consecutive steps, or units of time. An essential approach consists in using, as an "a priori" information, previous observations and measurements. The formal tools used in our analysis consists in: (1) associating a digraph to the network where the edges correspond to arteries and the vertices with loops represent the sites. (2) From an initial set of numbers, that are the distribution of vehicles within the network, we construct a matrix that we transform into a stochastic matrix (SM) by normalizing the rows, whose entries are now transition probabilities. This matrix becomes the generator of the evolution of the traffic flow. And (3), we use the Perron-Frobenius theory for a formal analysis. We investigate three models: (a) a closed four-site network having a conserved number of vehicles; (b) to this network we add an influx and an outflux of vehicles to characterize an open system; asymptotically, $n \rightarrow \infty$, the SM raised to the power $n$ goes to a unique stationary matrix. And (c), we construct a nonlinear model because the formal structure permits the existence of several ($L$) stationary states for the distribution of vehicles at each site, that alternate cyclically with time. Each state represents the traffic for $L$ different moments. These models were used to analyze the traffic in a sector of the city of Tigre, located in the province of Buenos Aires, Argentina. The results are presented in a following paper.