Science and Technology

Modelling traffic congestion as a contagion

Modelling traffic congestion

In this PhD research the aim is to understand the spatio-temporal characteristics of congestion described by a contagion process. The proposed model enables observation, forecast and control of traffic congestion in a network over time.

Traffic congestion is among the biggest problems most metropolitan areas face. Congestion propagation behaves as a spreading phenomenon moving from one link to another. During peak periods, it spreads throughout the network and then recovers.

This behaviour resembles the spread of a virus throughout a population that could be explained with a contagion model known as Susceptible-Infectious-Recovered (SIR) model. An analogous model can be used to describe propagation and dissipation of traffic congestion in the network.

The preliminary analysis suggested the dependence of congestion propagation on two newly introduced parameters, congestion propagation rate and dissipation rate as macroscopic characteristics of network traffic.

The proposed traffic model is examined using a large-scale simulation-based dynamic traffic assignment model of Melbourne as well as empirical link speed data from Google.

Participants

Project background

The spread of traffic congestion in urban networks has long been viewed as a complex spatiotemporal phenomenon that often requires computationally intensive microscopic models. The spread of congestion at the link level is well theorised and understood with queuing and kinematic wave theories.

However, our understanding of congestion propagation dynamics at the network level is still incomplete. Further, the lack of available transportation network data in many countries, especially those that are developing, poses a challenge for traffic modellers. However, the fast-paced development and deployment of mobile sensors offers the opportunity to generate continuous spatial data, which further enables the estimation of road traffic conditions in real time and at the macroscopic level.

The high spatial correlation in the road network suggests, propagation of congestion has behavioural resemblance to spreading phenomenon (e.g. an infectious disease, internet virus etc.) moving from one link to another. During a peak period, it spreads throughout the network and then recovers. The widely used contagion model known as Susceptible-Infectious-Recovered (SIR) model has been most resourceful in epidemic case studies.

Hence, this analogous model can also be used to describe propagation and dissipation of traffic congestion in the network. The proposed model can be used for adaptive and predictive control of congestion in an urban network.

By monitoring the network in real time and observing the number of congested links, the model can be applied to develop optimal control strategies with different objectives such as minimising the total duration of congestion, the total number of congested links, and the recovery time.

See the full list of iMOVE projects here

Project objectives

The project aims to describe propagation and dissipation of congestion in an urban network using a contagion model.

Therefore, the following objectives are specified:

  • Proposing a new model for network traffic congestion propagation based on spreading phenomena
  • Validate the proposed model with simulation-based and empirical data
  • Analyse the impact of disruption on network performance

The project will be undertaken in three phases. In the first stage, a comprehensive literature review on existing approaches and methods for macroscopic traffic modelling will be conducted. This includes more recent developments in large-scale network traffic modelling.

In the second stage, new methods and algorithms will be developed to model congestion propagation and dissipation at the network level using contagion-based approaches with homogenous mixing assumption.

The third stage will see the extension of the proposed modelling framework to networks with non-homogenous mixing. A range of simulation software (e.g. AIMSUN, VISSIM, etc.), TRACSLab (driving simulator laboratory) and empirical traffic data from Google will be used by the PhD candidate at each stage.

World interest already!

Work on this topic by supervisor Dr Meead Saberi, of the CityX Lab, Research Centre for Integrated Transport Innovation (rCITI) at the University of NSW, and his PhD student Mudabber Ashfaq has already generated interest from press across the world.

See:

A simple contagion process describes spreading of traffic jams in urban networks
Traffic jams are contagious. Understanding how they spread can help make them less common
Turns out, traffic spreads like the coronavirus

More from iMOVE Australia

Most Popular

ADVERTISEMENT

To Top