IntroImage for Research Focus 2

Simulating Traffic and Intelligent Traffic Network Control

The Modern City – A Traffic Network

Congested traffic networks have become a common condition during rush hours causing economic losses by increasing travel time and air pollution. The potential to optimise traffic flow within the existing road network infrastructure has been overlooked in the past. However, modern cities experience increasing traffic, and the need for intelligent network control to minimise negative effects is currently driving research worldwide.

Traffic Models and Simulation

Traffic flow optimisation requires a sufficiently accurate and adaptive traffic model that is applicable  online. A macroscopic model that meets these criteria has been developed at our Institute. The general idea of macroscopic traffic models is to treat traffic flow like a compressible fluid that can be described by a nonlinear partial differential equation. The number of model parameters in macroscopic traffic models is comparatively small. Hence, large-scale networks can be simulated at acceptable computational cost.

Utilising traffic sensor data, a method to automatically estimate the model parameters is used to update the model parameters online. Both temporal and spatial variations in driving behaviour are taken into account. Thereby, the model accurately represents the average driving behaviour and predicts the traffic situation in the near future.  Current signalling is incorporated in the simulation, such that the evolution of the network-wide traffic situation is accurately represented.

The concept of the Fisher Information Matrix is utilised to investigate optimum Design-of-Experiment (traffic sensor type and placement) as well as identifiability of model parameters. Additionally, approaches for model reduction in situations where not all model parameters can be  identified from the data set at hand are developed and tested.

Simulation results are validated through comparison with results from commercial traffic simulation software packages. Together with our industry partner,  we plan and conduct a model validation experiment on Austrian motorways as well as in a small traffic network in Hallein, Salzburg.

Intelligent Network Control Algorithms

The particular problem of traffic control is that several different goals are reasonable depending on the current local and global state of traffic. For low network loads, it is desirable to minimise travel time, whereas for high network loads the maximization of total flow is preferred. Additionally, significant disturbances both known (major events, long-term construction sites, etc.) and unknown (accidents, lane closures, weather changes) occur. To capture these disturbances, the underlying traffic model must be adapted online, ideally, fully automatically and reliably. Several control strategies seem promising at the current state of research and will be investigated, tested, and validated.

Projects of this Research Unit

ConnAT - Connecting Austria

Contact

TU Wien

Institute of Mechanics and Mechatronics

Research Unit of Control and Process Automation

Getreidemarkt 9 / E325-04 / 6th floor

1060 Vienna

E-Mail to Secretary

Location and how to get there

Publication

E. Thonhofer, E. Luchini, A. Kuhn, S. Jakubek
Online Parameter Estimation for a Flexible, Adaptive Traffic Network Simulation, Proceedings of the ICCVE, 2014