MSc Thesis Proposal

Anticipative control of traffic networks

Mentor: B. De Schutter

Keywords: traffic, control, optimization

Description:
The congestion problem on the Dutch highways is an everyday issue in national politics. The direct costs of the delays in 1998 were estimated about 1.7 billion guilders. Since the societal relevance is so high, several attempts are made to reduce the number and length of the traffic jams. One way to improve the situation is to better utilize the existing infrastructure by controlling the traffic flows.

An important topic in this context is the short or long term reaction reaction of the drivers. Consider, e.g., the route information that is given to the drivers via DRIPs (Dynamic Route Information Panels) or via radio broadcast. This information can (and hopefully will) influence the drivers' route choice. So, these messages can be used to control traffic flows. The given information usually describes the current situation or a prediction for the short term future. However, when drivers react to these messages, the predictions will be no longer valid, since the actual traffic situation changes as a consequence to these reactions. Then the messages will be adjusted, and the drivers will react again. This is an undesirable situation and can even become unstable. This is illustrated in the figure below.

Unstable cycle due to adaptive DRIP messages.
An example of an unstable situation that could arise when drivers react to DRIP messages.

When a congestion warning is given for the northern part of the ring rood, many drivers will shift their route choice to the southern route. However, this may cause congestion on the south ring also while at the same time alleviating congestion on the north ring. As a consequence, the DRIP message is adapted, and more drivers are again taking the route via the north ring, which may again result in extra congestion on the north ring, and less on the south ring, so that we are again in the original situation.

Furthermore, encountering situations that differ from what has been announced can frustrate the drivers and might decrease their confidence in these messages. This is highly undesirable, because the drivers' confidence is necessary to be able to control the traffic. Thus, if we want to control the traffic via route information, he given predictions (and control) should take both the drivers' reaction and the control objectives into account. (Model) predictive control seems to be a good approach to tackle these problems, because predictions about the future behavior and the development of traffic flows can be considered. The students task is to select or develop a model which can describe the network effects of the most important control measures and the drivers' reactions, and find an optimal way to operate these control measures. Optimality in the traffic context could mean: improved flow and/or speed, decreased travel times, etc.


If you are interested in selecting this project as your MSc project, please come along or send an email for more information.

This page is maintained by Bart De Schutter.