Dec. 20 2005 3:40PM

Dec. 20 2005 4:00PM

Increasing number of cars, especially in big cities, becomes very pressing. Existing ways of traffic control are insufficient and it is necessary to look for some more efficient way of control, such that is able to “see” the traffic situations
more globally and which is able to learn and adapt according to the actual traffic situation. In the lecture, a new hierarchical approach to the control of a large transportation area will be presented. The basic modelled element is a traffic microregion. It is a collection of junctions controlled by signal lights and the arms connecting the junctions and possessing measurement devices giving transportation data. The result is state-space model describing the lengths of queues standing in the arms of controlled junctions. The model is estimated using Kalman filter. After estimation of the queues, the optimization is performed, minimizing the weighted sum of the next queues by proper setting of greens proportions. For optimization, linear programming is used. Now, optimal control of each microregion means a fight between them. That is why some superior layer of control is necessary to coordinate them. This
control is achieved with practically the same type of model, when the analogy between junction and microregion is considered (e.g. queue length - number of cars in microregion, etc.). The goal is to keep all the microregions equally
loaded. The algorithms of the estimation and optimization have been tested on traffic simulator AIMSUN and they seem very promising.

International Workshop on Data - Algorithms - Decision Making (Contact person: Martin Janžura)