Faculty of Science and Technology

This talk will focus on compositional methods for modelling vehicular traffic and on distributed estimation methods for real time traffic systems.

Compositional models describing the evolution of the traffic density, and possibly also of the average speed of traffic, in different sections of a road network, are proposed. Hybrid models - combining discrete event models with very simple continuous evolution equations - allow compact modelling of fairly complicated behaviour.

In many cases road traffic controllers will have available a fairly large, but often an inaccurate and unreliable, set of measurements of the traffic state at the current time. Combining the on-line measurements with the compositional models we develop recursive estimators for the traffic state, and detection of faults (sensor faults, or traffic incidents). Particle filtering and other Bayesian techniques can be used. Distributed implementations of the estimation algorithms make possible their application in practice. Finally, these state estimates can be used for on-line control actions, e.g., synchronising traffic in closed loop.