Modelling and control of road traffic networks

Road traffic networks offer a particularly challenging research subject to the control community. The traffic congestion around big cities is constantly increasing and is now becoming a major problem. However, the dynamics of a road network exhibit some complex behaviours such as nonlinearities, delays and saturation effects that prevent the use of some classical control algorithms. This thesis presents different models and control algorithms used for road traffic networks. The dynamics are represented using a "fluid-flow" approach. This leads to a system of quasi-linear hyperbolic partial differential equations which represents the behaviour of the drivers on each road. The boundary conditions are represented by a set of algebraic relations describing the behaviour of the drivers at the junctions. Two models with different complexities are introduced and their properties analysed. Different control algorithms are presented. One method is focused on the steady state case and intends to minimise a "sustainable cost" function. This function takes into account a time cost, the pollution and the accident risk. Two other methods which are able to deal with transient effects are also presented. The first one is a routing strategy expressing how to spread the traffic flow between two paths leading to the same destination. The second one is a ramp metering strategy using linear feedback.