Optimisation of polling networks with limited service disciplines


One of the ultimate goals of this research topic is the performance analysis of multiple traffic intersections in a network, and finding optimal control policies for the traffic signals at these intersections. Among the existing queueing models, polling systems are a natural choice to model traffic intersections, because they have similar features: multiple queues being served by a single server in cyclic order. Nevertheless, traffic intersections have some complications compared to polling models:

  • multiple flows (corresponding to queues in the polling model) can be served simultaneously
  • vehicle-actuated traffic signals have minimum and maximum green times
  • urban traffic intersections are part of a network, implying that arrivals at (some) flows are generated by the departure process of another traffic intersection.

The performance analysis and optimisation of networks of polling systems form the two parts of this sub-project. We study networks of polling systems with limited service disciplines. This means that either the length of one service period, or the number of customers that can be served in one service period, is constrained. A large part of this project focusses on networks operating in the critically loaded regime (heavy traffic).

Supervisors Rudesindo Nunez Queija (CWI/UvA), Marko Boon (TU/e)
PhD Student Abhishek
Location University of Amsterdam (UvA)