Modelling time-varying traffic road networks

Project title:   A more flexible framework for modelling time-varying traffic road networks.

Project duration: 2/1998 - 8/2001 

Project staff:    Principal Investigator: Professor Malachy Carey

                         Research Associate:  Initially Dr Mark McCartney, then Dr Y. E. Ge

Abstract

There are currently two main approaches to modelling and predicting traffic flows and travel times varying over time on congested road networks (referred to as dynamic traffic assignment, DTA).  These are, mathematical programming and simulation, both of which have well-known advantages and deficiencies.  We propose to develop a new framework that will combine advantages of these existing approaches, while avoiding their deficiencies.  To ensure tractability, mathematical programming (MP) models for DTA are very restrictive in their treatment of link flow behaviour.  As a result, existing MP models for DTA are little used in practice and have been subjected to increasing criticism.  To overcome this, we have developed a new approach (or 'framework') that decomposes the DTA problem into a network flow MP, that coordinates flow across the network, and a set of link flow submodels, one for each link.  This allows a range of more flexible, realistic link flow models to be used, while retaining the tractability and desirable properties of the math programme network model.  To solve, we iterate between these components until mutually consistent solutions are found.  We developed user equilibrium and system optimal versions of this framework, and devised, tested and demonstrated associated versions of the solution algorithms.  Our experiments with this approach have been very successful, the algorithms showing rapid convergence to desired solutions.  We also analysed properties of the above framework models and algorithms, and derived externalities and optimal tolls (with and without queues).