Colleagues
The next CTS seminar at UCL will be given by
Prof Mike Smith
University of York
16.00 Wednesday 9 March 2011
Chadwick Lecture Theatre B05
Centre for Transport Studies
UCL
Integrated models of travel demand and network equilibrium using
tours:
theoretical properties and solution algorithms*
Abstract
Predictions are essential if interventions in transport systems are to be
sensibly considered prior to implementation. Transportation models, which
form the core of such predictions, usually comprise (a) a network
equilibrium model for forecasting congestion levels and travel costs at
given demand levels and (b) a travel demand model for predicting trip
patterns at given travel cost levels.
Historically, these elements were considered separate problems, with the
result that a plethora of algorithms and theory have been developed for
addressing them in isolation, but with a relative paucity of research into
formulating and solving the integrated model: namely, determining consistent
solutions to (a) and (b). Published academic research on integrated problems
provides convergent algorithms but only for highly restrictive demand
models, precluding many realistic behavioural dimensions required in
practice. Given the need to solve these latter problems in practice, it is
common to see heuristics adopted by the practising community that may fail
to converge, or are so time-inefficient that they are terminated far from
convergence. Furthermore, current techniques used in practice generally
focus on models dominated by origins and destinations in a single time
period (usually the peak period); or in a few periods which are unconnected.
This seminar introduces a tour-based, multi-period network equilibrium model
embracing destination choice, time period choice and out-and-return route
choice, as implied by current best practice in travel demand modelling, and
which we believe yields a problem of central importance to transport
planning. We show that the resulting variable demand network equilibrium
problem for tours stretching over a number of distinct time periods may be
written in a form which is identical to the network equilibrium model
introduced by Aashtiani and Magnanti (1983) for a single time period. It
then follows that a recently designed, convergent, two-direction algorithm
solves this problem under reasonable conditions despite the demand
interactions across time intervals. We compare the structural properties of
the two-direction algorithm with the well-known
cobweb method and another one-direction approach.
*This seminar reports on work that was undertaken jointly with Andrew Daly
and David Watling of the Institute for Transport Studies at the University
of Leeds
We hope that you will be able to join us for this
Benjamin Heydecker
________________________________________________
Benjamin Heydecker
Professor of Transport Studies
Centre for Transport Studies
UCL
Tel: 020 7679 1553
Fax: 020 7679 3042
[log in to unmask]
________________________________________________
|