Home > People > Joan Serras > My PhD Thesis
My PhD Thesis
Multidimensional Multilevel Representation for Microscopic Traffic
Simulation Models
ABSTRACT:
Computer technology has enabled new breeds of transport models which
can simulate millions of vehicles in very large areas. TRANSIMS, one
of the most powerful, models at the disaggregate level of millions
of individuals using a synthetic micropopulation. Like other
simulation models, TRANSIMS models links and routes at a single
level. Although TRANSIMS has modelled the road traffic in
Switzerland, it cannot model the whole of continental Europe in an
integrated way.
This thesis investigates a Multilevel Multidimensional Representation (MMR) devised in the nineteen seventies that claims it can represent road systems of any size allowing microdynamics to aggregate into the higher level dynamics, and macro level dynamics to disaggregate to the lower level dynamics in a coherent way.
The MMR was devised long before the development of large dynamic modelling systems, and is untested. Here TRANSIMS has been used to generate dynamic traffic flows over the large area of Milton Keynes to test the MMR hypothesis.
Generally there are many equally attractive routes between two nodes
in a network. The central idea in the MMR involves representing sets
of routes as a more abstract higher level link between two nodes.
When these are combined they give combinatorially many combinations
of sub-routes. At the lowest level, N, are the 'links on the
ground'. Combinations of these make 'routes on the ground'. These
are aggregated to make links at level N+2, which can be aggregated
to form N+2 routes, which can be aggregated into N+3 links, and so
on indefinitely. The higher level links have flow and travel time
distributions derived from the links and routes at lower level. The
MMR is based on the definition of hierarchical zones at levels using
boundary nodes to define the higher level links. MMR enables
holistic road network modelling with an unlimited numbers of levels.
Generating traffic flow and travel time data from TRANSIMS showed volatility when trying to get the system to converge, even after many feedback iterations. This nearly compromised the whole research project, since it was unclear if the chaotic and unnatural-looking traffic patterns were due to misuse of TRANSIMS or bugs in our implementation. However, in combination with our interest in higher level link data, it has opened up a new and surprising research direction.
Some very simple networks studies of TRANSIMS' convergence give unexpected results: the classical Fundamental Diagram of Traffic Flow is not applicable to the majority of urban road systems; another diagram better reflects the dynamics of relatively short road segments between intersections; and this provides a new way to model very large transport systems.
Not only does MMR provide a practical way of modelling very large road systems in a holistic way, but it supports new ways of modelling road traffic dynamics and extends the power of modelers such as TRANSIMS.
Thesis: Multidimensional Multilevel Representation for Microscopic Traffic Simulation Models.
Supervisors:
Jeff Johnson and Stephen Potter
Page Last Updated: 26 October, 2008