Empirical Comparison of Simulation Models with Different Input Data Structures
No Thumbnail Available
Files
Date
2002-11-18
Authors
Journal Title
Series/Report No.
Journal ISSN
Volume Title
Publisher
Abstract
This thesis focuses on an empirical comparison of CORSIM and Paramics, two commonly used traffic simulation models with different input data structures.
The case comparison was executed between a field-validated CORSIM model and a fully calibrated Paramics network. These two models were constructed based on the same physical network dataset, which was originally created for the CORSIM simulation purposes. For those input data that were necessary for Paramics, but not available in this dataset, estimations were performed based on the known data and, sometimes, based on CORSIM default values. Of these the most important one was the Origin-Destination (OD) matrix.
To enter traffic demand in Paramics, an OD matrix was derived using two different methods, namely a statistical fitting method and a stochastic assignment method. The feedback results from a Paramics test network showed that the stochastic assignment method was more effective in deriving a good OD solution.
One straightforward finding of the comparison was that Paramics generated what appeared to be a larger percentage of unsuccessful runs than CORSIM. That was possibly because Paramics created more link flow fluctuations with the dynamic feedback traffic assignment algorithm; therefore, it had a higher chance of spillback or blockage for overloaded links or turn movements.
A comparison of link flows in the two simulation models was executed based on the sample replications after excluding outliers. It displayed that there were some apparent link flow discrepancies between these two models. To ensure a meaningful comparison of other selected traffic performance measures, two critical corridors with minor vehicle flow discrepancies were selected as the comparison sites.
By comparing the results on one corridor (NB LaSalle) , Paramics generated fewer vehicle trips and a higher vehicle travel speed, while on the other corridor (WB Ontario), the reverse occurred: although Paramics had fewer vehicle trips on that corridor, it still produced lower vehicle speeds than CORSIM.
The research suggests that empirical comparisons of simulation models with different input data structures are feasible and informative for model validation and selection. Further, for the same traffic demand, Paramics generated traffic performance that is at variance with CORSIM?s when using dynamic feedback traffic assignment algorithm.
Description
Keywords
OD matrix, model comparison, Paramics, CORSIM, traffic simulation
Citation
Degree
MS
Discipline
Civil Engineering
