With the recent widespread deployment of intelligent transportation systems
(ITS) in North America there is an abundance of data on traffic systems and thus an
opportunity to use these data in the calibration of microscopic traffic simulation models.
Even though ITS data have been utilized to some extent in the calibration of microscopic
traffic simulation models, efforts have focused on improving the quality of the
calibration based on aggregate form of ITS data rather than disaggregate data.
In addition, researchers have focused on identifying the parameters associated
with car-following and lane-changing behavior models and their impacts on overall
calibration performance. Therefore, the estimation of the Origin-Destination (OD)
matrix has been considered as a preliminary step rather than as a stage that can be
included in the calibration process.
This research develops a methodology to calibrate the OD matrix jointly with
model behavior parameters using a bi-level calibration framework. The upper level seeks
to identify the best model parameters using a genetic algorithm (GA). In this level, a
statistically based calibration objective function is introduced to account for disaggregate
form of ITS data in the calibration of microscopic traffic simulation models and, thus,
accurately replicate dynamics of observed traffic conditions. Specifically, the
Kolmogorov-Smirnov test is used to measure the "consistency" between the observed
and simulated travel time distributions. The calibration of the OD matrix is performed in
the lower level, where observed and simulated travel times are incorporated into the OD
estimator for the calibration of the OD matrix. The interdependent relationship between travel time information and the OD matrix is formulated using a Extended Kalman filter
(EKF) algorithm, which is selected to quantify the nonlinear dependence of the
simulation results (travel time) on the OD matrix.
The two test sites are from an urban arterial and a freeway in Houston, Texas.
The VISSIM model was used to evaluate the proposed methodologies. It was found that
that the accuracy of the calibration can be improved by using disaggregated data and by
considering both driver behavior parameters and demand.
Identifer | oai:union.ndltd.org:tamu.edu/oai:repository.tamu.edu:1969.1/4409 |
Date | 30 October 2006 |
Creators | Kim, Seung-Jun |
Contributors | Burris, Mark, Rilett, Laurence R. |
Publisher | Texas A&M University |
Source Sets | Texas A and M University |
Language | en_US |
Detected Language | English |
Type | Book, Thesis, Electronic Dissertation, text |
Format | 1289818 bytes, electronic, application/pdf, born digital |
Page generated in 0.0014 seconds