Establishing the Variation of Dynamic Traffic Assignment Results Using Subnetwork Origin-Destination Matrices
Traffic simulation often uses random number generation to incorporate stochasticity into the model. Generally, traffic simulation should attempt to incorporate randomness associated with the probabilities observed in real world traffic scenarios. Difficulties arise, however, when an analyst attempts to use traffic simulation data to predict changes in traffic flow based on an impact to the network. Effects on network operation could be due to the impact scenario, or it could be a result of randomness associated with the model. Common statistical measures for this analysis assume independent and identical distributions, but traffic simulation results are inherently spatially autocorrelated. This paper investigates the proposed structural similarity index, a statistic that incorporates the spatial relationship among subnetwork ODs, as a method to quantify the baseline variation that can be expected from simulated traffic flow. Stochasticity is particularly influential on the results of dynamic traffic assignment simulation. In light of many different software packages containing these random effects, some of which have proprietary barriers that prevent them from being controlled, this study attempts to quantify the variation associated with multiple runs of the same network. This study focuses on the subnetwork size required to evaluate local rerouting due to a traffic control plan. Once the subnetwork boundary demand of the base and impact scenarios are no longer statistically different, it is assumed that the subnetwork size is sufficient. It was found that the root mean squared error was the most useful measure for evaluating the subnetwork size relative to an impact scenario.
-
Supplemental Notes:
- This paper was sponsored by TRB committee ADB30(8) Paper Review Group #4.
-
Corporate Authors:
500 Fifth Street, NW
Washington, DC United States 20001 -
Authors:
- Bringardner, Jack W
- Gemar, Mason D
- Boyles, Stephen D
- Machemehl, Randy B
-
Conference:
- Transportation Research Board 93rd Annual Meeting
- Location: Washington DC
- Date: 2014-1-12 to 2014-1-16
- Date: 2014
Language
- English
Media Info
- Media Type: Digital/other
- Features: Figures; Maps; References; Tables;
- Pagination: 18p
- Monograph Title: TRB 93rd Annual Meeting Compendium of Papers
Subject/Index Terms
- TRT Terms: Dynamic traffic assignment; Forecasting; Geographic information systems; Origin and destination; Traffic flow; Traffic simulation
- Subject Areas: Highways; Operations and Traffic Management; Planning and Forecasting; I72: Traffic and Transport Planning; I73: Traffic Control;
Filing Info
- Accession Number: 01520100
- Record Type: Publication
- Report/Paper Numbers: 14-3107
- Files: TRIS, TRB, ATRI
- Created Date: Mar 26 2014 10:13AM