01506167

Component

http://onlinepubs.trb.org...ces/2011/RSS/2/Duong,D.pdf

A Pareto Archived Dynamically Dimensioned Search (PA-DDS) algorithm is introduced for the calibration of microscopic traffic simulation platforms and compared to other single and multi-criteria calibration approaches. PA-DDS algorithm explicitly considers trade-off in errors among the various constituent fitting function such that any solution on the non-dominated curve cannot be improved without incurring a corresponding degradation in error in at least one of the constituent fitness criteria. For example, improvement in speed error cannot be achieved without increased error in either volume or safety performance or both. In this paper, PA-DDS is used to calibrate selected VISSIM model parameters based on observed traffic data obtained from the Federal Highway Administration (FHWA) Next Generation Simulation (NG-SIM) vehicle tracking study. The calibration seeks to obtain best-estimate parameter values that minimize residual mean square percentage error (RMSPE) for three fitness criteria: speed, volume, and Crash Potential Index per vehicle (a surrogate safety performance metric). This comparison clearly demonstrates that the multi-criteria PA-DDS algorithm yields best-estimate parameter values with acceptable residual errors for the two traffic factors of speed and volume, as well as for the safety performance criteria. The best estimate VISSIM parameters obtained from the PA-DDS application were found to differ significantly from default values and from values obtained based on other calibration methods that do not explicitly consider trade-off errors in fitness criteria. A number of solution sets were obtained from the PA-DDS algorithm, with a range of parameter values. The best estimate solution (lowest overall model goodness-of-fit) yielded parameters that differed from other PA-DDS solution sets (with higher overall error). This suggests that trade-offs (non-dominated sets) of parameter values can have significant implications for values that correspond to the lowest overall model goodness-of-fit.

Distribution, posting, or copying of this PDF is strictly prohibited without written permission of the Transportation Research Board of the National Academy of Sciences. Unless otherwise indicated, all materials in this PDF are copyrighted by the National Academy of Sciences. Copyright © National Academy of Sciences. All rights reserved

01501394

English

21p

2011

3rd International Conference on Road Safety and Simulation

Digital/other

Figures; References; Tables

Algorithms; Microscopic traffic flow; Speed; Traffic data; Traffic safety; Traffic simulation; Traffic volume

Next Generation Simulation program; U.S. Federal Highway Administration; VISSIM (Computer model)

Surrogate safety measures

Highways; Safety and Human Factors; I80: Accident Studies

TRIS, TRB, ATRI

Jan 29 2014 9:55AM