Analysis of distribution and calibration of car-following sensitivity parameters in microscopic traffic simulation models

Grant G. Schultz, L. R. Rilett

Research output: Contribution to journalArticle

28 Citations (Scopus)

Abstract

The calibration of microscopic traffic simulation models has been a topic of increased attention in the transportation and traffic engineering profession. These calibration analyses have traditionally been concerned with identifying the "best" parameter set from a range of acceptable values. A methodology to introduce and calibrate a low parameter distribution is examined; it uses measures of central tendency and dispersion (i.e., mean and variance) to generate input parameters for car-following sensitivity factors in microscopic traffic simulation models. The approach is applied to IH-10 in Houston, Texas, with the CORSIM model and subsequently calibrated with an automated genetic algorithm methodology to examine the effectiveness of the distribution alternatives. An overview of car-following sensitivity parameters is provided, parameter distribution alternatives are outlined, and an application of the car-following distribution alternatives is discussed and compared with default distributions. The results of this analysis indicate that the distribution of car-following sensitivity parameters can be modeled in microscopic traffic simulation models and that with automated calibration methodologies such as genetic algorithms, the mean absolute error between simulated and observed traffic volume and travel time data can be minimized. The results further indicate that both lognormal and normal distribution alternatives are very effective at replicating observed conditions and representing the full distribution of car-following sensitivity factors from the calibration of the mean and variance of these parameters.

Original languageEnglish (US)
Pages (from-to)41-51
Number of pages11
JournalTransportation Research Record
Issue number1876
DOIs
StatePublished - Jan 1 2004

Fingerprint

Railroad cars
Calibration
Genetic algorithms
Travel time
Normal distribution

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Mechanical Engineering

Cite this

@article{c326bef96b4546b88a41319c9792e892,
title = "Analysis of distribution and calibration of car-following sensitivity parameters in microscopic traffic simulation models",
abstract = "The calibration of microscopic traffic simulation models has been a topic of increased attention in the transportation and traffic engineering profession. These calibration analyses have traditionally been concerned with identifying the {"}best{"} parameter set from a range of acceptable values. A methodology to introduce and calibrate a low parameter distribution is examined; it uses measures of central tendency and dispersion (i.e., mean and variance) to generate input parameters for car-following sensitivity factors in microscopic traffic simulation models. The approach is applied to IH-10 in Houston, Texas, with the CORSIM model and subsequently calibrated with an automated genetic algorithm methodology to examine the effectiveness of the distribution alternatives. An overview of car-following sensitivity parameters is provided, parameter distribution alternatives are outlined, and an application of the car-following distribution alternatives is discussed and compared with default distributions. The results of this analysis indicate that the distribution of car-following sensitivity parameters can be modeled in microscopic traffic simulation models and that with automated calibration methodologies such as genetic algorithms, the mean absolute error between simulated and observed traffic volume and travel time data can be minimized. The results further indicate that both lognormal and normal distribution alternatives are very effective at replicating observed conditions and representing the full distribution of car-following sensitivity factors from the calibration of the mean and variance of these parameters.",
author = "Schultz, {Grant G.} and Rilett, {L. R.}",
year = "2004",
month = "1",
day = "1",
doi = "10.3141/1876-05",
language = "English (US)",
pages = "41--51",
journal = "Transportation Research Record",
issn = "0361-1981",
publisher = "US National Research Council",
number = "1876",

}

TY - JOUR

T1 - Analysis of distribution and calibration of car-following sensitivity parameters in microscopic traffic simulation models

AU - Schultz, Grant G.

AU - Rilett, L. R.

PY - 2004/1/1

Y1 - 2004/1/1

N2 - The calibration of microscopic traffic simulation models has been a topic of increased attention in the transportation and traffic engineering profession. These calibration analyses have traditionally been concerned with identifying the "best" parameter set from a range of acceptable values. A methodology to introduce and calibrate a low parameter distribution is examined; it uses measures of central tendency and dispersion (i.e., mean and variance) to generate input parameters for car-following sensitivity factors in microscopic traffic simulation models. The approach is applied to IH-10 in Houston, Texas, with the CORSIM model and subsequently calibrated with an automated genetic algorithm methodology to examine the effectiveness of the distribution alternatives. An overview of car-following sensitivity parameters is provided, parameter distribution alternatives are outlined, and an application of the car-following distribution alternatives is discussed and compared with default distributions. The results of this analysis indicate that the distribution of car-following sensitivity parameters can be modeled in microscopic traffic simulation models and that with automated calibration methodologies such as genetic algorithms, the mean absolute error between simulated and observed traffic volume and travel time data can be minimized. The results further indicate that both lognormal and normal distribution alternatives are very effective at replicating observed conditions and representing the full distribution of car-following sensitivity factors from the calibration of the mean and variance of these parameters.

AB - The calibration of microscopic traffic simulation models has been a topic of increased attention in the transportation and traffic engineering profession. These calibration analyses have traditionally been concerned with identifying the "best" parameter set from a range of acceptable values. A methodology to introduce and calibrate a low parameter distribution is examined; it uses measures of central tendency and dispersion (i.e., mean and variance) to generate input parameters for car-following sensitivity factors in microscopic traffic simulation models. The approach is applied to IH-10 in Houston, Texas, with the CORSIM model and subsequently calibrated with an automated genetic algorithm methodology to examine the effectiveness of the distribution alternatives. An overview of car-following sensitivity parameters is provided, parameter distribution alternatives are outlined, and an application of the car-following distribution alternatives is discussed and compared with default distributions. The results of this analysis indicate that the distribution of car-following sensitivity parameters can be modeled in microscopic traffic simulation models and that with automated calibration methodologies such as genetic algorithms, the mean absolute error between simulated and observed traffic volume and travel time data can be minimized. The results further indicate that both lognormal and normal distribution alternatives are very effective at replicating observed conditions and representing the full distribution of car-following sensitivity factors from the calibration of the mean and variance of these parameters.

UR - http://www.scopus.com/inward/record.url?scp=15544378089&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=15544378089&partnerID=8YFLogxK

U2 - 10.3141/1876-05

DO - 10.3141/1876-05

M3 - Article

AN - SCOPUS:15544378089

SP - 41

EP - 51

JO - Transportation Research Record

JF - Transportation Research Record

SN - 0361-1981

IS - 1876

ER -