Characterizing regimes in daily cycles of urban traffic using smooth-transition regressions

Yiannis Kamarianakis, H. Oliver Gao, Poulicos Prastacos

Research output: Contribution to journalArticle

40 Citations (Scopus)

Abstract

During the past few years, researchers have found evidence indicating that various time series representing daily cycles of traffic, such as volumes, speeds and occupancies, may be nonlinear. In this paper it is shown that such nonlinearities can be adequately described by smooth-transition regression (STR) models which may characterize distinct regimes for free flow, congestion, and asymmetric behavior in the transition phases from free flow to congestion and vice versa. STR models are advantageous compared to the - frequently adopted in traffic modeling - Artificial Neural Networks models because their parameters are interpretable. An exposition of smooth transition models is presented with a focus on logistic multi-regime models that are deemed to be most appropriate for modeling traffic variables. The methodology is illustrated by an application to data on speeds volumes and occupancies, obtained from two loop detectors located in a major arterial of Athens, Greece. Tests on nonlinearity provided ample evidence of regime-dependent dynamics for all traffic time series examined. The following research questions are examined using STR models: How many regimes, described by linear dynamics, characterize the cycle of traffic for a typical weekday? Which time interval corresponds to each regime and how fast is the transition from one regime to another? Are the estimated dynamics of daily cycles stable across weekdays?

Original languageEnglish (US)
Pages (from-to)821-840
Number of pages20
JournalTransportation Research Part C: Emerging Technologies
Volume18
Issue number5
DOIs
StatePublished - Oct 2010
Externally publishedYes

Fingerprint

traffic
regime
regression
time series
Time series
neural network
Greece
evidence
Smooth transition regression
Logistics
logistics
Detectors
Neural networks
methodology
Regression model
Congestion
Nonlinearity
Modeling

Keywords

  • Nonlinear traffic dynamics
  • Short-term prediction
  • Smooth-transition regression
  • Traffic states
  • Traffic-flow patterns

ASJC Scopus subject areas

  • Computer Science Applications
  • Management Science and Operations Research
  • Automotive Engineering
  • Transportation

Cite this

Characterizing regimes in daily cycles of urban traffic using smooth-transition regressions. / Kamarianakis, Yiannis; Oliver Gao, H.; Prastacos, Poulicos.

In: Transportation Research Part C: Emerging Technologies, Vol. 18, No. 5, 10.2010, p. 821-840.

Research output: Contribution to journalArticle

Kamarianakis, Yiannis ; Oliver Gao, H. ; Prastacos, Poulicos. / Characterizing regimes in daily cycles of urban traffic using smooth-transition regressions. In: Transportation Research Part C: Emerging Technologies. 2010 ; Vol. 18, No. 5. pp. 821-840.
@article{8fd223b70fdb45bca85d55dc6c862b3c,
title = "Characterizing regimes in daily cycles of urban traffic using smooth-transition regressions",
abstract = "During the past few years, researchers have found evidence indicating that various time series representing daily cycles of traffic, such as volumes, speeds and occupancies, may be nonlinear. In this paper it is shown that such nonlinearities can be adequately described by smooth-transition regression (STR) models which may characterize distinct regimes for free flow, congestion, and asymmetric behavior in the transition phases from free flow to congestion and vice versa. STR models are advantageous compared to the - frequently adopted in traffic modeling - Artificial Neural Networks models because their parameters are interpretable. An exposition of smooth transition models is presented with a focus on logistic multi-regime models that are deemed to be most appropriate for modeling traffic variables. The methodology is illustrated by an application to data on speeds volumes and occupancies, obtained from two loop detectors located in a major arterial of Athens, Greece. Tests on nonlinearity provided ample evidence of regime-dependent dynamics for all traffic time series examined. The following research questions are examined using STR models: How many regimes, described by linear dynamics, characterize the cycle of traffic for a typical weekday? Which time interval corresponds to each regime and how fast is the transition from one regime to another? Are the estimated dynamics of daily cycles stable across weekdays?",
keywords = "Nonlinear traffic dynamics, Short-term prediction, Smooth-transition regression, Traffic states, Traffic-flow patterns",
author = "Yiannis Kamarianakis and {Oliver Gao}, H. and Poulicos Prastacos",
year = "2010",
month = "10",
doi = "10.1016/j.trc.2009.11.001",
language = "English (US)",
volume = "18",
pages = "821--840",
journal = "Transportation Research Part C: Emerging Technologies",
issn = "0968-090X",
publisher = "Elsevier Limited",
number = "5",

}

TY - JOUR

T1 - Characterizing regimes in daily cycles of urban traffic using smooth-transition regressions

AU - Kamarianakis, Yiannis

AU - Oliver Gao, H.

AU - Prastacos, Poulicos

PY - 2010/10

Y1 - 2010/10

N2 - During the past few years, researchers have found evidence indicating that various time series representing daily cycles of traffic, such as volumes, speeds and occupancies, may be nonlinear. In this paper it is shown that such nonlinearities can be adequately described by smooth-transition regression (STR) models which may characterize distinct regimes for free flow, congestion, and asymmetric behavior in the transition phases from free flow to congestion and vice versa. STR models are advantageous compared to the - frequently adopted in traffic modeling - Artificial Neural Networks models because their parameters are interpretable. An exposition of smooth transition models is presented with a focus on logistic multi-regime models that are deemed to be most appropriate for modeling traffic variables. The methodology is illustrated by an application to data on speeds volumes and occupancies, obtained from two loop detectors located in a major arterial of Athens, Greece. Tests on nonlinearity provided ample evidence of regime-dependent dynamics for all traffic time series examined. The following research questions are examined using STR models: How many regimes, described by linear dynamics, characterize the cycle of traffic for a typical weekday? Which time interval corresponds to each regime and how fast is the transition from one regime to another? Are the estimated dynamics of daily cycles stable across weekdays?

AB - During the past few years, researchers have found evidence indicating that various time series representing daily cycles of traffic, such as volumes, speeds and occupancies, may be nonlinear. In this paper it is shown that such nonlinearities can be adequately described by smooth-transition regression (STR) models which may characterize distinct regimes for free flow, congestion, and asymmetric behavior in the transition phases from free flow to congestion and vice versa. STR models are advantageous compared to the - frequently adopted in traffic modeling - Artificial Neural Networks models because their parameters are interpretable. An exposition of smooth transition models is presented with a focus on logistic multi-regime models that are deemed to be most appropriate for modeling traffic variables. The methodology is illustrated by an application to data on speeds volumes and occupancies, obtained from two loop detectors located in a major arterial of Athens, Greece. Tests on nonlinearity provided ample evidence of regime-dependent dynamics for all traffic time series examined. The following research questions are examined using STR models: How many regimes, described by linear dynamics, characterize the cycle of traffic for a typical weekday? Which time interval corresponds to each regime and how fast is the transition from one regime to another? Are the estimated dynamics of daily cycles stable across weekdays?

KW - Nonlinear traffic dynamics

KW - Short-term prediction

KW - Smooth-transition regression

KW - Traffic states

KW - Traffic-flow patterns

UR - http://www.scopus.com/inward/record.url?scp=77953362241&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77953362241&partnerID=8YFLogxK

U2 - 10.1016/j.trc.2009.11.001

DO - 10.1016/j.trc.2009.11.001

M3 - Article

AN - SCOPUS:77953362241

VL - 18

SP - 821

EP - 840

JO - Transportation Research Part C: Emerging Technologies

JF - Transportation Research Part C: Emerging Technologies

SN - 0968-090X

IS - 5

ER -