The Heterogeneity of Capacity Distributions Among Different Freeway Lanes

The stochastic nature of highway capacity has drawn increasing attention in recent studies, where the capacity distributions are usually estimated for whole freeway sections using aggregated data from multiple lanes. However, those cross section–based capacity models cannot be used to assess the risk of semi-congested states observed, in which traffic breaks down on certain lanes while still flowing uninterruptedly on others. This study attempts to investigate the freeway capacity heterogeneity among individual lanes using robust statistical estimation methods. Four diverge sections of different interstate highways in California were selected for the study. Optimal threshold speed was identified for each lane by maximizing the average reduction of traffic efficiency, which was defined as the product of mean speed and volume. Capacity observations were obtained for individual lanes, and censoring was indicated, both based on the optimal threshold speeds. Log-rank and Wilcoxon tests were conducted, and the results confirmed the heterogeneity of capacity distributions among lanes of the same freeway section and showed the necessity of estimating capacity distributions for individual lanes separately. A Bayesian hierarchical Weibull model based on censored capacity data was used to estimate these lane-specific capacity distributions. The model parameters are allowed to vary across freeway sections to account for unobserved heterogeneity and accordingly to improve the accuracy of estimations. In addition, censored data issues are adequately addressed in the proposed model. It is found that breakdown probability would be overestimated if censoring is ignored. The proposed model can provide useful insights when diagnosing bottlenecks with semi-congested cases, which may be neglected by the cross section–based models.

Language

  • English

Media Info

  • Media Type: Digital/other
  • Features: Figures; References; Tables;
  • Pagination: pp 161-176
  • Monograph Title: Celebrating 50 Years of Traffic Flow Theory: A Symposium. August 11-13, 2014, Portland, Oregon
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 01604570
  • Record Type: Publication
  • Files: TRIS, TRB, ATRI
  • Created Date: Jul 5 2016 10:53AM