Citation

Orsini F, Gecchele G, Gastaldi M, Rossi R. Transportmetrica A: Transp. Sci. 2019; 15(2): 556-572.

Copyright

(Copyright © 2019, Informa - Taylor and Francis Group)

DOI

10.1080/23249935.2018.1515271

PMID

unavailable

Abstract

This work contributes to study the application of extreme value theory (EVT) in road safety analysis, estimating the risk of being involved in an entering-circulating collision in single-lane roundabouts. Detailed trajectory data of the vehicles were derived from a driving simulator experiment, and the time-to-collision (TTC) was used as a surrogate measure of safety. Three EVT approaches were applied, tested and compared: (1) the Generalized Extreme Value distribution used in the block maxima (BM) approach, (2) the Generalized Pareto Distribution used in the peak-over-threshold approach (POT), with negated-TTC (nTTC), and (3) shifted-reciprocal-TTC (srTTC). Case-study results analysis showed that BM and POT with shifted-reciprocal-TTC confidence intervals included the number of observed crashes, while POT with negated-TTC did not include it. According to these findings, both BM and POT-with-shifted-reciprocal-TTC appear promising and deserve further attention in order to develop effective ready-to-practice crash prediction models, useful in intersection design and operational analysis.

Language: en

Keywords

collision prediction; driving simulator; proactive crash prediction models; Road safety; surrogate safety measures