Citation

Yang H, Wang Z, Xie K, Dai D. Transp. Res. C Emerg. Technol. 2017; 82: 138-160.

Copyright

(Copyright © 2017, Elsevier Publishing)

DOI

10.1016/j.trc.2017.06.016

PMID

unavailable

Abstract

Secondary crashes are non-recurrent incidents that frequently affect traffic operations and safety. They are an important performance measure in evaluating traffic incident management programs. Although several methods (e.g., static, contour map-based, and shockwave-based) have been introduced to identify secondary crashes, the applications of the existing methods are often limited by their shortcomings such as the needs for extra incident information, assumptions, simplified model structures, etc. As an alternative, this paper aims to develop a new data-driven analysis framework to support the identification of secondary crashes. Unlike existing methods, the proposed approach is concentrated on exploring the untapped potential of ubiquitous probe vehicle data for secondary crash analysis. It consists of three major components: detection of the impact area of a primary crash, estimation of the boundary of the impact area, and identification of secondary crashes within the boundary. The first component uses clustering methods to highlight the congested area induced by a primary crash. The second component develops metaheuristic optimization algorithms to approximate the boundary of the congested area. With the estimated boundary, a novel identification method is introduced to automatically identify secondary crashes within the boundary. The performance of the proposed approach has been tested under a set of simulation scenarios. The test results show that the proposed approach based on the ant colony optimization can best describe the impact area and re-identify up to 95 percent of the simulated crashes. Although the performance of the proposed approach is related to the market penetration rate, the results suggest that a relatively low market penetration rate can already achieve promising performance.

Language: en

Keywords

Clustering; Impact area; Probe vehicle data; Secondary crashes; Shockwave; Traffic incident