Vulnerable road user safety evaluation using probe vehicle data with collision warning information. (February 2022)

• Record Type: Journal Article
• Authors: Matsuo, Kojiro; Chigai, Naoki; Chattha, Moazam Irshad; Sugiki, Nao
• Is Part Of: Accident analysis and prevention. Volume 165(2022)
• Publication Date: 2022-02
• Subjects: Crash risk evaluation -- Vulnerable road user (VRU) -- Pedestrian collision warning (PCW) -- Probe-vehicle data -- Empirical Bayesian (EB) estimation -- Advanced driver assistance system (ADAS); Accidents -- Prevention -- Periodicals; Accident Prevention -- Periodicals; Accidents -- Prévention -- Périodiques; 363.106
• DOI: 10.1016/j.aap.2021.106528
• Languages: English
• ISSNs: 0001-4575

Abstract

Highlights: Develop a safety performance function (SPF) for crashes against vulnerable road users, utilizing pedestrian crash warning (PCW) information acquired by connected advanced probe vehicles (APVs). Proposes a two-step empirical Bayesian estimation based on the SPF (2-step EB-SPF) to consider the issue regarding the limited number and vehicle types of APVs. Assess the effectiveness of traffic-safety treatment based on the developed 2-step EB-SPF. Abstract: Recently, connected vehicle (CV) and advanced driver assistance system (ADAS) technologies, including retrofit ADAS products, have been introduced in the real-world market. This study focuses on pedestrian collision warning (PCW) as an intensive function of the ADAS, which operates when a vehicle is at a collision risk with a vulnerable road user (VRU). Although several studies have been conducted on surrogate safety measures for crashes against VRUs, none of these studies used real-world CV data with collision warning information. Thus, the current study aims to i) develop a safety performance function (SPF) for crashes against VRUs at unsignalized intersections, where the PCW information was acquired using connected advanced probe vehicles (APVs), and ii) assess the effectiveness of a traffic-safety treatment implemented at an unsignalized intersection based on the developed SPF. In particular, this study proposes a two-step empirical Bayesian estimation based on the SPF model (2-step EB-SPF) to consider the issue regarding the limited number and vehicle types of APVs that can obtain PCW information. Based on the APV data, the vehicle–VRU crash-count negative binomial (NB) models were separately estimated using the actual PCW incidence rate and the EB estimate of PCW incidence rate, respectively. Although the actual PCW incidence rate was not statistically significant in the former model, the EB estimate of the PCW incidence rate was statistically significant and positively related to the crash count in the latter model. Moreover, a traffic-safety treatment was implemented at an unsignalized intersection and subsequently assessed as a case study based on the estimated 2-step EB-SPF model. Consequently, the model with the EB estimate of PCW incidence rate revealed that the vehicle–VRU crash risk was reduced by approximately 70%, and it was statistically significant at the 99% confidence level, which diminished the confidence interval in comparison to the model without the PCW incidence rate. Thus, the APV data including collision warning information can improve the estimation accuracy of determining the effect of the traffic-safety treatment, which can considerably contribute toward traffic safety assessment, especially for short after-treatment periods such as that prevailing in this case study.