A novel agent-based framework for evaluating pedestrian safety at unsignalized mid-block crosswalks

• Published in: Accident Analysis & Prevention
• Main Authors: Zhu, Hong, Almukdad, Abdulkarim, Iryo-Asano, Miho, Alhajyaseen, Wael K.M., Nakamura, Hideki, Zhang, Xin
• Format: Article in Journal/Newspaper
• Language: English
• Published: Elsevier Ltd.
• Subjects: Agent-based model; Pedestrians-vehicles interaction; Post encroachment time; Reaction time; Unsignalized crosswalks; Visual obstacle; Refuge Islands
• Online Access: http://hdl.handle.net/10576/35298; https://doi.org/10.1016/j.aap.2021.106288; https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85109463144&origin=inward

A critical safety problem in road networks is the conflicts on unsignalized crosswalks. Thus, a proactive approach to assess pedestrian safety performance is required. Simulating the microscopic road user behavior in a virtual platform is one of the typical approaches. However, current simulation tools are not flexible enough to properly reproduce various behaviors of pedestrians and drivers considering their interactions with the road environment and other road users. Therefore, the objective of this paper is to propose a novel agent-based framework for evaluating pedestrian safety at unsignalized crosswalks. Unsignalized mid-block crosswalks with refuge islands (UMCR) are considered as an example facility to implement the proposed framework, where relevant behavioral elements such as the reaction time, visual field with obstacles, and minimum safety margin time are addressed. Not only the pedestrian-vehicle interaction is modeled but also the vehicle-vehicle interaction is considered. Empirical validation on a UMCR shows that the proposed framework can reproduce reliable distributions of the post encroachment time compared to the observed distributions within an acceptable error range. The sensitivity analysis reveals that the high reaction time of drivers, small safety margin time, and visual obstacles near crosswalks increase the probability of serious conflicts. The results also show that the framework can reproduce traffic crashes under assumed extreme road conditions. The Qatar–Japan Research Collaboration Application Award [M-QJRC-2020-8] from Qatar University. The JSPS KAKENHI Grant Number 20 K04720.