Xu Z, Zheng N, Logan DB, Vu HL. Assessing bicycle-vehicle conflicts at urban intersections utilizing a VR integrated simulation approach.
ACCIDENT; ANALYSIS AND PREVENTION 2023;
191:107194. [PMID:
37402331 DOI:
10.1016/j.aap.2023.107194]
[Citation(s) in RCA: 1] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Subscribe] [Scholar Register] [Received: 03/28/2023] [Revised: 06/12/2023] [Accepted: 06/26/2023] [Indexed: 07/06/2023]
Abstract
Animosity between drivers and cyclists has existed on urban road networks for many years. Conflicts between these two groups of road users are exceptionally high in the shared right-of-way environments. Benchmarking methods of conflict assessments are mostly based on statistical analysis with limited data sources. The actual crash data would be valuable to understand the features of bike-car collisions, however the available data are spatially and temporally sparse. To this end, this paper proposes a simulation-based bicycle-vehicle conflict data generation and assessment approach. The proposed approach uses a three-dimensional visualization and virtual reality platform, integrating traffic microsimulation to reproduce a naturalistic driving/cycling-enabled experimental environment. The simulation platform is validated to reflect the human-resembled driving/cycling behaviors under different infrastructure designs. Comparative experiments are carried out on bicycle-vehicle interactions under different conditions, with data collected from a total of 960 scenarios. Based on the results of the surrogate safety assessment model (SSAM), the obtained key insights include: (1) scenarios of a high conflict probability do not lead to actual crashes, which suggests that the classic SSM-based measurements such as TTC or PET values may not sufficiently reflect real cyclist-driver interactions; (2) the major cause of conflicts is variation in vehicle acceleration, which suggests that drivers are considered to be the main party responsible for bicycle-vehicle conflict/crash occurrence; (3) the proposed approach is able to generate near-miss events and reproduce interaction patterns between cyclists and drivers, facilitating experiments and data collections which would be typically unavailable for this type of study.
Collapse