Icon Legend

This session is not in your schedule.

This session is in your schedule. Click again to remove it.

12 Views

ICTD 2025

Transportation Safety

Cybersecurity Challenges in AI-Driven Intelligent Transportation Systems: Securing Autonomous, Connected, and Electric Vehicles

How to reduce the spatial scope of sections with high crash risks on mountainous roads?

Wednesday, June 11, 2025

8:30 AM – 8:45 AM MT

Primary Author(s)

CZ

Changjian Zhang, Ph.D

Postdoctoral Resesarch Associate
University of Hawaii
Honolulu, Hawaii, United States

Presenter(s)

CZ

Changjian Zhang, Ph.D

Postdoctoral Resesarch Associate
University of Hawaii
Honolulu, Hawaii, United States

Co-Author(s)

JH

Jie He, n/a

Prof.
Southeast University
Nanjing, Jiangsu, China (People's Republic)
XY

Xintong Yan, n/a

Postdoc
Southeast University
Nanjing, Jiangsu, China (People's Republic)
Hanyi Yang, Ph.D

Postdoctoral Research Assistant
University of Hawaii at Manoa
Honolulu, Hawaii, United States
Guohui Zhang, PhD

Professor
University of Hawaiʻi at Mānoa
Honolulu, Hawaii, United States
Yiwei Wang, n/a

Ph.D Candidate
University of Hawaii at Manoa
HONOLULU, Hawaii, United States

Submitter(s)

CZ

Changjian Zhang, Ph.D

Postdoctoral Resesarch Associate
University of Hawaii
Honolulu, Hawaii, United States

Abstract: In practice, identifying the long-term road crash risk still relies primarily on crash-based methods, which are significantly limited by issues such as long collection times, underreporting, and being reactive. This study developed a generic framework to address these problems through proactive detection. The development of the generic framework involves four steps. First, the Symbolic Aggregate Approximation algorithm is used to convert the smoothed time series into a lower-dimensional symbolic sequence. Second, a Negative Binomial Regression model is employed to test the statistical relationship between the characters and long-term road crash risk. Next, the SHAP value algorithm is applied to analyze the importance of the significant characters. Finally, a Binary Logit model is constructed using characters with above-average importance as explanatory variables to estimate the conditional probability of a road segment being a high-risk area. We tested the performance of the framework using real-world tire force time series collected from a mountainous road in China. Initially, the framework was designed to simultaneously detect the long-term risks for four types of crashes (total crashes, rear-end, fixed-object hitting, and overturning). However, the results indicate that tire forces are only significantly related to the long-term risk of rear-end. The framework ultimately outputs the rear-end probability for each individual road segment, and classical probability theory is used to demonstrate the reliability of the identification results. This framework can also provide the potential spatial extent of high-risk road segments, which helps reduce screening costs for road safety optimization and risk management. Additionally, this is a type of proactive detection technology that can serve as a reference for safety management in areas with limited crash and traffic flow data.

Learning Objectives:

Attendees can expect to learn the following from this session:

proactively reduce the spatial scope of potential crash black spots
conduct risk analysis using time series of factors closely related to long-term crash risk
recognize the role of tire mechanical properties in long-term road crash risk analysis