Purdue University Graduate School
Browse
- No file added yet -

INTERSECTION CRASH EXPANSION FACTORS BASED ON PROBABILITY MODELS APPLICABLE TO TRAFFIC CONFLICTS

Download (867.43 kB)
thesis
posted on 2022-07-27, 15:27 authored by Xueqian ShiXueqian Shi

  

The major concern about vehicle crashes has led to a great amount of research on the topic in the road safety area. Nevertheless, real-world crash data collection periods are often extensive and they result in a great delay in improving safety. Therefore, surrogate measures of safety, such as traffic conflicts, are considered for safety management.

The definition of a traffic conflict has evolved over the course of half a century. One of the current definitions encompasses a failure-based road event that inevitably results in a crash if no evasive action is taken by involved road users. This counterfactual concept was validated with specific road events datasets, including rear-end events and vehicle-bicycle encounters. However, observing conflicts for an extended period is still a major difficulty. For example, a LIDAR-based technique applicable to intersections can collect conflict data for a relatively short period of several days. The LiDAR-collected data are then converted to the corresponding expected crash frequency during the observation period, which eventually must be expanded to the corresponding annual value. The conversion step has not been sufficiently addressed in the past research. Thus, an important task of estimating the annual expected crash frequency based on a short-term estimate remains unanswered. Addressing this need is the research objectives and contribution of this study.

Advanced statistical methods allow developing models to estimate expected crash frequencies for annual and short periods. The ratio of such two estimates is defined as an expansion factor in this study. This thesis presents the modeling effort and its results for different types of crashes at signalized and unsignalized intersections in Indiana. Traditional and emerging data, such as traffic volume, speed, road characteristics, weather, and other features were collected and assembled at randomly selected 194 intersections. Then, they were used to estimate the logistic models of hourly crash probability. The models were then utilized to calculate expansion factors for a specific intersection.to evaluate the method.

Funding

JTRP/INDOT

History

Degree Type

  • Master of Science

Department

  • Civil Engineering

Campus location

  • West Lafayette

Advisor/Supervisor/Committee Chair

Andrew P. Tarko

Additional Committee Member 2

Qifan Song

Additional Committee Member 3

Samuel Labi

Usage metrics

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC