We've Moved!
Visit SDSU’s new digital collections website at https://digitalcollections.sdsu.edu
Description
Continued emphasis on active transportation has led to a proliferation of vulnerable road users (VRUs) such as bicyclists and pedestrians at intersections. Intersections are critical locations as many crashes occur due to mixed traffic flow and various conflicting patterns between road users. Intersection safety has primarily relied on historical crash data. However, due to several limitations including unpredictability and irregularity of occurrence of crashes in real environment, quantitative and qualitative determination of crashes may not be accurate. This study explores alternative measures to the traditional safety analysis known as surrogate safety measures (SSMs). SSMs such as Time to Collision (TTC), Post Encroachment Time (PET) and a variant of TTC, Relative Time to Collision (RTTC) were used to evaluate safety at ten signalized intersections in the city of San Diego. The analysis was conducted in two main parts: proactive safety evaluation for VRUs at signalized intersections by comparison of SSMs and predicting critical bicycle-vehicle conflicts at signalized intersections. In part one, frequency of each SSMs was estimated to identify critical intersections for VRUs and then a comparative study of each SSMs were conducted. It was found that RTTC alone was insufficient to accurately identify critical conflicts. Furthermore, safety evaluation results showed that a single SSM was not reliable but a combination of different SSMs was necessary to ensure the reliability of evaluations. In part two, logistic regression model was developed in R to predict critical conflicts based on PET measure. Bicycle-vehicle kinematics data were monitored for certain period before predicting critical conflicts. Several scenarios were analyzed considering different combinations of PET threshold value and monitoring period, and it was found that a scenario with PET threshold value of 3s and monitoring period of 2s led to the best model based on its statistical performance. Of the many input variables investigated, velocity of the conflicting objects and minimum relative approach velocity were found to be statistically more significant. The model was tested under two cases; sensitivity maximization and maximum overall accuracy, and it was found that sensitivity maximization was more suitable as it ensured accurate prediction of critical conflicts.
