Citation

Arun A, Haque MM, Lyon CA, Sayed T, Washington S, Loewenherz F, Akers D, Bandy M, Bahl V, Ananthanarayanan G, Shu Y. City of Bellevue, Washington USA. Bellevue, Washington USA: City of Bellevue, Washington USA, 2022.

Copyright

(Copyright 2022, City of Bellevue, Washington USA)

The full document is available online.

Abstract

Pedestrian safety is of grave concern to traffic engineers. In the United States, 6,205 pedestrians were killed, and approximately 76,000 were injured in 2019. More concerning is the fact that the relative proportion of pedestrian fatalities among all crash fatalities has been steadily increasing over the past ten years (National Center for Statistics and Analysis (NCSA), 2020, December). A pedestrian was estimated to be killed every 85 minutes and injured every 7 minutes in traffic crashes in 2019. Therefore, targeted treatments for improving pedestrian safety at crucial locations such as urban intersections are needed.

Leading Pedestrian Intervals (LPI) are an innovative signalized intersection treatment involving a pre-timed allocation to allow pedestrians to begin crossing the street in advance of the next cycle of vehicle movements (AASHTO, 2014). It helps reduce the “element of surprise” for right-turning vehicles. However, the evidence from the literature on the effectiveness of the LPI treatment is mixed.(King, 2000, Van Houten et al., 2000, Fayish and Gross, 2010, Sharma et al., 2017, Goughnour et al., 2021).

Modern advancements in sensor technology offer an unprecedented opportunity to assess the effectiveness of safety treatments remotely and proactively using traffic conflict analysis (Tageldin et al., 2018, Zheng et al., 2018, Guo et al., 2020a, Guo et al., 2020b). Therefore, this study undertook a traffic conflict-based before-after type safety evaluation of a 5-second LPI treatment implemented at three signalized intersections in the city of Bellevue, Washington. The traffic conflicts were automatedly extracted from video-captured traffic movements using advanced Computer Vision and Deep Learning techniques. The study applied a bivariate peak-over threshold modelling approach to model the tail distributions of pedestrian-vehicle Time-to-collision (TTC) conflicts, with the objective of studying whether the LPI treatment leads to a reduction in such conflicts. Additionally, the study also modelled rear-end TTC conflicts between vehicles to investigate whether the LPI treatment, on the other hand, led to an increase in vehicular conflicts ....