Research Papers

A Novel Framework to Evaluate Yielding Behavior and Crossing Decisions at Non-signalized Locations using Video-based Trajectory Data

Filename 6C_1_Fu_Paper.pdf
Filesize 1 MB
Version 1
Date added June 28, 2017
Downloaded 6 times/fois
Category 2017 CARSP XXVII Toronto
Tags Research and Evaluation, Session 6C, Student Paper Award Winner
Author/Auteur Ting Fu, Luis F. Miranda-Moreno, Nicolas Saunier
Stream/Volet Research and Evaluation
Award/Prix Étudiant 3 Student

Slidedeck Presentation

6C_1_Fu

Abstract

The vulnerability of pedestrians explains why vehicles should yield right-of-way to pedestrians at crosswalks. Yielding is therefore a critical part of interactions at unsignalized intersections;

Past research has considered yielding compliance, but their definition is ambiguous. Besides, there are situations where it is impossible for the vehicle to yield considering their proximity to the crosswalk. Such situations are likely considered as violations in most previous studies.

Different studies have investigated pedestrian safety at the interaction level; however, most studies have not investigated pedestrian-vehicle interactions in a microscopic way, i.e. how vehicles and pedestrians compromise to each other under different position and speed situations. Aiming to address the above-mentioned research gaps in the pedestrian safety literature, the main purpose is to propose a new framework to investigate pedestrian safety at unsignalized crosswalks, and to apply it to explore safety issues and countermeasure efficiency. In the framework, the vehicle yielding maneuver to a pedestrian is split into reaction and braking. The distance required for a yielding maneuver depends on driver reaction time and deceleration rate that the vehicle can achieve. The framework then categorizes vehicles approaching crossing pedestrians with a certain speed as: 1) unable to make a full stop, 2) able to yield depending on the driver reaction time, and 3) able to and legally required to yield. Based on this, non-yielding maneuvers are classified as non-infraction maneuvers, confusing maneuvers and violations. Pedestrian crossing decisions are classified as dangerous, risky and safe crossings. Yielding compliance and yielding rate, as the ratio-based measure, are redefined. Time to crossing and deceleration rate required for the vehicle to stop are used as measures of collision probability. The framework, represented in the distance-velocity diagram, is called the DV model. It is demonstrated through a case study involving three different types of unsignalized crossings including: painted, unprotected, and stop-sign-controlled locations, with the help of video-based tracking techniques which extract trajectory data including distance and velocity information.

Results generally meet the framework assumptions. For instance, no single yielding maneuver is observed for interactions in situation 1).

Significant differences (a maximum difference of 16.9 % points) with large variance between the overall yielding rate and the compliance were observed.

Stop sign controlled crosswalk has the highest yielding ratios, highest mean TC and lowest mean DRS at pedestrian occurrences and crossing decisions, while unpainted crosswalk is the opposite. • Results indicate that the framework works reasonably.

Differences between the yielding rate and the compliance show that the yielding rate used in many past studies does not properly describe yielding compliance.

Comparison results show that the stop sign controlled crosswalk performs best for pedestrian safety, while the unprotected crosswalk is the worst. A new framework is proposed to learn pedestrian-vehicle interactions in a potentially more precise and microscopic way. This framework can be used for different purposes including treatment evaluation, behavior analysis, safety monitoring (violation detecting), improvements of yielding enforcement policies, and even, pedestrian-vehicle interaction simulation. However, the model needs to be validated further through sufficiently large numbers of observations. "

Ting Fu, Luis F. Miranda-Moreno, Nicolas Saunier