Comparison of Various Time-to-Collision Prediction and Aggregation Methods for Surrogate Safety Analysis

摘要: INTRODUCTION Traditional methods of road safety analysis rely on direct accident observations, data sources which are rare and expensive to collect also carry the social cost placing citizens at risk unknown danger. Surrogate is a growing discipline in field that promises more pro-active approach diagnosis. This methodology uses non-crash traffic events measures thereof as predictors collision probability severity they significantly frequent, cheaper collect, have no impact. Time-to-collision (TTC) an example indicator indicates primarily: smaller TTC, less likely drivers time perceive react before collision, thus higher outcome. Relative positions velocities between users or user obstacles can be characterised by course corresponding TTC. Meanwhile, driving speed (absolute speed) primarily severity. The travelling speed, stored kinetic energy dissipated during impact . Similarly, large differentials with stationary may contribute severity, though TTC depends relative distance well. Driving used extensively stopping-sight models , some even suggesting modulate their emergency braking response travel Others content there little empirical evidence relationship Many surrogate been literature, especially recently renewal automated collection methods, but consistency definitions indicators, interpretation, transferability results still lacking. While wide diversity demonstrates research thriving, remains need comparison for order make practical practitioners. For example, time-to-collision events, definition lacks rigour literature. Also lacking systematic validation different techniques. Some early attempts made Swedish Traffic Conflict Technique  using trained observers, recent across methodologies, preferably objectively-defined measures, needed. Ideally, this would done respect crash crash-based second best method compare characteristics all same set, public benchmark very limited despite efforts . objectives paper review interpretation one most ubiquitous least context-sensitive namely time-to-collision, i) consistent, recent, and, importantly, objective ii) set numerous sites, iii) latest developments analysis. work examines use various motion prediction constant velocity, normal adaptation observed patterns, (for its properties transferability), space aggregation continuous indicators. represents application largest sets date.