Using reaction times and accident statistics for safety impact prediction of automated vehicles on road safety of vulnerable road users. (June 2023)

• Record Type: Journal Article
• Authors: Hula, Andreas; de Zwart, Rins; Mons, Celina; Weijermars, Wendy; Boghani, Hitesh; Thomas, Pete
• Is Part Of: Safety science. Volume 162(2023)
• Publication Date: 2023-06
• Subjects: Automated Vehicles -- Accident Statistics -- Power Model -- Vulnerable -- Road Users -- Reaction Times -- Dose-Response Curve; Industrial accidents -- Periodicals; Accident Prevention -- Periodicals; Safety -- Periodicals; Travail -- Accidents -- Périodiques; 363.11
• DOI: 10.1016/j.ssci.2023.106091
• Languages: English
• ISSNs: 0925-7535

Abstract

Highlights: We investigate reductions in crash numbers to be expected by the introduction of Level 5 autonomous Vehicles. Estimates are based on accident causes eliminated and potential improvements in braking capacity. We derive formulae that can be used with common parameters in microsimulation (Reaction Times, Decelerations). We discuss how to connect resulting estimates with the outcomes of microsimulations. Abstract: The global effort to make automated vehicles a common reality on the roads of the world is intensifying in recent years and the challenges of automated driving are being investigated in ever more detail and variety. Among the foremost promises of using automated vehicles (AVs) in daily commute is their assumed benefit on road safety, which should offer additional safety to the most vulnerable road users (pedestrians, cyclists), as well as other vehicles. In this article we aim to derive a functional relationship (dose–response curve) between the proportion of automated vehicles on the road (penetration rate) and the expected accident numbers/fatalities of interactions with vulnerable road users. Our approach is built upon two fundamental components: Firstly, based on an analysis of current accident causes, we can make a projection of which causes of accidents between cars and vulnerable road users could ideally be mitigated by AVs and which not. Secondly, for the accidents that are not mitigated, we still assume a potential for reduction of accident occurrence and accident severity, based on the assumed reaction time of an automated vehicle, compared to the reaction time of a human driver. Based on statistics, braking distances and reaction times, as well as the power model, we derive an estimate of the potential reductions in accidents and fatalities in the presence of automated vehicles, ultimately expressed as a relationship between the proportion of automated vehicles (penetration rate) and the accident numbers/fatality rates of accidents between motorized vehicles and vulnerable road users.