A farewell to brake reaction times? Kinematics-dependent brake response in naturalistic rear-end emergencies

https://doi.org/10.1016/j.aap.2016.07.007Get rights and content

Highlights

  • In naturalistic rear-end emergencies, driver braking is highly kinematics-dependent.

  • Most brake onsets occur within one second after certain looming thresholds are reached.

  • Patterns of brake onset can be well explained by models assuming accumulation of looming.

  • Deceleration ramp-up (jerk) increases with kinematical urgency at brake onset.

  • Brake reaction time is a problematic concept for describing behavior in emergencies.

Abstract

Driver braking behavior was analyzed using time-series recordings from naturalistic rear-end conflicts (116 crashes and 241 near-crashes), including events with and without visual distraction among drivers of cars, heavy trucks, and buses. A simple piecewise linear model could be successfully fitted, per event, to the observed driver decelerations, allowing a detailed elucidation of when drivers initiated braking and how they controlled it. Most notably, it was found that, across vehicle types, driver braking behavior was strongly dependent on the urgency of the given rear-end scenario⿿s kinematics, quantified in terms of visual looming of the lead vehicle on the driver⿿s retina. In contrast with previous suggestions of brake reaction times (BRTs) of 1.5 s or more after onset of an unexpected hazard (e.g., brake light onset), it was found here that braking could be described as typically starting less than a second after the kinematic urgency reached certain threshold levels, with even faster reactions at higher urgencies. The rate at which drivers then increased their deceleration (towards a maximum) was also highly dependent on urgency. Probability distributions are provided that quantitatively capture these various patterns of kinematics-dependent behavioral response. Possible underlying mechanisms are suggested, including looming response thresholds and neural evidence accumulation. These accounts argue that a naturalistic braking response should not be thought of as a slow reaction to some single, researcher-defined ⿿hazard onset⿿, but instead as a relatively fast response to the visual looming cues that build up later on in the evolving traffic scenario.

Introduction

When the driver of a vehicle is suddenly faced with an unexpected, critical risk of collision, how does he or she respond? If evasive maneuvering is applied, when does it begin? How is it carried out?

Conclusive answers to these questions have been a long-standing objective of traffic safety research, and have a range of implications: In the design of roads, vehicles, or vehicle support systems for safety and automation, quantitative models of driver behavior can be very directly applied, for example in system algorithms or in computer simulations of crashes (e.g., Perel, 1982, Fambro et al., 2000a, MacAdam, 2001, Brännström et al., 2010, Markkula, 2015). In the broader study of traffic safety, the way one thinks about drivers⿿ emergency responses can also be important in more subtle ways, for example by shaping design of experiments and subsequent interpretations of results, or by guiding one⿿s analysis of actual crashes to understand their causation (e.g., Naing et al., 2009, Engström et al., 2013b), sometimes for purposes of litigation (e.g., Maddox and Kiefer, 2012).

The driver⿿s reaction time (RT) is a concept that traffic safety researchers have repeatedly made use of in models, when designing studies, and when analyzing driver behavior close to crashes. The RT usually represents the time duration from the appearance of a potential hazard, such as a lead vehicle⿿s brake lights activating, until the driver under study initiates some form of evasive response (Society of Automotive Engineers, 2015). Especially for braking responses, there is a considerable literature measuring brake reaction times (BRTs) and how they are influenced by factors such as driver age, gender, cognitive load, situation urgency, number of stimuli for the driver to consider, warnings, and so on (see for example the studies by Barrett et al., 1968, Olson and Sivak, 1986, Fambro et al., 1998, McGehee et al., 1999, Lee et al., 2002, Jurecki and Staſczyk, 2009, Jurecki and Staſczyk, 2014, Fitch et al., 2010, Ljung Aust et al., 2013; and the reviews by Olson, 1989, Green, 2000, Muttart, 2003, Muttart, 2005).

Green⿿s much-cited review (2000) aimed to determine typical RT values for different driving conditions. Expectancy was identified as the major factor determining BRT, with estimated values of 0.70⿿0.75 s for fully anticipated events, 1.25 s for unexpected but common events such as brake light onsets, and 1.5 s for surprise events such as sudden path intrusions. These canonical, situation-independent, BRT values drew criticism from Summala (2000), who pointed to evidence that BRTs for highly unexpected events can, if the traffic scenarios in question are sufficiently urgent, decrease to 1 s or lower. Similar dependencies between situation kinematics (the relative motion of involved road users, in terms of distances, speeds, etc.) and BRT have been reviewed by Muttart, 2003, Muttart, 2005 and have also been demonstrated in more recent test track and driving simulator studies (Jurecki and Staſczyk, 2009, Jurecki and Staſczyk, 2014, Engström, 2010, Ljung Aust et al., 2013). However, a detailed, large-scale analysis is still outstanding, especially for naturalistic (i.e. real-traffic) emergencies.

As for what happens beyond the point of brake onset, it has been reported from both controlled and naturalistic studies that drivers will often, but not always, show maximum deceleration levels close to their vehicle⿿s limits on the given road (McGehee et al., 1999, Fambro et al., 2000b, Lee et al., 2007). From some controlled studies, there are also reports of progressive or step-wise ramping up towards these maximum levels (Prynne and Martin, 1995, Fambro et al., 2000b, Lee et al., 2002). Again, a detailed, quantitative account of emergency braking control is lacking, especially for naturalistic data.

This paper presents time-series analyses of situation kinematics and driver braking behavior observed in naturalistic rear-end crashes and near-crashes, continuing from the work by Victor et al. (2015, pp. 76⿿84). They showed, for one set of naturalistic passenger car data, that when visually distracted drivers looked back to the road to find a rear-end collision threat, the time delay before they exhibited any discernible physical reaction to the situation was strongly kinematics-dependent. Here, these results are extended by including (1) not only driver physical reaction but also actual measured deceleration behavior, (2) events without any off-road eye glances, and (3) an additional data set of recorded events that includes truck and bus drivers in addition to car drivers.

It will be described here how drivers⿿ deceleration behavior in the studied events varied markedly with situation kinematics, in certain rather specific manners, across data sets and vehicle types. Statistical-level descriptions of this variability, potentially useful in quantitative approaches to traffic safety, will be provided. Possible psychological mechanisms behind the observed behaviors will be discussed, and it will also be argued that the findings make the concept of a ⿿brake reaction time⿿ seem inadequate as a means for describing and understanding driver behavior in surprise emergencies.

Section snippets

Data sets

The naturalistic events analyzed here came from two different sources: passenger car events from the Second Strategic Highway Research Program (SHRP 2), and passenger car, heavy truck, and bus events from the Analysis of Naturalistic External Datasets (ANNEXT) project. Table 1 provides an overview of the number of events per data set and vehicle type. In the remainder of this paper, the truck and bus events will be combined and treated together.

Within SHRP 2, the world⿿s largest naturalistic

Results

Below, the results reported in Victor et al. (2015) on physical reaction timing in the SHRP 2 data set will first be reiterated and extended with the ANNEXT data. Next, it will be described to what extent the piecewise linear model was able to describe the naturalistic deceleration behavior. Finally, the obtained model fits will be analyzed with respect to deceleration onset timing (tB), deceleration ramp-up (jB), and maximum deceleration (a1), and summarized by means of probability

Discussion

The present analyses have provided several novel insights into driver braking behavior in emergency situations. One important finding is that drivers who returned their eye gaze to the forward direction at some point before the crash almost always applied their brakes in response to the collision threat (96% of all events, 90% if considering only crashes). Previously, based on crash statistics from police reports, it has been proposed that it is fairly common for crash-involved drivers to not

Conclusions

The model-based analyses of naturalistic rear-end near-crashes and crashes presented here have provided a number of novel insights into driver emergency braking: (1) Drivers who looked toward the collision threat before impact did, with few exceptions, initiate defensive braking. (2) Brake onset almost always occurred within ±0.5 s of a visually discernible physical reaction by the driver to the collision threat. (3) Crucially, brake onset timing defied description in terms of a single value or

Acknowledgments

This work was supported by a grant from the VINNOVA Swedish Governmental Agency for Innovation Systems (2009-02766). The 2nd Strategic Highway Research Program (SHRP 2) Naturalistic Driving Study data were provided under Project 8A, Analysis of Naturalistic Driving Study Data: Safer Glances, Driver Inattention, and Crash Risk, Transportation Research Board of the National Academies of Science (2014). The findings and conclusions of this paper are those of the authors and do not necessarily

References (75)

  • J. Engström et al.

    Effects of working memory load and repeated scenario exposure on emergency braking performance

    Hum. Factors

    (2010)
  • J. Engström et al.

    Attention selection and multitasking in everyday driving: a conceptual model

  • J. Engström et al.

    Analysis of the role of inattention in road crashes based on naturalistic on-board safety monitoring data

  • Engström, J., 2010. Scenario criticality determines the effects of working memory load on brake response time. In: J....
  • C. Erbsmehl et al.

    Statistical driver model for accident simulation: using a statistical driver model for benefit estimation of advanced safety systems with warning interfaces

  • B.R. Fajen

    Calibration, information, and control strategies for braking to avoid a collision

    J. Exp. Psychol. Hum. Percept. Perform.

    (2005)
  • B.R. Fajen

    Perceptual learning and the visual control of braking

    Percept. Psychophys.

    (2008)
  • B.R. Fajen et al.

    Learning to Control Collisions: The Role of Perceptual Attunement and Action Boundaries

    J. Exp. Psychol.: Hum. Percept. Perform.

    (2006)
  • D.B. Fambro et al.

    Driver perception-brake response in stopping sight distance situations

    Transp. Res. Rec.

    (1998)
  • Fambro, D. B., Fitzpatrick, K., Koppa, R. J., 2000. New stopping sight distance model for use in highway geometric...
  • Fambro, D. B., Koppa, R. J., Picha, D. L., Fitzpatrick, K., 2000. Driver braking performance in stopping sight distance...
  • A. Field

    Discovering Statistics Using SPSS

    (2009)
  • G.M. Fitch et al.

    Driver braking performance to surprise and expected events

    Proceedings of the Human Factors and Ergonomics Society Annual Meeting

    (2010)
  • J.I. Gold et al.

    The neural basis of decision making

    Annu. Rev. Neurosci.

    (2007)
  • C. Gold et al.

    Take over! How long does it take to get the driver back into the loop?

    Roceedings of the Human Factors and Ergonomics Society 57th Annual Meeting

    (2013)
  • M. Green

    How long does it take to stop? Methodological analysis of driver perception-brake times

    Transp. Hum. Factors

    (2000)
  • M. Green

    Perception-reaction time: is olson & sivak all you need to know?

    Collision Int. Compend. Crash Res.

    (2009)
  • R.B. Heusser

    Heavy Truck Deceleration Rates as a Function of Brake Adjustment Technical Report 910126

    (1991)
  • R.S. Jurecki et al.

    Driver model for the analysis of pre-accident situations

    Veh. Syst. Dyn.

    (2009)
  • Jurecki, R. S., Stanczyk, T. L., 2014. Driver reaction time to lateral entering pedestrian in a simulated crash traffic...
  • R.J. Kiefer et al.

    Forward Collision Warning Requirements Project Final Report⿿Task 1. Technical Report DOT HS 809 574

    (2003)
  • S.G. Klauer et al.

    An Analysis of Driver Inattention Using a Case-Crossover Approach on 100-Car Data. Final Report DOT HS 811 334

    (2010)
  • T. Kondoh et al.

    Identification of visual cues and quantification of drivers⿿ perception of proximity risk to the lead vehicle in car-Following situations

    J. Mech. Syst. Transp. Logist.

    (2008)
  • T. Kondoh et al.

    Direct evidence of the inverse of TTC hypothesis for driver⿿s perception in car-closing situations

    Int. J. Automot. Eng.

    (2014)
  • K.D. Kusano et al.

    Safety benefits of forward collision warning, brake assist, and autonomous braking systems in rear-end collisions

    IEEE Trans. Intell. Transp. Syst.

    (2012)
  • K.D. Kusano et al.

    Population distributions of time to collision at brake application during car following from naturalistic driving data

    J. Saf. Res.

    (2015)
  • D. Lamble et al.

    Detection thresholds in car following situations and peripheral vision: implications for positioning of visually demanding in-car displays

    Ergonomics

    (1999)
  • Cited by (118)

    View all citing articles on Scopus
    View full text