Proceedings of 8th Transport Research Arena TRA 2020, April 27-30, 2020, Helsinki, Finland
Reaction Time Variability Association with Unsafe Driving
M. G. Altarabichi
a*, M. U. Ahmed
a, M. R. Ciceri
b,
S. Balzarotti
b, F. Biassoni
b, D. Lombardi
b, P. Perego
baSchool of Innovation Design and Engineering (IDT), Mälardalen University, Västerås, Sweden bUniversità Cattolica del Sacro Cuore, L.go Gemelli, 1, 20123 Italy
Abstract
This paper investigates several human factors including visual field, reaction speed, driving behavior and
personality traits based on results of a cognitive assessment test targeting drivers in a Naturalistic Driving Study
(NDS). Frequency of being involved in Near Miss event (fnm)and Frequency of committing Traffic Violation (ftv)
are defined as indexes of safe driving in this work. Inference of association shows statistically significant correlation between Standard Deviation of Reaction Time (σRT) and both safe driving indexes fnm and ftv. Causal
relationship analysis excludes age as confounding factor as variations in behavioral responses is observed in both younger and older drivers of this study.
Keywords: Road Safety, Naturalistic Driving, Vienna Test, Cognitive Assessment, Reaction Time Variability.
* Corresponding author. Tel.: +46-76-4291956; E-mail address: mohammed.ghaith.altarabichi@mdh.se
1 1. Introduction and Related Work
Naturalistic Driving Studies (NDS) investigates real traffic situations by observing drivers in their natural settings. This type of applied traffic research provides useful insights about critical driving events like accidents and near misses (Muronga K, 2017; Feng G, 2019; Barnard, Y, et al., 2016). Expert analysis of observed driving events is carried out to investigate factors that contribute to unsafe driving behaviour. Factors could be broadly categorized into external and internal factors. Examples of external factors include environment characteristics (e.g. bumpy road, night driving) and event characteristics (e.g. roundabout, emergency braking) (Altarabichi et al, 2019). While internal refers to human factors like driver status (e.g. emotional state, attention) and driving performance (e.g. committing traffic violation, exploration of visual field).
As found in (Elander et al. 1993; Feyer et al. 1997) around 90% of road-traffic crashes are caused by driver error/human factors (i.e. as inattention, loss of vigilance, mental under/overload) and unsafe behavior (i.e. inadequate training or lack of experience). Several human factors contribute to a higher risk of traffic accidents as observed in the literature (Ahmed et al. 2017). Driving skills and driver’s attitude toward safety are associated with a higher risk levels of accidents as mentioned in (Quimby A.R. and Watts G.R., 1981). In a literature review on cognitive, sensory, motor and physical factors (Anstey K. J., et al., 2005.) Anstey highlights correlations between attention, reaction time, memory, executive function, mental status, visual function, and physical function variables with driving outcome measures. Several researches studied additional factors that might contribute to a higher risk level like sleeping disorders (Fairclough S.H. and Graham R., 1999), alcohol (De W D, and Brookhuis K A, 1991), aging (Mori Y and Mizohata M, 1995) and distractions (Horberry et al., 2006). Again, authors in (Petridou E. and Moustaki M., 2000) classified behavioral factors into factors that reduces capability on a long-term basis (e.g. diseases), short-long-term basis (e.g. fatigue), long-long-term impact (e.g. overestimation of capabilities) and short-term impact (e.g. suicidal behavior). Authors in (Wood J M, et al. 2008) proposed an approach based on multidisciplinary tests to predict unsafe driving of elderly, while (Johansson G. and Rumar K., 1971) investigated distribution of brake reaction of drivers in unexpected traffic situations.
This study is established based on NDS data of SimuSafe* Project. Recorded data part of NDS cycle of the project
varied in nature and was originated from multiple sources. For example, cognitive assessment was conducted to evaluate variety of human factors of participating subjects. Vehicular sensors were used to register different aspects of driving dynamics and motion (e.g. Inertial Measurement Unit IMU). Road cameras were used to capture driving events and the surrounding environment. While in-vehicle cameras were utilized to record driver’s status for further evaluation of behaviour and emotional state by experts in psychology. The variety in data sources encouraged the use of causal inference and data mining techniques to generate useful insights from data. The derived knowledge could be then utilized to tune driving functions and update safety regulations.
Primarily, this analysis inspects the roles and associations of human factors with safe driving behavior by analyzing the cognitive assessment results part of SimuSafe NDS. The major contribution of this work demonstrates correlation between Standard Deviation of Reaction Time (
σ
RT) and safe driving indexes based on the observedpopulation data.
2. Materials and Method
The analysis of this work is performed using cognitive assessment tests results conducted on volunteers’ part of SimuSafe project. The performed tests aim to evaluate several cognitive aspects of participants including visual field, reaction speed, driving behavior and personality traits. Table 1 provides information about conducted cognitive tests.
These cognitive assessments took place prior to the volunteers’ participation in NDS cycle of SimuSafe between May and August of 2018. Video analysis were conducted by psychologists to investigate events of interest that occurred during NDS. A total of 7173 videos that correspond to 2040 events were viewed and studied by the psychologists. 565 events of interest were identified that represents 280 roundabouts, 171 intersections, 53 aggressive acceleration and 75 emergency braking. The psychologists identified a total of 26 near miss events and 133 traffic violations committed by the participating 15 drivers, with no accidents recorded during the NDS cycle.
2
Table 1. Cognitive assessments conducted on volunteers who participated in SimuSafe project.
# Cognitive Test Description
1 Active Visual Field † Detection of visual stimuli presented in peripheral vision while performing a central vision task.
2
Vienna Test – Adaptive Tachistoscopic Traffic Perception test ‡
Recognition of traffic elements (pedestrian, vehicles, bicycles, road signs, traffic lights) after brief presentation of road pictures.
3 Vienna Test – Determination test §
Response to rapidly changing visual and acoustic stimuli by pressing appropriate buttons and foot pedals. 4 Vienna Test – Reaction test ** Go/no go reaction time.
5 Vienna Test – Aggressive Driving Behaviour ††
Using an eight-point answer scale, respondents indicate how often they exhibit a particular behaviour when driving.
6
Vienna Test – Inventory of Driving-Related Personality Traits ‡‡
Using a free scale (agree-don’t agree), respondents rate the extent to which 48 concrete statements about driving, leisure, and work apply to them.
Near miss is by definition an event that had potential to cause but doesn’t actually result in human injury. Based on the results coming from the video analysis the Frequency of Near Miss events (fnm) was calculated and identified
as the first metric of safe driving according to equation (1):
𝑓nm
=
𝐸nm𝐸all (1)
Where fnm is the frequency of near miss events of the observed driver,
𝐸
nm is the count of near miss events,𝐸
allis the count of overall viewed events of the same driver. A lower value of fnm indicates safer driving behavior.
While a higher value suggests that the driver is more likely to be involved in risky events based on the observed NDS videos. The second safe driving metric, Frequency of committing Traffic Violation (ftv) is calculated
according to equation (2):
𝑓tv
=
𝐸tv𝐸all (2)
Where ftv is defined as the frequency of committing traffic violations by the observed driver,
𝐸
tv is the count oftraffic violations committed by the driver. A lower value of ftv indicates safer driving behavior. While a higher
value suggests that the driver is more likely to commit traffic violations based on the observed NDS videos. The psychologists identified 8 types of traffic violations,
𝐸
tv of a driver is increased by one for committing any of the below violations:• The driver didn’t use indicators.
• The driver didn’t explore the visual field. • The driver ignored side mirrors.
• The driver ignored rear view mirror. • The driver didn’t maintain safety distance. • The driver didn’t give way to others. • The driver ignored road signs. • The driver performed a heavy braking.
† Psytest Active Visual Field (https://www.psytest.net/index.php?page=Aktives-Gesichtsfeld&hl=en_US).
‡ Schuhfried Vienna Test - ATAVT Adaptive Tachistoscopic Traffic Perception (https://www.schuhfried.com/test/ATAVT). § Schuhfried Vienna Test - Determination Test (https://www.schuhfried.com/test/DT).
** Schuhfried Vienna Test - Reaction Test (https://www.schuhfried.com/test/RT).
†† Schuhfried Vienna Test - AVIS Aggressive Driving Behaviour (https://www.schuhfried.com/test/AVIS).
3 The evaluation in this work aims to identify important measured variables of cognitive assessment tests. These measurements could be utilized as an independent variable to reliably predict the dependent variables fnm and ftv as
indexes of safe driving performance.
Pearson correlation coefficient is used as a measure to identify linear relation between independent variables identified from cognitive tests and metrics of safe driving as dependent variables. Correlation is reported along with the P-value to indicate the probability of an uncorrelated system producing datasets that have a Pearson correlation at least as extreme as the one computed from these tests. A (P-value=0.05) is identified as cut-off to identify statistically significant correlations. Scatter plots of variables were examined visually according to (Anscombe, 2010) to graph the distribution and account for the effect of outliers on statistical properties.
3. Results and Discussion
As observed from Table 2, variables associated with Reaction Time (RT) of volunteers are showing the highest linear correlation and the lowest P-values of all tested variables. Slower Mean of Reaction Time (𝑅𝑇)are strongly correlated with fnm. The fact that three independent tests (Reaction, Determination and Active Visual tests) returned
statistically significant results of variables measuring RT of respondent confirms validity of this research results that is aligned with the literature (Đurić P and Filipović D, 2009). Table 2 shows that Standard Deviation of
Reaction Time (
σ
RT) measured in Reaction Test is significantly correlated with both metrics of safe driving fnm andftv. This finding highlights the significance of variations in the behavioural responses of the respondent.
Fluctuations represented by the value of
σ
RT are statistically correlated with increased levels of unsafe drivingbehaviour.
Table 2. Correlation analysis results ranked by P-value of different variables of cognitive assessment tests.
Near Miss fnm
# Variable Test Correlation P-Value
1 𝑅𝑇 Reaction 0.78 0.002
2 𝑅𝑇 Determination 0.64 0.02
3 σRT Reaction 0.58 0.04
4 Delayed responses for upper left
side of visual field Active Visual
0.57
0.04
Violations ftv
# Variable Test Correlation P-Value
1 σRT Reaction 0.73 0.004
In order to analyse causal relationship, age of participants is validated as a potential confounding factor that might be influencing both variables 𝑅𝑇,
σ
RT and safe driving indexes fnm and ftv. The correlation matrix in Figure 1confirms age is linearly correlated with higher 𝑅𝑇 (P-value= 0.005). The correlation between age and
σ
RT is alsopositive, which is aligned with results of (Hultsch D F, et al., 2002) that found greater RT variability in older subjects when compared to younger ones. However, the correlation between age and safety indexes are statistically insignificant fnm (P-value= 0.14) and ftv (P-value= 0.95). This result is aligned with (Ryan G A, et al. 1998; Guo,
F, et al. 2017) findings which confirms young adults along with elderly as two risk groups. Thus, the analysis confirms age as a causing factor of higher 𝑅𝑇 and
σ
RT. However, variations in behavioural responses are observedin both younger and older drivers of this study.
The association of fluctuations in responses with an increased level of unsafe driving behaviour may have an intuitive explanation behind it. This uncertainty complicates the process of planning actions in response to external events. An inconsistent driver occasionally acts quickly in response to certain events, but not as sharp for others. A driver with a slow consistent reaction on the other hand could still accommodate a driving style that fit slower, yet predictable responses (e.g. by planning actions ahead, driving slowly).
4 Fig. 1. Correlation matrix of driver age, fnm and ftv, 𝑅𝑇 and
σ
RTResults presented in Table 2 suggest that variables of the two Vienna tests measuring aggressive driving behaviour and driving related personality traits didn’t show significant results based on the observed population
.
It is necessary to investigate if they are related to other indexes of unsafe driving behaviour (e.g. frequency of braking and/or heavy braking) or to specific risky events (e.g. inattention).4. Acknowledgements
The authors would like to acknowledge the project SimuSafe, the project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 723386.
5. References
Ahmed M.U, Begum S., Catalina C. A, Limonad L., Hök B., Flumeri G. D, Cloud-based Data Analytics on Human Factor Measurement to Improve Safer Transport (Nov 2017), 4th EAI International Conference on IoT
Technologies for HealthCare (HealthyIoT'17)
Altarabichi M. G, Ahmed M. U, Begum S., Supervised Learning to Identify Roundabouts Using IMU
(Mar 2019), First International Conference on Advances in Signal Processing and Artificial Intelligence (ASPAI' 2019)
Anscombe, F. (1973). Graphs in Statistical Analysis. The American Statistician, 27(1), 17-21. doi:10.2307/2682899
Anstey, K.J., Wood, J., Lord, S. and Walker, J.G., 2005. Cognitive, sensory and physical factors enabling driving safety in older adults. Clinical psychology review, 25(1), pp.45-65.
Barnard, Y., Utesch, F., van Nes, N. et al, The study design of UDRIVE: the naturalistic driving study across Europe for cars, trucks and scooters, Eur. Transp. Res. Rev. (2016) 8: 14.
De Waard, D. and Brookhuis, K.A., 1991. Assessing driver status: a demonstration experiment on the road. Accident analysis & prevention, 23(4), pp.297-307.
Đurić, P. and Filipović, D., 2009. Reaction time of drivers who caused road traffic accidents. Medicinski pregled, 62(3-4), pp.114-119.
Elander, J., West, R, French D, Behavioural correlates of individual differences in road traffic crash risk: An examination of methods and findings, Psychological bulletin 113.2 (1993): 279
Fairclough, S.H. and Graham, R., 1999. Impairment of driving performance caused by sleep deprivation or alcohol: a comparative study. Human factors, 41(1), pp.118-128.
5 Feng G, Statistical Methods for Naturalistic Driving Studies, Annual Review of Statistics and Its Application, Vol. 6:309-328, March 2019
Feyer, A.M., Williamson, A. & Friswell, R., Balancing work and rest to combat driver fatigue: an investigation of two-up driving in Australia. Accident Analysis and Prevention 1997 Jul, 29,541–553.
Guo, F., Klauer, S.G., Fang, Y., Hankey, J.M., Antin, J.F., Perez, M.A., Lee, S.E. and Dingus, T.A., 2017. The effects of age on crash risk associated with driver distraction. International journal of epidemiology, 46(1), pp.258- 265.
Horberry, T., Anderson, J., Regan, M.A., Triggs, T.J. and Brown, J., 2006. Driver distraction: The effects of concurrent in-vehicle tasks, road environment complexity and age on driving performance. Accident Analysis & Prevention, 38(1), pp.185-191.
Hultsch, D.F., MacDonald, S.W. and Dixon, R.A., 2002. Variability in reaction time performance of younger and older adults. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 57(2), pp. P101- P115.
Johansson, G. and Rumar, K., 1971. Drivers' brake reaction times. Human factors, 13(1), pp.23-27.
Mori, Y. and Mizohata, M., 1995. Characteristics of older road users and their effect on road safety. Accident Analysis & Prevention, 27(3), pp.391-404.
Muronga, K, The effectiveness of the naturalistic driving studies in improving driver behavior, Submitted as a Mini-Dissertation and Partial Requirement for The Degree: Magister Technologiae: Business Information Systems, Tshwane University of Technology, 2017
Petridou, E. and Moustaki, M., 2000. Human factors in the causation of road traffic crashes. European journal of epidemiology, 16(9), pp.819-826.
Quimby, A.R. and Watts, G.R., 1981. Human factors and driving performance (No. LR 1004 Monograph). Ryan, G.A., Legge, M. and Rosman, D., 1998. Age related changes in drivers' crash risk and crash type. Accident Analysis & Prevention, 30(3), pp.379-387.
Wood, J.M., Anstey, K.J., Kerr, G.K., Lacherez, P.F. and Lord, S., 2008. A multidomain approach for predicting older driver safety under in‐traffic road conditions. Journal of the American Geriatrics Society, 56(6), pp.986 993.
