Modelling Driver Distraction effect due to Mobile phone use on reaction time

• Objective is to examine the effect of distractions [Due to Phone use] on the

reaction time of driver on an Undivided 2 lane two-way road.

• 5 Different Driving Conditions

i. No phone use
ii. Simple Conversation: "Where did you go for last trip?"
iii. Complex conversation: Multiplication, addition, logical puzzle
iv. Simple Texting: "Whats your favourite Game?"
v. Complex texting: Address?

• 2 Hazardious Events
i. Pedestrial crossing: pedestrian will start crossing the road in front of subject
vehicle when it is 130m away.
ii. Parked Vehicle crossing: Vehicle will start crossing the road in order to join the
oncomming traffic stream in front of subject vehicle when it is 130m away

iii.

• Age Group:
i. Young: <30 [55%] [Avg. Exp.: 4.5 yrs]
ii. Mid age: 30 - 50 [39%] [Avg. Exp.: 11.9 yrs]
iii. old Age: >50 [6%] [Avg. Exp.: 34 yrs]

• Pressing of braking pedal was considered as the first reaction.

• Specification of Driving Simulator: Open Cab Simulator, 150 degree horizontal view
on three 42 inch LED, Cab with power steering(with active feedback), brake, gear
slector, turn signal.
• Test procedure: Simple road geometry and traffic conditions are designed in the
simulator using Simcreator and Simvista.
Drive of 3.5 Km with one hazardous event to measure driver rection time.
four drives were combined in a single closed loop road network. 100
participants of different age group participated in study out of which 27
were professional drivers. Order of presenting distraction and hazard event is
randomised.
• Result: Reaction time in case of pedestrian crossing event were increased by 40,
95, 137 and 204% in presence of simple conversation, complex conversation, simple
texting and complex testing respectively. In case of praking vehicle crossing event,
the reaction time increased by 48, 65, 121 and 171% in presence of simple
conversation, complex conversation, simple texting and complex testing
respectively.

• The 90th percentile value of Reaction time data was found to be more than 2.5 Sec
in all distrac ted condition.
• Drivers who are involved in talking during driving, reduce the visual scanning of the
peripheral envirnoment but focus more at the roadway to compensate for the
increased workload. This causes the delay in driver response to detect the vent
starting from the sides of the road for ex. pedestrian crossing.
 • The increment observed in terms of reaction time in vehicle crossing event is lesser
then pedestrian crossing event because driver find pedestrian crossing event more
sudden.
• Participant who had habit of recieving phone calls during driving perform better
than the others.
Mean reaction time of driver: For No phone condition: 0.99 Sec; Simple
conversation: 1.49 Sec; Complex conversation; 1.84 Sec; Simple Texting: 2.56 Sec;
Complex Texting: 2.72 Sec

Modelling of Reaction time.

Modelling is done using Survival Analysis
Survival analysis is a set a statistical tool to address the questions such as " How long
would it be, before a particular event occur" in other word "time to event" .

Event in this case event is the "Detection of pedestrian or Vehicle"

Duration Variable: It is the time elapsed until an event happens. i.e. Reaction time.
Reaction time is a continous random variable for which Probability distribution function
and cumulative distribution function are commonly used to characterized the distribution.

Survival Function: It is the probability of duration variable beeing greater than some
specified time t. It is the probability of not detecting the hazardous event (pedestrian
crossing & Parked vehicle).
S(t)= P(T>t) = 1-F(t)

Hazard Rate: conditional probablity of that event will occur between time t and t + dt.
h(t) gives the rate at which event is ending at time t, given that the event duration has
not ended up to time t.

As the time increases, the increase in probability of detecting the pedestrian is also
increases.

For accounting the influence of covariates(variables) on the probability, 2 methods are

used: 1. Proportional Hazard; 2. Accelerated lifetime model (Accelerated failure model)
Proportional Hazard: this method assumes that the covariate, act multiplicatively to the
base line hazard function. For ex. Base line hazard function is denoted by h 0(t) and it is
assumed that covariates influence the base line hazard with the function EXP(BX), where B
is a vector of estimable parameter. thus hazard rate with covariate is given by:
h(t|X) = h0(t) EXP(BX)

Accelerated lifetime model (Accelerated failure model): Second approach is to assume

that the covariate rescale time directly in a baseline survival function.
S(t|X) = S0[t EXP(BX)]
which lead to Hazard function, given by:
h(t|X) = h0[t EXP(BX)]EXP(BX)

Exponential distribution

Log-logistic

Weibull Distribution

Survival rates is given by:

Here lamda is location and p is scale parameter.

P value: 2.97 for pedestrian crossing and 2.63 for road crossing event.

Now for different value of t, probability of not detecting the vent can be plotted from the
equation of survival rate.
Why only Simulator testing?
It is not safe to conduct distraction studies in the field, therefore large number of
researchers have used driving simulator. Moreover data collection and extraction is also
more precise.

How 5 distraction conditions and 2 hazard events are finalised?

5 Distraction conditions of conversation and texting are selected to reflect the true
behaviour of driver under low and high level of distraction.

User Survey:
T-test, Wilcoxon test, Contigency tyable method, ANOVA, Linear mixed model, Parametric
Duration model

Effect of texting on accident risk during a sudden hazardous event: Analysis of

predetection and Post detection Phases

• Objective is to determine the effect of texting age and experience of Driver

on the accident risk associated with sudden event. Also to compare the
effect of driving behaviour of young inexperienced and professional drivers
on risk during Predetection nand post detection phases of event.

• Age Group and Driver's Experience: [total 49 drivers]

i. Inexperienced Young Driver: with experience less than 4 years; age < 25 years
ii. Professional drivers: Driving experience > 5 years; Age > 25 years; Annual mileage
>15,000 Km
[Experience was determined based on previous literature review studies].

• 4 lane undivided Rural Highway (2 way traffic) with free flow condition with a
speed limit of 110 km/hr
 • 3 Different Driving Condition:

i. Base Line
ii. Short Text: Limited to 10 character
iii. Long Text: Longer than 10 character

• 2 Hazardious Events
i. Pedestrial crossing: pedestrial will start crossing the road in front of subject
vehicle when it is 130m away.
ii. Parked Vehicle crossing: Vehicle will start crossing the road in front of subject
vehicle when it is 130m away

• Predetection: Time lapsed until detection ( i.e. Time difference between the start
of the sudden event until the driver made the first response)
• Post detection Phase (From the end of predetection phase): Time lapsed in
responding to the sudden event (Time lapsed in reducing the speed to 5 Km/hr)

• Specification of Driving Simulator: Open Cab Simulator, 150 degree horizontal

view on three 42 inch LED, Cab with power steering(with active feedback), brake,
gear slector, turn signal.

• Results:

For younger driver, the likelyhood of being involved in an accident was 6.4
times greater than professional drivers.

Predetection Phase: A significant deterioration in reaction time was observed

in case of short and long texts. However, there were no significant difference
between short and long text condition in drivers reaction time. The deterioration in
reaction time was same for young and professional driver.
 For Post Detection Phase: In base line condition, both categories took same
time to reduce the speed . However, Professional drivers were able to reduce the
speed in relatively less time during texting as compared to young drivers. This is due
to the reason that young drivers approached the event at a higher speed than
the professional driver.

Approach Speed: Young drivers had 65-75% higher speed than the professional
drivers during texting task. Also speed reduction among young driver was
observed only while long texting task. Young drivers showed risky behaviour
while driving, which led to higher accident risk for young drivers.

Professional driver reduced their speed by 43 -50% while texting, however

young driver reduced their speed by 15% only while long texting.

Compensation stratergy adopted by young drivers to reduce the increased

workload (i.e. reducing the speed of vehicle while texting) was minimal as
compared to professional drivers.
 Texting while Driving increased the accident probability by 2 times for
professional drivers and 4 times for young drivers.

Effect of long text was higher as compared to short text.

1 sec Delay in reducing speed could lead to a 37% increase in accident risk.

Model Used: Accident probability was estimated by generalized linear mixed

model.
Analysis of Vehicle-Based Lateral performance measures during distracted
driving due to phone use

Objective: To examine the effect of different phone use conditions on the Vehicle
based performance parameter.

Vehicle Performance Parameters:

 Standard Deviation of Lane positioning (SDLP): lateral offset of vehicle

from the centre line (measured in mtrs)
 Number of Lane Excursion: Outer side of the vehicle went off the road.
Frequency of lane excursion made by a driver in each drive is counted as
number of lane excursion.
 Mean and Standard deviation of lateral acceleration: measured in m/s2
 Mean and Standard deviation in Steering angle: measured in degrees
 Steering wheel Reversal rates: This is the number of corrections made to
the direction of steering wheel movement per minute by the driver.

Secondary Task:

 No Phone Condition
 Simple conversation: Based on study
 Complex conversation: Based on study
 Simple texting: Based on study
 Complex texting: Based on study

Participants:

Young: Age < 30; Average Experience: 4.50 years

Mid Age: 30 < Age < 50; Average Experience: 11.92 years

Old age: Age > 50; Average Experience: 34.33 years

87% Male and 13% Female

36% frequent conversing category

Driving Environment: 4 lane Road (3.5 Mtr width) undivided highway with posted
speed of 110 Kmph. Drive comprise of a 3.5 Km straight section.

Driving Simulator: Open cab driving simulator with horizontal view of 150° with
seat, steering wheel, gear selector, accelerator, brake, speedometer, turn signals,
headlights switches and presence of rear and side view mirror. Sound system under
the floor pan in order to recreate vehicle noise and wind noise, Simulator run at the
update speed of 60 Hz.
Main limitation of simulator was the unavailability of vibration transducer in the floor
pan

Modelling Braking Behaviour and accident probability of drivers under

increasing time pressure conditions

 Introduction: Road traffic accident statistics, reason for accident, speed

thrilling behaviour, Late to work, Time pressure
 Objective: To analyse the braking behaviour and accident probability of
drivers under increasing time pressure. Comparison is made between
gender, driving profession, approach speed, age, driving history
 Two hazardous event:
1. Pedestrian crossing on a 4 lane undivided highway:
7 pedestrian were placed on left side in the direction of traffic with
speed from 0.63 m/s to 1.63 m/s. If the driver speed was less than
20m/s (72 Km/hr) then pedestrian speed was ranged between 0.63-
1.35 and if speed of vehicle is more than 20 m/s (72 Km/hr) then the
speed of pedestrian was kept between 1.0-1.63 m/s. pedestrian
started crossing the road when the vehicle was 1.20 m away from the
pedestrian.
How 120 m distance was selected: firstly a distance of 110m was
selected and pilot study was performed. It was observed that the
driver reached earlier at the point of event and crossed the
intersection point before all the pedestrian covered the whole road
width (14m).
For 130m distance it was observed that pedestrian covered ¾ lane
width with an opening of 3.5 m
Finally a distance of 120m was selected and it was observed that all
the pedestrian covered the whole road width providing no space for
vehicle to move forward thus driver has to stop and wait till the
pedestrian crossed the road.
2. Vehicle overtaking on a two lane undivided highway:
A jersey barrier is used as an obstacle in the direction of lane to
hinder free movement of vehicle. 4 Vehicles in the opposite lane
 started moving with the speed of 60 km/hr when the subject vehicle
is 90 m away from them. Event was designed to stop the subjected
vehicle at the beginning of jersey-barrier.
Scenario was designed considering: (1). that the subjected vehicle
must stop before the obstacle (2). The waiting time should not make
the driver frustrated
First vehicle was placed at the beginning of the obstacle and other
vehicles are placed by leaving a gap of 20 m. it was considered that
the subjected vehicle would reach the obstacle in 4-5 second after
observing the obstacle in the opposite lane. Last vehicle require
approximately 4.1 s to cross the obstacle.
 Pressure condition:
1. No time pressure
2. Low time pressure: 10% travel time was reduced from the time
required in base line condition
3. High time pressure: 20% travel time was reduced from the time
required in base line condition
 Brake paddle force: force applied by the driver on the brake paddle
 Brake to maximum brake: duration of time required by the driver to
achieve maximum deceleration from the beginning of the brake application
to the response to the occurrence of the event.
 Experimental Design: 6 Km long urban undivided road with 4 lane and 2
lane with posted speed of 60 km/hr and 30 km/hr. low traffic volume
 Pedestrian crossing and obstacle overtaking are considered as perilous
events and are extensively used in transportation safety research to
examine driver’s response and effectiveness of emergency steering evasions
(ESE) assistance system. Advance driver assistance system (ADAS).
 NASA TLX mental workload questionare
