This document summarizes research on the effects of texting while driving on driver glance patterns and vehicle lane position on horizontal curves. The research analyzed driver glance duration and frequency at a mobile phone, as well as standard deviation of lane position, for drivers in a simulator under control and texting treatment conditions on four curves. The results showed no significant differences in glance duration or lane position between curves for the texting drivers. However, texting did increase lane position variability and time spent looking away from the road compared to the control drivers. The conclusions were that texting impairs driving performance but not specifically on curves versus other road sections.
Presentation by Professor Oliver Carsten at ETSC event hosted at European Parliament, Brussels on 3 November 2014.
www.its.leeds.ac.uk/people/o.carsten
http://etsc.eu/fitting-safety-as-standard
Highway crash data with average of 39 thousand fatalities and 2.4 million nonfatal injuries per year have repetitive and predictable patterns, and may benefit from statistical predictive
models to enhance highway safety and operation efforts to reduce crash fatalities/injuries. Highway crashes have patterns that repeat over fixed periods of time within the data set for
crashes such as motorcycle, bicycles, pedestrians, nighttime, fixed object, weekend, and winter crashes. In some States, these crashes are weekly, monthly, or seasonally. Contributing
factors such as: age category, light condition, weather, weekday, underlying state of the economy, and others impact these variations.
Presentation by Professor Oliver Carsten at ETSC event hosted at European Parliament, Brussels on 3 November 2014.
www.its.leeds.ac.uk/people/o.carsten
http://etsc.eu/fitting-safety-as-standard
Highway crash data with average of 39 thousand fatalities and 2.4 million nonfatal injuries per year have repetitive and predictable patterns, and may benefit from statistical predictive
models to enhance highway safety and operation efforts to reduce crash fatalities/injuries. Highway crashes have patterns that repeat over fixed periods of time within the data set for
crashes such as motorcycle, bicycles, pedestrians, nighttime, fixed object, weekend, and winter crashes. In some States, these crashes are weekly, monthly, or seasonally. Contributing
factors such as: age category, light condition, weather, weekday, underlying state of the economy, and others impact these variations.
Highway crash data with average of 39 thousand fatalities and 2.4 million nonfatal injuries per year have repetitive and predictable patterns, and may benefit from statistical predictive models to enhance highway safety and operation efforts to reduce crash fatalities/injuries. Highway crashes have patterns that repeat over fixed periods of time within the data set for crashes such as motorcycle, bicycles, pedestrians, nighttime, fixed object, weekend, and winter crashes. In some States, these crashes are weekly, monthly, or seasonally. Contributing factors such as: age category, light condition, weather, weekday, underlying state of the economy, and others impact these variations.
Presentation on Spot Speed Study Analysis for the course CE 454nazifa tabassum
This presentation describes the process of Spot Speed Study Analysis, how it can be performed and how the findings from such studies can help to improve road design in urban areas.
Scientific evidence on road safety effects of section control and red light c...Charles. Goldenbeld
The presentation summarises scientific evidence on road safety benefits from red light cameras and section control. The major research and reviews up to 2013 are covered. Separate attention is given to Dutch experiences with red light cameras and section control.
Cities operate ambient air quality monitoring networks but often do not analyze and interpret the data. Data gets simply "stacked". Networks are not configured correctly capturing the data trends and monitoring objectives. This presentation provides guidance and uses Mumbai's ambient air quality data to illustrate application
The purpose of this section is to increase Council’s understanding of residents’ knowledge, attitudes, behaviour and perception of safety with regards to road use within the City of Stonnington. The results will assist Council’s Transport Department in future program planning and service delivery.
Breakout Session 9: Improving Safety through Enforcement
2015 Traffic Safety Conference
by Nicole Zanier, Technology Transfer Coordinator, ATLAS Center/University of Michigan Transportation Research Institute
[Amended upload]
Presented by PhD student Segun Aluko at UTSG2014.
www.its.leeds.ac.uk/people/s.aluko
www.utsg.net/web/uploads/UTSG%202014%20Newcastle%20Programme.pdf
The study aims to quantify how people make trade-offs to avoid junctions by taking additional time along routes with and without cycle facilities in the UK context. A video based stated preference survey and analysis is undertaken to investigate how people feel approaching junctions, determine the relative importance of the features of junctions, determine how cycle facilities compensate the exposure of right turn risks at junctions at the cost of additional time and identify the person type factors that also influence choice. Primary data is used for the study. The survey work for the primary data constituted a major part of the study.
A case study utilizing the Six Sigma data analysis toolkit to examine a 15.5-mile daily morning commute completed on bicycle. The case first explores the usage of control charts to examine the total completion time in addition to various waypoints along the route. It then utilizes hypothesis testing to attempt to prove if a statistically significant improvement has occurred. It then demonstrates a multifactor regression model to predict the time needed to traverse the route. Finally it does a cost comparison between cycling, taking the metro and driving to work.
Highway crash data with average of 39 thousand fatalities and 2.4 million nonfatal injuries per year have repetitive and predictable patterns, and may benefit from statistical predictive models to enhance highway safety and operation efforts to reduce crash fatalities/injuries. Highway crashes have patterns that repeat over fixed periods of time within the data set for crashes such as motorcycle, bicycles, pedestrians, nighttime, fixed object, weekend, and winter crashes. In some States, these crashes are weekly, monthly, or seasonally. Contributing factors such as: age category, light condition, weather, weekday, underlying state of the economy, and others impact these variations.
Presentation on Spot Speed Study Analysis for the course CE 454nazifa tabassum
This presentation describes the process of Spot Speed Study Analysis, how it can be performed and how the findings from such studies can help to improve road design in urban areas.
Scientific evidence on road safety effects of section control and red light c...Charles. Goldenbeld
The presentation summarises scientific evidence on road safety benefits from red light cameras and section control. The major research and reviews up to 2013 are covered. Separate attention is given to Dutch experiences with red light cameras and section control.
Cities operate ambient air quality monitoring networks but often do not analyze and interpret the data. Data gets simply "stacked". Networks are not configured correctly capturing the data trends and monitoring objectives. This presentation provides guidance and uses Mumbai's ambient air quality data to illustrate application
The purpose of this section is to increase Council’s understanding of residents’ knowledge, attitudes, behaviour and perception of safety with regards to road use within the City of Stonnington. The results will assist Council’s Transport Department in future program planning and service delivery.
Breakout Session 9: Improving Safety through Enforcement
2015 Traffic Safety Conference
by Nicole Zanier, Technology Transfer Coordinator, ATLAS Center/University of Michigan Transportation Research Institute
[Amended upload]
Presented by PhD student Segun Aluko at UTSG2014.
www.its.leeds.ac.uk/people/s.aluko
www.utsg.net/web/uploads/UTSG%202014%20Newcastle%20Programme.pdf
The study aims to quantify how people make trade-offs to avoid junctions by taking additional time along routes with and without cycle facilities in the UK context. A video based stated preference survey and analysis is undertaken to investigate how people feel approaching junctions, determine the relative importance of the features of junctions, determine how cycle facilities compensate the exposure of right turn risks at junctions at the cost of additional time and identify the person type factors that also influence choice. Primary data is used for the study. The survey work for the primary data constituted a major part of the study.
A case study utilizing the Six Sigma data analysis toolkit to examine a 15.5-mile daily morning commute completed on bicycle. The case first explores the usage of control charts to examine the total completion time in addition to various waypoints along the route. It then utilizes hypothesis testing to attempt to prove if a statistically significant improvement has occurred. It then demonstrates a multifactor regression model to predict the time needed to traverse the route. Finally it does a cost comparison between cycling, taking the metro and driving to work.
1. Influence of Texting on Driver Glance
Patterns and Vehicular Lane Position on
Horizontal Curves
Presented by:
Makenzie Ellett
Research Assistant
Oregon State University
School of Civil and Construction Engineering
March 20th, 2015
Forsyth
2. Background – Cell Phone Use
1
• The first cell phone, 1983
• 1985: 340,000 subscribers
• 2000: 100 million subscribers
• 94% of people in the US aged 16+
owned a mobile device in 2013
• The first text message, 1992
• “Merry Christmas”
• 1997, USA: 40,000 text messages/day
• 2012, USA: 6 billion text
messages/day
• Most popular cell phone feature
(CTIA, 2013)
Number of Cell Phone Subscribers in the United States
0
50
100
150
200
250
300
350
1980 1990 2000 2010 2020
NumberofSubscribers(in
millions)
Year
0
1
2
3
4
5
6
7
1995 2000 2005 2010 2015
NumberofTextMessages(billions)
Year
Number of Text Messages Sent Per Day in the United States
5. Background – Safety of Texting and Driving
4
• Risk of crash increases by 23.24
times (Olson et al., 2009)
• Conversing on a hand-held mobile
phone increases crash risk 1.04 times
• Texting is the most dangerous activity
while driving
• National Survey on Distracted
Driving Attitudes and Behaviors
• 32.9% believe there is no difference in
their driving
• 92.2% feel at least “somewhat
uncomfortable” when riding with a
driver who texts
Percentage of Population Observed
Manipulating Hand-Held Devices
(NOPUS)
Driver Type 2010 2011
All Drivers 0.90% 1.30%
Age 25-69 0.80% 1.10%
Age 16-24 1.50% 3.70%
6. Background – Legality of Texting while Driving
5
Laws Regarding Texting While Driving By StateLEGEND
No Ban
Total Ban (Primary Law)
Total Ban (Secondary Law)
Partial Ban (School Bus & Novice Drivers)
Partial Ban (Novice Drivers Only)
7. Literature Review – Glance Patterns
• The longer a driver’s eyes are away from the roadway, the greater
the odds ratios of a crash incident
• For an “incident” to occur, driver glances of 1.1 sec. were observed (Klauer et
al., 2006)
• Texting defined as a “complex, tertiary task” (Olson et al., 2009)6
Odds Ratios Associated with Eyes Off of the Forward Roadway (Klauer et al., 2006)
Total Eyes off Forward Roadway Odds
Ratios
Lower Control Limit
Upper Control
Limit
Time (seconds) (LCL) (UCL)
t ≤ 0.5 1.13 0.67 1.92
0.5 < t ≤ 1.0 1.12 0.79 1.59
1.0 < t ≤ 1.5 1.14 0.79 1.65
1.5 < t ≤ 2.0 1.41 0.98 2.04
t > 2.0 2.27 1.79 2.86
8. Literature Review – Lateral Position
• As distraction levels increase, the vehicle’s standard deviation of
lateral position (SDLP) also increases
• 70% increase in lane position variability compared to baseline (Hosking et al.,
2006)
• Lane excursions increase when texting (Reed et al., 2008)
• Reading: 8 to 18
• Writing: 4 to 42
7
Measuring Standard Deviation of Lateral Position (Verster et al., 2011)
9. Methodology - Research Hypotheses
H0: There is no difference in the duration of driver
fixations on a mobile phone while completing a text
messaging task between four horizontal curves.
8
? ? ?Curve 1 Curve 2 Curve 3 Curve 4
10. Methodology - Research Hypotheses
H0: There is no difference in the lateral position of a vehicle
between baseline driving and driving while completing a text
messaging task between four horizontal curves.
9
? ? ?Curve 1 Curve 2 Curve 3 Curve 4
(Larmoyeux & Bone)
11. Methodology - Research Hypotheses
H0: There is no difference in the lateral position of a
vehicle before, during, or after the text messaging task
between four horizontal curves.
10
? ? ?Curve 1 Curve 2 Curve 3 Curve 4
12. Methodology – Dependent Variables
• Glance frequency towards mobile phone
• Duration of glances towards mobile phone
• Percentage of time on curve subject’s eyes are on the mobile
phone
• SDLP of vehicle throughout curve
11
17. Methodology – Participants
• Data obtained from Joshua Swake, MS 13’ Thesis
• Texting while driving was used as a distractor for the original study
• Original research studied driver behavior in work zones
• Original Study: 36 participants
• Current Study: 18 participants
• Control Group: 4 subjects (did not text)
• Treatment Group: 14 subjects (responded to texting cues)
16
18. Results - Data Collection
17
Result
Data Collection
Method
Reduction of Data
Driver Glance Patterns Mobile Eye XG Videos
Researcher
Observation
Lateral Position of
Vehicle
OSU Driving Simulator CSV Files
19. Results - Analysis
• Paired T-test
• R-studio
• Adjusted for multiple comparisons with
the Benjamini and Yekutieli adjustment
• Statistically significant p-values < 0.05
• 95% confidence intervals
18
20. Results – Average Duration of Driver Fixations
19
Average duration of driver fixations
21. Results – Average Duration of Driver Fixations
20
Average duration of driver fixations
Average Duration of Driver Fixations (sec)
Curve
Paired T-test
Curve 1 Curve 2 Curve 3 Curve 4 P-value Significant
1.078 1.091 1.090 1.146
1 v 2 0.7311 No
1 v 3 0.6817 No
1 v 4 0.8329 No
2 v 3 0.9922 No
2 v 4 0.5374 No
3 v 4 0.3525 No
22. Results – Maximum Duration of Driver Fixations
21
Maximum duration of driver fixations
23. Results – Maximum Duration of Driver Fixations
22
Statistical summary comparing maximum duration of fixations between curves
Maximum Duration of Driver Fixations (sec)
Curve
Paired T-test
Curve 1 Curve 2 Curve 3 Curve 4 P-value Significant
4.04 2.54 2.61 2.87
1 v 2 0.1953 No
1 v 3 0.2701 No
1 v 4 0.1983 No
2 v 3 0.5397 No
2 v 4 0.4081 No
3 v 4 0.7664 No
24. Results – Percentage of Time with Eyes Off Roadway
23
Percentage of time with eyes off roadway
25. Results – Average Percentage of Time with Eyes Off Roadway
24
Statistical summary of average percentage of time with eyes off roadway
Average Percentage of Eyes off Forward Roadway
Curve
Paired T-test
Curve 1 Curve 2 Curve 3 Curve 4 P-value Significant
30.2 20.1 27 24.7
1 v 2 0.06212 Suggestive
1 v 3 0.5885 No
1 v 4 0.2548 No
2 v 3 0.06371 Suggestive
2 v 4 0.05607 Suggestive
3 v 4 0.5473 No
26. Results – Average Overall SDLP
25
Overall SDLP for control condition Overall SDLP for treatment condition
27. Results – Average Overall SDLP
26
Average overall SDLP for control and treatment conditions
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
2.00
Curve 1 Curve 2 Curve 3 Curve 4
SDLP(ft)
Control
Treatment
28. Results – Average Overall SDLP
27
Statistical summary of average overall SDLP for control condition
Average SDLP of Control Condition
Curve
Paired T-test
Curve 1 Curve 2 Curve 3 Curve 4 P-value Significant
1 1.19 1.06 1.05
1 v 2 0.26 No
1 v 3 0.60 No
1 v 4 0.12 No
2 v 3 0.49 No
2 v 4 0.36 No
3 v 4 0.94 No
29. Results – Average Overall SDLP
28
Statistical summary of average overall SDLP for treatment condition
Average SDLP of Treatment Condition
Curve
Paired T-test
Curve 1 Curve 2 Curve 3 Curve 4 P-value Significant
1.77 1.29 1.25 1.26
1 v 2 0.10 No
1 v 3 0.16 No
1 v 4 0.13 No
2 v 3 0.80 No
2 v 4 0.78 No
3 v 4 0.94 No
30. Results – Average Overall SDLP Comparison
29
0
5
10
15
20
25
250 260 270 280 290 300
LanePosition(ft)
Video Time (s)
Control Subject - Curve 1
0
5
10
15
20
25
230 235 240 245 250 255 260 265
LanePosition(ft)
Video Time (s)
Treatment Subject - Curve 1
Comparison of control and treatment subjects’ SDLP
31. Results – Average Interval SDLP
30
SDLP for before interval SDLP for during interval
SDLP for after interval
32. Results – Average Interval SDLP
31
Average overall SDLP for control and treatment conditions
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1 2 3 4
SDLP(ft)
Curve
Before
During
After
33. Results – Average Interval SDLP
32ns
Statistical summary of average SDLP for before interval
Average SDLP of before Period
Curve
Paired T-test
Curve 1 Curve 2 Curve 3 Curve 4 P-value Significant
0.6880 0.5473 0.5652 0.6646
1 v 2 0.1678 No
1 v 3 0.2738 No
1 v 4 0.7468 No
2 v 3 0.9487 No
2 v 4 0.2346 No
3 v 4 0.1675 No
34. Results – Average Interval SDLP
33
Statistical summary of average SDLP for during interval
Average SDLP of during Period
Curve
Paired T-test
Curve 1 Curve 2 Curve 3 Curve 4 P-value Significant
1.1948 1.0280 1.1671 1.0826
1 v 2 0.5976 No
1 v 3 0.7777 No
1 v 4 0.5857 No
2 v 3 0.3269 No
2 v 4 0.7846 No
3 v 4 0.6845 No
35. Results – Average Interval SDLP
34
Statistical summary of average SDLP for after interval
Average SDLP of after Period
Curve
Paired T-test
Curve 1 Curve 2 Curve 3 Curve 4 P-value Significant
1.0918 0.9341 0.8377 0.8057
1 v 2 0.7251 No
1 v 3 0.6745 No
1 v 4 0.4962 No
2 v 3 0.6771 No
2 v 4 0.2655 No
3 v 4 0.8039 No
36. Conclusions – Duration and Frequency of Fixations
H0: There is no difference in the duration of driver fixations on a mobile
phone while completing a text messaging task between four horizontal
curves.
H0 is not rejected
• No statistically significant differences were found between the fixation
durations
• No statistically significant difference was found between the maximum
fixation durations
35
37. Conclusions –SDLP of Treatment and Control Groups
H0: There is no difference in the lateral position of a vehicle between
baseline driving and driving while completing a text messaging task
between four horizontal curves.
H0 is not rejected
• No statistical difference was found in the average SDLP of the treatment
group
• No statistical difference was found in the average SDLP of the control group
• Treatment group exhibited increased SDLP compared to control group on
all four curves
36
38. Conclusions – SDLP of Before, During, & After Intervals
H0: There is no difference in the lateral position of the vehicle before,
during, or after the text messaging task between four horizontal
curves.
H0 is not rejected
• No statistically significant difference was found in the average SDLP of the
before intervals
• No statistically significant difference was found in the average SDLP of the
during intervals
• No statistically significant difference was found in the average SDLP of the after
intervals
• Average SDLP was least for before interval on all four curves
• Average SDLP was greatest for during interval on all four curves
• Average SDLP was noticeably increased during after interval, compared to
before interval
37
39. • A larger, more diverse sample size could result in more specific
conclusions relating the effects of age, gender, and driving
experience
• A larger sample size could result in statistical conclusions being
drawn between the control and treatment groups
• Analysis on the addition of ambient traffic
• Varying the text messaging cues by category, complexity, or prompt-
type to see their effects on driver behavior
• Direct comparison of SDLP and glance patterns of texting on
horizontal curves and tangent sections
38
Future Work
40. • Dr. David Hurwitz
• Justin Neill
• Joshua Swake
• OSU Transportation Department
• OSU Honors College
39
Acknowledgements