1. Effects of an Aftermarket Crash Avoidance System on Warning Rates and Driver Acceptance in Urban and Rural Environments
ADTSEA 2019
Burlington VT
July 22, 2019
Ian J. Reagan, Ph.D.
Senior Research Scientist

2. Vehicle Automation and Crash Avoidance Features
Lane Maintenance
Lane Departure Warning (LDW)
Lane Keep Assist
Front Crash Prevention
Forward Collision Warning (FCW)
Automatic Emergency Braking (AEB)
Adaptive Cruise Control (ACC)
Blind Spot Monitoring / Park Assist
Rear cross traffic alert
Rear crash prevention
Fatigue Alert
Electronic Stability Control

3. Crash avoidance technologies are working
Effects on relevant police-reported crash types
all severities
injury
statistically significant
FCW, AEB: front-to-rear strikes
LDW: single-vehicle, head-on, sideswipes
Blind spot: lane-change crashes

4. New vehicle series with forward collision warning
By model year

5. Registered vehicles with forward collision warning
By calendar year

6. Estimated registered vehicles by feature
Calendar years 2016 and 2021

7. Aftermarket crash avoidance systems
Mobileye-630 crash avoidance technologies
Telematics
record of warnings issued, speed, posted limit, GPS, hard accelerations
forward collision warning (FCW) and urban forward collision warning (UFCW)
lane departure warning (LDW)
headway monitoring (HWM)
pedestrian collision warning (PCW)
speed limit information (SLI)
Mobileye-630 crash avoidance technologies
Compatible with nearly all MY1994 or newer vehicles in US market
FCW and LDW passed confirmation tests included in federal New Car Assessment Program (NHTSA, 2012)
Other features: speed limit information display, intelligent high beam assist
Paired with telematics: record of warnings issued, speed, posted limit, GPS, hard accelerations

8. 22 participants w/ Mobileye
Study Methods
22 participants w/ Mobileye
15 also with telematics
Independent variables
Primary roadway type:
urban (Arlington, VA) and rural (Ruckersville, VA)
Study period:
Mobileye data (baseline, treatment);
self-report data (week 2 and week 8 surveys)
Mobileye-630 aftermarket systems and telematics installed in IIHS employees’ vehicles
103 employees invited
22 employees volunteered for Mobileye
17 also agreed to telematics installation
Independent variables
Office location: urban (Arlington, VA) and rural (Ruckersville, VA)
Study period: Mobileye data (baseline, treatment); self-report data (week 2 and week 8 post-treatment surveys)
Hypotheses
Rates would decrease from baseline to treatment and would be lower for Ruckersville than Arlington drivers
Perceived usefulness would be lower and annoyance would be higher at week 8 compared with week 2.

9. Study timeline and procedure
1 week washout
4 week baseline, alerts silent
8 week treatment, alerts active
week 2 post-activation survey
week 8 post-activation survey
installations, May 2017
alerts activated
Warning thresholds and volumes locked for duration of study
Week 2 and week 8 post-activation surveys
10-point Likert-scale statements about the usefulness, criticality, urgency, and annoyance of the FCW, LDW, HMW, PCW, and SLI warning displays
Desire to disable warnings, self-reported behavioral change

10. Forward collision warning
study period: F(1,13)=6.01, p=0.03
location: F(1,13)=5.06, p=0.04
interaction: F(1,13)=0.34, p=0.57
Mean warning rate per 100 miles driven
Mean FCW rates decreased at both locations during treatment relative to baseline. Controlling for study period, FCW rates were lower at the rural location.

11. Forward collision warning
study period: F(1,13)=6.01, p=0.03
location: F(1,13)=5.06, p=0.04
interaction: F(1,13)=0.34, p=0.57
Mean warning rate per 100 miles driven
Mean FCW rates decreased at both locations during treatment relative to baseline. Controlling for study period, FCW rates were lower at the rural location.

12. Headway monitoring warnings (HMW)
study period: F(1,13)=7.30, p=0.02
location: F(1,13)=7.55, p=0.02
interaction: F(1,13)=1.96, p=0.18
Mean warning rate per 100 miles driven
Mean HMW rates decreased at both locations during treatment relative to baseline. Controlling for study period, HMW rates were lower at the rural location.

13. Lane departure warning (LDW)
Mean warning rate per 100 miles driven
During treatment, mean lane departure warning rates decreased significantly. Controlling for study period, lane departure warning rates were lower in the rural location but this effect only approached significance.
study period: F(1,13)=4.67, p=0.05
location: F(1,13)=3.80, p=0.07
interaction: F(1,13)=2.04, p=0.18

14. Mobileye safety score Rural Urban study period: F(1,13)=7.35, p=0.02
location: F(1,13)=6.10, p=0.03
interaction: F(1,13)=2.53, p=0.14
Rural
Urban
Mean safety scores increased significantly during treatment, while holding the effect of location constant. Across the study period, mean safety scores were lower in the urban location.

15. Usefulness of Mobileye Features*
Mean ratings showed a small, consistent trend of decreased usefulness of the different warnings at week 8. No significant differences in subjective ratings due to office location.
Effects for study period:
FCW : F(1,13)=0.65, p=0.43
HMW: F(1,19)=0.99, p=0.33
LDW: F(1,19)=4.84, p=0.04
SLI: F(1,19)=2.97, p=0.10
PCW: F(1,19)=0.13, p=0.73

16. Criticality, urgency, and annoyance for FCW, LDW, and HMW
Mean Rating
Situations that led to FCW were rated to be more critical than those for LDW and HMW
FCW had higher mean ratings for urgency of the alert at week 2 and week 8
HMW had highest mean ratings of annoyance at week 2 and week 8
HMW criticality and urgency and LDW annoyance omitted from week 2 survey

17. Self-report experience (agree/disagree)
Would like to turn off (FCW or LDW)
Increased following distance because of HMW
Reduced time spent driving over the posted limit because of SLI
Mobileye helped improve safety while driving

18. Self-report experience (agree/disagree)
Would like to turn off (FCW or LDW)
FCW: 84 percent disagreed, LDW: 81 percent disagreed
Increased following distance because of HMW
48 percent disagreed, 48 percent agreed
Reduced time spent driving over the posted limit because of SLI
67 percent disagreed, 15 percent agreed
Mobileye helped improve safety while driving
19 percent disagreed, 67 percent agreed

19. Conclusion
Warning-triggering behaviors less frequent when warnings present
The warnings work!
Roadway conditions modify driving risk, change possible with feature
Research need: Effects of age?
Robust reduction in warning rates for vehicles with telematics
30-70 percent lower in treatment than baseline, depending on location and warning system
Compelling changes given a sample of employees who work for a highway safety organization
Effect of office location suggests adjusting expectations for behavior change to account for regional differences (traffic density, geography)
Safety score of rural Ruckersville drivers at baseline (84 percent) exceeded Mobileye’s targeted score for treatment (80 percent)
Different systems may be more relevant in certain contexts, e.g., lane departure crashes more common in less dense areas
Research need: compare crash rates of vehicles with aftermarket systems with and without telematics

20. Ian J. Reagan, Ph.D.
Senior research scientist