Driver behavior analysis on rural 2-lane, 2-way highways using SHRP 2 NDS data

Abstract

Objective: Rural 2-lane, 2-way (RTLTW) highways comprise a large portion of the rural road network in the United States. Crashes on RTLTW roads have been a major concern for traffic safety and may impact traffic significantly. Driver behavior on RTLTW roads has not been analyzed thoroughly. This article analyzed the role of driver behavior fault in crashes/near-crashes on RTLTW roads.

Methods: A total of 467 crash/near-crash records and 160 baseline events (normal driving) were extracted from the Strategic Highway Research Program 2 Naturalistic Driving Study (NDS) database. The contribution of driver behavior to crash risk compared to roadway geometry and environmental factors was ranked using random forest. The influence of different factors on driver behavior fault in a crash/near-crash was further examined using a probit model. The reliability of the model was evaluated using 3 different parameters.

Results: It was found that driver behavior fault was the most important factor in crashes that occurred on RTLTW roads. Driver age, traffic density, alignment, presence of an intersection, and hands on wheel were found to be highly related to driver behavior fault.

Conclusions: Young drivers (16–19 years old) were more likely to have driving behavior fault compared to other age groups. The presence of an intersection may also slightly increase the probability of driver behavior fault. The curve on the road can increase the risk of driver behavior fault. Drivers were more likely to have driver behavior fault when no hands were on the wheel. Sites with curves near intersections are locations where drivers are more likely to have behavior fault.

Keywords:

• Driver behavior
• rural 2-lane
• 2-way roads
• naturalistic driving study data
• near crash

Acknowledgments

The author acknowledges Alyssa Hernandez of the Transportation Research Board and Suzie Lee at Virginia Tech Transportation Institute for their support in obtaining the NDS data.

Additional information

Funding

This article was sponsored by the SHRP 2 Student Paper Competition.