Investigation of driver performance in different horizontal and vertical curve combinations using SHRP2 naturalistic driving data

The goal of this study was to examine driver performance under different combinations of horizontal and vertical alignment in rural two-lane roads. The SHRP2 roadway information database (RID) was used to identify study sites on rural two lane roads from New York and Pennsylvania that included tangent sections, straight grades, horizontal curves, vertical curves, and combinations. The study included the development of an algorithm that used the RID to identify vertical curves. For these study sites, time series data, forward video, and eye glance data were extracted for a sample of drivers and trips from the SHRP2 Naturalistic Driving Study (NDS). The time series data included kinematics variables such as speed, acceleration, lane position, distance from left and right lane markings, and steering wheel position. The forward video provided the driver’s viewpoint, and was used to record events such as presence of lead vehicle, presence of vehicle in the opposing lane, and intrusions from vehicles and other road users. The eye glance data provided information on where a driver was looking every fraction of a second. The analysis of the NDS data showed that the alignment categories that include a horizontal curve have the worst performance in terms of the lane deviation measures. In addition, sharper horizontal curves are associated with the higher absolute values of lane deviation. The analysis was conducted using both disaggregate and aggregate methods. Comparison of the CMFs from a recent study by the Federal Highway Administration [1] with the performance measures indicates a good match with the lane deviation measures, indicating that the lane deviation measures could serve as good surrogates for crash propensity.