Search

​

Horizontal curves make up only a small share of mileage in the United States, yet accounted for 23% of all fatal crashes. Overall, research has shown that the average crash rate for horizontal curve locations is approximately three times the average crash rate for the tangent section. Thus, it is important to understand how drivers can more safely transverse these sections of roadways, such as slowing to a safe speed. Drivers’ speed at horizontal curves is due, in part, to their perception of the way in which curves look upon approach. This perception is linked to the curve radius and deflection angles of horizontal curves.
Previous research found that on the tangent section before a curve began, drivers cruised at their highest speeds before slowing most significantly right before the start of the curve (Point of Curvature). Following this, drivers increased their speed slightly until the midpoint of the curve, before exiting the curve at approximately the same speed as the midpoint. However, to date, research has not considered a continuous study of driver speed throughout the entire length of a curve in terms of the impact of the curve deflection angle alone. Thus, it is not known how the intensity of a curve impacts driver performance throughout the time a driver spends traversing the length of a curve, and what the relationship between speed, lane position, curve radii, and deflection angle is throughout each portion of a horizontal curve, from the tangent section prior to the curve, through the tangent section following the curve.
Phase 1: Virtual simulation scenarios will be developed using recorded video of representative horizontal curves from the instrumented vehicle (with camera and LiDAR) and with scenarios developed by GTA5 mod and/or CARLA. These scenarios do not require the physical presence of the participants, but instead can be completed remotely. The scenarios will record the actions of the participants (as if they are playing a video game or playback a video of the horizontal curve). Although this is a reduced version of the actual driving simulator, the recorded actions (e.g., key pressing for gas, brake, steering, etc.) will provide many insights without the need for the physical presence. In addition, this approach provides flexibility for different curves. Phase 2: When the human subject research resumes on campus, we will proceed with a series simulator tests to collect continuous lateral position, acceleration, and velocity data of each driver for each experimental drive on the simulator guided by the outcomes and configurations from Phase 1. Modeling techniques will be utilized to investigate the relationship between continuous speed (and speed differential), braking time, pedal movements, distance traveled along the curve, lateral position, curve radii, and deflection angle.