Safety Analysis of Horizontal Curves Using Real Traffic Data

Abstract

Researchers are still seeking a better understanding of the parameters that affect safety in horizontal curves. Curves are one of the most critical sections of the road network contributing to a high percentage of serious runoff accidents and lane-changing crashes. Moreover, driving in curves requires combined control of both steering and speed, taking into account the dynamic response and limits of the car. The objectives of this study were to evaluate the safety performance of horizontal curves by analyzing vehicle dynamic signals, such as lateral acceleration and speed, as well as quantitative analysis of lane-changing maneuvers. The study uses real traffic environments where driver behavior and vehicle response data were recorded and stored during regular operations without subjecting the driver to any experimental controls. A total of 96 curves, equally distributed for left and right turn directions, have been collected and grouped according to their radii. The analysis identified frequent overtaking and lane-change maneuvers on the curves, of which 20% more lane changes occurred on right curves than on left curves. Lane-change maneuvers also increased significantly with increasing curve radius. The curve entrance was found to be the most dangerous segment of a curve. Current design practice assumes the safety risk is constant when driving along horizontal curves. The results also showed that drivers consider curve radius in choosing their driving speed rather than the posted speed limit of the curves. The study showed how road design influences the driver’s strategy by establishing links between curve features, vehicle dynamic responses, and the driver’s behavior. Analyzing road characteristics gave insight into how road geometry affects the vehicle dynamics relevant to safety and driving strategy through curves. The findings are useful inputs for reviewing curve design, selecting appropriate countermeasures, and improving active safety devices.