370 An empirical analysis of driving patterns on horizontal curves using trajectory data

<h3>Background</h3> The vehicles traversing a highway passing through hilly terrain are prone to crashes due to complex highway geometries. It is evident from past studies and accident statistics that many of the crashes in hilly terrain occur on or near the curves as compared to tangent sections. Out of several accident types, road run-off is one of the most common crash types, which primarily occurs due to the presence of unexpected geometric features. <h3>Objective</h3> The present study analyses the curve negotiation behavior of drivers using videography-based trajectory data. <h3>Methods</h3> Videography data was collected on fourteen horizontal curves of a Guwahati-Shillong highway with an average duration of 1 hour. The study section is a two-lane, undivided rural highway passing through mountainous terrain. The vehicle trajectories from the videograph were extracted using a semi-automated image processing technique, and a multiple homography coordinate transformation technique was performed to project the non-planar image scene. The data was only extracted for the free-flowing vehicles whose actions only depend on the highway geometry. The present study analyses the vehicle trajectories of primary commuting vehicles observed in the study locations in both directions. A Dynamic Time Warping (DTW) method was used to cluster the driving behavior, and a statistical analysis was performed to understand different driving patterns. <h3>Results</h3> The results indicate that the drivers’ behavior is different concerning the direction of travel, and the same is evident from the trajectories. <h3>Conclusion</h3> The analysis of trajectories shows that the vehicles are prone to depart the road space outside the designated right of way due to the erroneous perception of the road geometry. It was also observed from the vehicle trajectories that the different curve geometry and speeding behavior of vehicles contribute to different magnitudes of lane departure. <h3>Future Scope</h3> In the future, the authors will investigate the risk of conflicts for each driving pattern.