Evaluating pedestrian vehicle interaction dynamics at un-signalized intersections: A proactive approach for safety analysis.

The present research demonstrates the use of advanced trajectory based data to analyze road user interactions at an un-signalized intersection under heterogeneous traffic complexities. This study demonstrates an improvement over the conventional grid-based analysis to estimate surrogate safety measures (SSM). An advanced pattern-based approach to categorize pedestrian-vehicle interactions based on the road user behavior is proposed in the study. A concept of a two-interaction pattern has been applied, which deals with the responsive and non -responsive behavior of the road users, respectively. The behavior-based patterns were categorized based on the SSM like Speed, Time to Collision, and Gap Time profiles of the pedestrian and vehicle interacting on an un-signalized intersection. On conducting a variable importance test, i.e., k-fold test, it was comprehended that, for pattern-1, Time to collision (TTC), and for pattern-2 both TTC and Post Encroachment Time (PET) were showing required importance. Further, Import Vector Machine (IVM) approach was used to classify the severity levels based on selected indicators computed from 1486 events, occurring at three Un-Signalized intersections in India. The proposed severity levels will help to test and evaluate various infrastructure and control improvements for making urban intersections safe for road users. It was observed from the severity levels of both the patterns that, events involving non-evasive behavior can also result in critical interaction. Overall, the research provides an advanced framework for evaluating and improving the safety of the uncontrolled intersections.