Naturalistic yielding behavior of drivers at an unsignalized intersection based on survival analysis

TL;DR

This paper analyzes drivers' natural yielding behavior at unsignalized intersections using survival analysis, providing insights crucial for developing autonomous vehicle algorithms that better mimic human driving patterns.

Contribution

It introduces a novel application of survival analysis to model driver yielding behavior at unsignalized intersections, highlighting variations across different interaction scenarios.

Findings

Distinct yielding behavior characteristics identified

Speed reduction times vary significantly across scenarios

Implications for autonomous vehicle behavior modeling

Abstract

In recent years, autonomous vehicles have become increasingly popular, leading to extensive research on their safe and efficient operation. Understanding road yielding behavior is crucial for incorporating the appropriate driving behavior into algorithms. This paper focuses on investigating drivers' yielding behavior at unsignalized intersections. We quantified and modelled the speed reduction time for vulnerable road users at a zebra crossing using parametric survival analysis. We then evaluated the impact of speed reduction time in two different interaction scenarios, compared to the baseline condition of no interaction through an accelerated failure time regression model with the log-logistic distribution. The results demonstrate the unique characteristics of each yielding behavior scenario, emphasizing the need to account for these variations in the modelling process of autonomous vehicles.