The right turn: Modeling driver yielding behavior to e-scooter riders

Journal article, 2025

Electric scooters (e-scooters) are a relatively new and popular means of personal transportation in many cities. Notably, they have been involved in crashes with other road users. Crashes with motorized vehicles are particularly critical since they result in more severe injuries or even fatalities. While previous work has highlighted the consequences of failed interactions, we know little about drivers’ interactions with e-scooters and how to improve them. In this paper, we conducted a test-track experiment to study how drivers negotiate a right turn at an intersection with an e-scooter. Using Bayesian regression, we modeled whether drivers yield to the e-scooter according to their approaching speed and the difference in time-to-arrival, and we were able to predict drivers’ intentions with an AUC of 0.94 and an accuracy of 0.82 in cross-validation. The model coefficients indicate that drivers yield less often when approaching the intersection at a higher speed or larger projected gap. We further modeled drivers’ braking timing (time-to-arrival) and strength (mean deceleration), yielding RMSEs of 1.42 s and 0.33 m/s2, respectively. As a reference for driver behavior when interacting with an e-scooter rider, the model can inform the development and evaluation of support systems to warn drivers more effectively.