Modeling and Conflict Prediction of E-Bike Violations at Signalized Intersections

Abstract

Because e-bike violation behaviors significantly affect the efficiency and safety of signalized intersections, they should be investigated for making better countermeasures. Based on video surveillance data, the research compares and analyzes the e-bike violation behaviors and related traffic conflicts at signalized intersections under the influence of personal attributes and spatial-temporal scenarios. To reveal the impacts of multiple factors on e-bike violation behaviors, a Cox proportional risk regression model embedded with parametric and nonparametric features was developed, and the waiting tolerance time of violated and nonviolating e-bikes has been considered as the censored and complete data, respectively. To further understand the characteristics of traffic conflicts that are caused by the violated e-bikes and account for the heterogeneity in traffic conflicts, a generalized linear mixed model (GLMM) based traffic conflict prediction method is proposed to predict the frequency of e-bike related traffic conflicts. Based on the observed data of 5,435 e-bikes at signalized intersections, the results show that the overall violation rate of e-bikes is 44.01%, which is 1.21 times that of conventional bicycles. Different from the conformity behaviors of conventional bicycles, the narrowed nonmotorized vehicle lanes or the larger group size could restrict the e-bike violation behaviors. Additional traffic assistants and the presence of left-turn only phases would effectively reduce the e-bike violations. The proposed GLMM-based traffic conflict prediction method is better than the generalized linear model (GLM) method for modeling e-bike related traffic conflicts, and the prediction result for the ordinary conflict frequency is superior. The findings indicate that enhancing the management of e-bike violation behaviors, such as occupying motor vehicle lanes and over-line waiting, increasing the number of traffic assistants, and optimizing signal phases, could reduce the occurrences of e-bike related traffic conflicts.