Assessing the Complexity of Potential Bicycle Interference on Vehicles on Urban Road Segments

Abstract

Interacting with bicycles on urban road segments is complex for vehicles, due to the diverse and flexible cycling behavior that can cause interferences. Studies show that driving mixed with bicycles is also a great challenge for autonomous vehicles (AVs), and it is necessary to consider the interference of bicycles when selecting public roads for testing. However, existing road evaluation methods mostly focus on autonomous driving functions and accident analysis, although bicycles have been considered, often with insufficient consideration of their interference. This study analyzes two types of cycling behavior that could interfere with vehicles, including lateral (turning handlebars) and longitudinal (braking or accelerating) behavior, with each occurrence of such behavior considered as one potential lateral or longitudinal interference. From the perspective of cycling behavior, a framework is proposed to assess the complexity of potential bicycle interference on vehicles on road segments. A higher frequency of both potential lateral and longitudinal interference represents a higher complexity of potential interference. A naturalistic field experiment was conducted to collect the potential lateral and longitudinal interference frequency and the environmental parameters of road segments. The quantile regression model was applied to analyze the environmental factors influencing different interference frequencies separately and further establish the assessing model of the potential bicycle interference complexity, and the usability of the model has been demonstrated with a case study. Results show that the potential interference complexity varies across road segments, with some factors leading to more frequent potential lateral and longitudinal interference but with varying degrees of impact (such as the separation between bicycles and vehicles), while some only affect the lateral interference frequency (such as the on-street parking condition). The proposed framework can help autonomous driving companies or evaluation agencies to select appropriate testing roads, thus promoting the development of autonomous driving.