Development of Bicycle Level of Service Models for Urban Un-Signalized Intersections using Artificial Intelligence Approaches

Abstract

The use of bicycle mode is highly essential for a developing country like India. Therefore, to safeguard the interests of bicyclist in terms of safety and comfort, Bicycle Level of Service (BLOS) models are required to be built for heterogeneous traffic flow condition since most the research has been limited to homogeneous traffic type. In this regard, the work has been carried out to examine the riding quality of bicyclists while passing an un-signalized intersection. Un-signalized intersection forms complex situations for vehicles crossing over it since it is uncontrolled where priority movements are often violated leading to havoc and jams in the junction. In this study, video recording is done on different un-signalized intersections present in cities of Nagpur, Lucknow and Rourkela. The traffic and geometric data are also collected subsequently with perceptions being recorded from bicyclists to get the ratings of intersections. Several important geometric and operational parameters are inputted in the modeling process such as major road width, shoulder width, number of lanes, traffic volume, cross street volume, pedestrian volume, parking and commercial activity. BLOS models are developed by using two Artificial Intelligence (AI) techniques namely, Multivariate Adaptive Regression Splines (MARS) and Artificial Neural Network (ANN). MARS model (average R2 = 0.90) was observed to have better prediction accuracy than the ANN model (average R2 = 0.87).