On the fundamental diagram and driving behavior modeling of heterogeneous traffic flow using UAV-based data

Abstract

A significant difference in the behavior of heterogeneous and undisciplined traffic streams is observed when compared with the conventional traffic flow. Most of the existing traffic flow models are developed considering the traffic stream with strict lane discipline. Several studies from South Asian countries have reported high heterogeneity in the traffic stream with weak or no lane discipline. This study, for the first time, develops fundamental traffic flow diagrams for the heterogeneous and undisciplined traffic stream by analyzing traffic videos captured using Unmanned Aerial Vehicle (UAV) in Karachi, Pakistan. UAV-based speed-density data is modeled using a weighted least-square regression technique, and stochastic fundamental diagrams (FDs) are developed to represent the entire range of speed-density data. The stochastic FDs are used to determine the 85th percentile speeds to implement speed limits on major urban arterials. The multi-modal FDs show a significant difference in the behavior of different modes in the traffic stream. The aggressive behavior of motorbike riders, which put them at a higher risk of accidents, highlights the need for policy measures to enforce discipline in the traffic stream. The main contribution of this study is the utilization of a UAV-based geospatial analysis technique for accurate extraction of longitudinal and lateral distances between vehicles to determine the relationship between macroscopic and microscopic parameters of traffic flow. This study shows that lateral gaps between vehicles are inversely related to traffic density. The longitudinal gaps observed for local heterogeneous traffic show a significant difference with the longitudinal gaps estimated using a standard car-following model. The macroscopic and microscopic models for heterogeneous and undisciplined traffic flow presented in this study could be useful in developing novel traffic flow models and calibrating the existing microscopic/ macroscopic traffic flow models for the traffic streams with similar heterogeneity and lane behavior.