Citation

Li M, Chen XM, Lin X, Xu D, Wang Y. J. Intell. Transp. Syst. 2018; 22(3): 229-243.

Copyright

(Copyright © 2018, Informa - Taylor and Francis Group)

DOI

10.1080/15472450.2016.1215921

PMID

unavailable

Abstract

This article proposes a connected vehicle (CV)-based dynamic all-red extension (DARE) framework for adaptive signalized intersections to avoid potential crashes caused by red-light running (RLR) behavior. The conceptual framework consists of three components: CVs, road-ide equipment (RSE), and traffic control devices. The RLR prediction of signalized intersections is a crucial component of DARE that avoids potential collisions caused by RLR behavior. This article formulates the RLR prediction as a binary classification problem based on continuous trajectories measured by radar sensors. In the CV environment, vehicle trajectories and real-time signal timing could be obtained via vehicle-to-infrastructure and vehicle-to-vehicle communications. Using continuous trajectories, individual speed, acceleration, and distance to the stop line at the red-light onset time are selected as classification attributes. Nonweighted and weighted least square support vector machines (LS-SVM) are adopted to solve the RLR prediction problem. Parameter tuning is conducted by the cross-validation in the discrete parameter space and the Bayesian inference in the continuous parameter space, respectively. As a comparison, the existing DARE approach at adaptive signalized intersections based on inductive loop detectors is discussed. In field experiments, vehicle trajectories collected by radar sensors were used to simulate the CV environment.

RESULTS showed merits of the proposed continuous trajectories-based DARE approach in terms of prediction rate, missing rate and false alarm rate, and thus a better RLR prediction improved the intersection safety.

Language: en

Keywords

connected vehicle (CV); dynamic all-red extension (DARE); intersection; least squares support vector machines (LS-SVM); radar; red-light-running (RLR)