Mixed-flow intersections are prevailing in many developing countries such as China and India. At mixed-flow intersections, there is no clear lane discipline or regular trajectories within the intersection, especially for the non-motorized traffic. This leads to more interactions and encounters between the motorized traffic and the non-motorized traffic. Hence, critical instantaneous decision events such as abrupt accelerating, decelerating, jerking, swerving, and swinging, may occur more frequently, which result in potential traffic conflicts and crashes. This study presents a methodology to identify critical instantaneous decision events at the mixed-flow signalized intersections, based on the entropy theory and high-resolution vehicle trajectory data. A three-dimensional cube searching algorithm is firstly proposed to extract general traffic events by examining the proximity between trajectories. A novel model incorporating vehicle kinematics and Permutation Entropy is then developed to identify critical events, by quantifying driving volatility based on the time-serial trajectory data. Next, 3, 349 vehicle trajectories and 805 bicycle trajectories with a resolution of 0.12 s collected at a signalized intersection in Shanghai are used to demonstrate the proposed method. Results show that the proposed method is capable of identifying critical instantaneous decision events, and tends to produce a higher identification ratio comparing with the conventional method solely based on kinematic thresholds. A sensitivity analysis is also conducted to investigate the effects of model parameters on the performance of the proposed method. The presented work could be applied for traffic safety assessment, real-time driving alert systems, and early diagnosis of risk-prone road users at mixed-flow intersections.
Keywords: Critical instantaneous decision events; Cube searching algorithm; Mixed-flow intersections; Permutation entropy; Trajectory data.
Copyright © 2018. Published by Elsevier Ltd.