Study On The Evolution Characteristics And Simulation Control Of Traffic Flow Near Expressway On-ramp

With the acceleration of China's urbanization process and the increasing demand for transportation,the construction of urban expressway has also accelerated.However,in recent years,congestion on expressway has become a frequent traffic phenomenon in many cities.In fact,frequent congestion on expressway bottleneck is often the road section with complex traffic flow operation,such as the aggregated flow of on-ramp area.The frequent local traffic flow disturbance,but the metastable traffic flow which does not transition to an overall serious congestion,is the main factor causing the traffic capacity drop and macroscopic hysteresis.In this paper,the evolution process of traffic congestion was studied as a specific traffic flow condition,and the bottleneck area of the intersection of the on ramp of Yingtian Avenue elevated(near the 1865 Industrial Park)in Nanjing was taken as the research scene.The traffic flow was theoretically analyzed by using the high-precision vehicle trajectory data,and the improved ramp control algorithm and simulation analysis were proposed to reconstruct,analyze and predict the temporal and spatial characteristics of traffic conditions.Therefore,based on the extracted trajectory data,theoretically,this paper focused on the observation of the characteristics of the evolution of traffic congestion and the critical threshold of phase transition,as well as in-depth understanding of the impact of the car following and lane changing behavior on the formation and change of metastable traffic flow in the bottleneck area;In practical,this paper constructed a prediction model based on the direction of deep learning image recognition,and proposed a ramp linkage control algorithm which integrates the three-phase traffic theory critical phase transition threshold judgment,short-term traffic flow prediction and single point control.The secondary development simulation evaluated the improvement effect of expressway operation from the aspects of traffic efficiency and occupancy.As a whole,this paper includes the following three parts:Firstly,during the congestion formation,the unmanned aerial vehicle(UAV)was used to record all the vehicles which appearing on the bottleneck.Image recognition technology was used to calibrate and track all these vehicles,and then these vehicles trajectories data were obtained with the multi-steps process of data denoising and smoothing,splicing fitting and clustering reconstruction.Secondly,for the three-phase traffic theory,this paper proposed the microscopic lane-changing model to capture the disturbance propagation characteristics of the oscillation quantitatively,and the improved asymmetric model to study the hysteresis phenomenon and the asymmetric driving behavior,as well as studied the formation and evolution mechanism of congestion from the perspective of macroscopic traffic transition and microscopic driving behavior.Finally,based on a new deep residual network model,an on-ramp linkage control algorithm integrating various theories was proposed,and a secondary development simulation control platform was built to predict the short-term traffic flow phase transition and control the ramp metering rate;and the simulation effect was evaluated.This paper can provide theoretical and simulation support for the future development of traffic control technology.