Research On The Theories And Techniques Of Urban Expressway Traffic Congestion Evolution Mechanism And Condition Recognition

Urban expressways,connecting various areas of a city and providing long-distance express transportation services,have become the key to supporting the normal operation of urban transportation systems.However,increasing traffic pressures have caused a considerable proportion of short-distance transportation demand continue to lean towards expressways.The inefficient of expressways has become a well-known fact.Once a traffic jam occurs in an expressway,the relatively closed road structure is often not conducive to the control and grooming of congestion.It is imminent to carry out the research on the congestion evolusiton mechanism and traffic condition recognition.How to determine the traffic operation status quickly,reveal the internal mechanism of congestion evolution and forecast the development trend of the traffic status accurately are the pivotal and difficult points for expressway traffic control and management.Therefore,based on the provision of information services for traffic management,the paper proceeded from the practical problems of expressway traffic and carried out some relative research about the key theories and methods including congestion evolution mechanism,traffic condition recognition,and floating car data processing and traffic flow prediction,respectively.Based on the law of conservation of traffic flow,a homogenous abstraction of the bottleneck traffic flow with recurrent congestion and non-recurrent congestion was carried out.The equilibrium balance equation of cusp catastrophe theory was used to derivate a critical velocity equation for congestion propagation.Under consideration of some critical influencing factors,the parameters such as the propagation speed,duration,and scope of impact of congestion were quantified,and the model of the congestion evolution mechanism of the expressway was proposed.The proposed model was verified by simulation under two different traffic conditions.When traffic congestion occurred on the expressway,the propagation speed and the relevant parameters can be determined according to the flow difference between the bottle traffic flow and the congested-critical interface,thereby achieving appropriate and effective traffic control.In order to achieve accurate identification of traffic conditions,on the one hand,the mutation characteristics of the evolution process of traffic flow were analyzed based on the cusp catastrophe theory,and the spatial distribution of the traffic flow three parameters and the cusp balance surface were jointly analyzed to determine the best operate status and extract the three-phase traffic flow boundaries.On the other hand,clustering algorithms were used to analyze traffic flow characteristics from the perspective of data,and an improved semi-supervised K-means algorithm was proposed to achieve accurate traffic flow clustering.The results show that the three-phase traffic flow theory has a high degree of consistency with the real data.Finally the traffic condition recogonition were achieved based on the proposed traffic condition index and the predicted traffic flow speed data.Centering the prediction of traffic flow on expressways,the study was conducted at three levels: road segments,expressways,and regional expressway network.Aiming at the road segment,a macroscopic state model of traffic flow state was established.The improved optimized UKF algorithm was proposed to solve the model to realize the traffic flow state estimation.Aiming at the long-distance expressway traffic flow,the problems of data anomaly and data missing of floating car data were studied and analyzed at first.An anomaly data identification algorithm and a missing-data repairation algorithm were proposed.Then,a multi-scale hybrid prediction algorithm based on the characteristics of traffic flow was proposed.Using the processed floating car data,the velocity of the traffic flow was predicted and the three-phase traffic flow status was identified real-timely.For the regional expressway network,based on the framework of deep learning theory,the data standardization,data reconstruction,and traffic flow characteristics were studied,and a road network traffic prediction algorithm based on LSTM-RNN was proposed to ensure the stable recurring of cyclical traffic flow,meanwhile to amplify the fluctuation characteristics of the random traffic flow.The measured data proves that the improvement of data quality and model improvement help to reduce the training time of the model and improve the accuracy and practicality of the algorithm.Starting from the actual traffic problems in the urban expressway,the thesis completed the research on the four including traffic congestion evolution mechanism,traffic condition recognition,traffic flow prediction model construction and data processing and verification.It also proposed the prospect on the practicality and application scope about the model and algorithms,and others for further study,in order to provide a reliable theory and technical support for the urban expressway traffic management.