Research On Spatio-Temporal Features And States Identification Of The Urban Traffic System

Urban traffic system is an open complex system,and traffic state is a kind of distribution modality of the traffic flow in a specific time and space.However,there is a lack of enough understanding about the spatial and temporal state of urban traffic system,which leads to a certain blindness in taking corresponding measures to control and alleviate traffic congestion.With the rapid development of information technology,it is convenient to obtain the data of urban traffic detection,but it is more important that providing more valuable information for urban traffic management through analyzing and processing detection data to identify and measure the spatio-temporal features of urban traffic.As such,based on the traffic detection data of Huaiyin,this dissertation analyzed the spatial and temporal difference and correlation features,research the methods of building traffic flow prediction model,threshold estimation model of traffic states and traffic state composite index.Firstly,the connotation and basic characteristics of urban traffic system were analyzed,spatio-temporal difference and spatio-temporal correlation of urban traffic system was elaborated.The information entropy was used to measure spatio-temporal difference of urban traffic system,the spatio-temporal similarity was measured by similar coemcient and spatio-temporal correlation was measured by correlation coefficient.Chaotic characteristics of the traffic flow were analyzed by phase space reconstruction theory,and were verified with calculating maximal Lyapunov exponent of traffic flow and average velocity.Secondly,incomplete or missing information sometimes occurred in data collection,and lead to traffic flow data loss,in order to obtain complete traffic flow data and further explored traffic flow distribution,combining with cloud model,cloud-neural network fitting method was used in traffic flow prediction.Considering spatio-temporal characteristics of urban traffic system,spatio-temporal correlation which used in prediction samples selecting was structured by spatial weight matrix and time delay,and cloud-SVM method was structured for state classification of traffic flow.The results of the calculation showed that two methods were satisfactory in deal with traffic flow data loss partly.Then,considering the volatility and instability of the urban traffic flow data,functional data analysis method was introduced in transportation research.Through transforming original discrete data into functional continuous data,variation tendency was analyzed by functional based methods such as functions derivative.Through estimating traffic status threshold of recurring congestion and occasional congestion,traffic flow data fluctuations exception was monitored and diagnosed in combination with MEWMA control chart,the result showed that recognition efficiency of abnormal conditions was improved.At last,in consideration of providing quantitative indicators of comprehensive traffic status to public,this dissertation built urban traffic congestion index,evaluated the urban traffic congestion from a macro perspective,and volatility of traffic index was analyzed through method of multi-fractal diffusion entropy.Coupling index was built to describe coupling balance of urban traffic system with extensive saturation and congestion index,in order to analyze the problem of urban traffic unbalance of traffic supply and demand.This method could provide some scientific references for relevant urban transport management departments.The models and methods introduced in this dissertation were verified and analyzed by the data of urban traffic detection of Huaiyin.The research results provided effective models and methods to analyze urban traffic detection data,the results also were expected to provide valuable information for urban traffic management and control.