Research On Urban Expressway Traffic State Estimation And Forecasting Method

With the rapid increase of concentrated urban motorized travel demand, the urban road traffic congestion becomes more serious. As the artery of the urban traffic, the congestions in urban expressways are particularly heavy, which sharply reduce the efficiency of the urban road traffic system. Therefore, to solve the urban traffic congestion problem, this paper includes the traffic flow theory, fuzzy theory and data mining technology to utilize the existing traffic data, such as flow, velocity and occupancy. Based on these theories, an integrated qualitative and quantitative method is proposed to analyze the characteristics of urban expressway traffic state, and make an accurate and real-time evaluation and prediction of urban expressway traffic state. And the main theses and achievements of this research are as follows.(1) Establish a traffic state evaluating method based on weighted traffic parameters.Considering the different effect of each traffic parameter (traffic flow, speed and occupancy or density) on classifying of traffic state, this research proposes the method to classify traffic state of urban road based on the parameter weight. According to a similarity measurement method based on weighted Euclidean distance, parameter weight learning algorithm gives each parameter a parameter weight by using the gradient method to minimize the parameter evaluation function. After obtaining parameter weight values, this research uses the weighted Euclidean distance to replace the common Euclidean distance in Fuzzy C-means Clustering (FCM). And the validation results show that the method is effective and practical.(2) Propose’ urban expressway traffic state index’ based on traffic parameters.The time-varying urban expressway traffic state is closely related to the traffic flow parameters. Based on the relationship between the parameters, and the principle that the distances between different traffic states and a certain point are quite different, this paper has established a quantitative method based on the weighted Euclidean distance to evaluate the urban expressway traffic state. Then, an ’urban expressway traffic state index’is obtained by the distance between a combination point and the given point. And the validation results show that the method is effective and practical.(3) Establish a method to analyze the urban expressway traffic state characteristics based on coefficient of variation.Coefficient of variation (CV) is an index that can be used to quantitatively measure degree of dispersion of a time series. CV could effectively characterize the variability of the distributions of the variables. While there is a change in dynamic structure of a system, the CV of the system variable will have some changes correspondingly. According to this characteristic of CV, the paper proposes a moving coefficient of variation method to detect abrupt change in a times series. And the validation results show that the method is effective and practical.(4) Establish an urban traffic state correlation analysis based on variation trend.According to the variation trend of urban expressway traffic state index time series, the paper divides the trend of short-term traffic flow into three phases:the trend of downward phase, the stable fluctuation phase, and the trend of rising phase. Then the paper researches the variation trend urban expressway traffic state time series based on accumulated anomaly, and calculates the correlation coefficient of accumulated anomaly values. And the validation results show the urban expressway traffic state has a strong variation trend correlation between.(5) Establish an urban traffic state multi step prediction method based on threshold auto-regressive model.According to the analysis of mutation characteristics and relevant characteristic of urban expressway traffic state, the time-series data of traffic state is separated into several segments by a method of sliding coefficient, and a threshold auto-regressive prediction model based on periodic characteristics analysis is then established. In this model, the nonlinear characteristics of the urban traffic state are represented by a combined model with several linear evolution functions, which reflects the dynamic evolvement rules of the traffic state. And the validation results show that the method is effective and practical.