Research On Urban Road Travel Time Prediction Based On Information Fusion

With the rapid increase in car ownership and people travel, the urban traffic plays more important role in the economic development,but the bad thing is worsening environmental pollution,traffic congestion and frequently traffic accidents. Intelligent Transportation System (ITS) is recognized as the most effective way in dealing with traffic problems, and travel time is the key problem. To be effective control and guide the urban traffic flow,accurate travel time is an essential element. At the same time, in order to reduce the lag influence on traffic management and guidance, prediction of travel time has been continued to explore by many scholars. Research on urban street travel time prediction theory and method and relatively accurately predicting urban road travel time can effectively alleviate urban traffic congestion, speed up the construction of urban modernization. It has an irreplaceable significance in social economy development.To develop a effective way in travel time prediction, the paper introduces the theory of traffic data acquisition technology in the first time, then show the research objection of urban street that contain the intersection, and then develop the travel time estimation model, travel time prediction model and prediction time fusion model. In this paper the basic idea is:using improved HCM2010 built the city street travel time estimation model, and the historical trend model and Kalman filter established urban street travel time prediction model;but the predictions of the model itself has its deep-rooted bad habits,and the detector itself precision also had influence in the predictive value in the single source traffic data;the paper constructed fusion model of travel time prediction value based on a neural network model in the last time; and the output variables (the value of travel time) of city street travel time estimation model and two prediction model is the input variables of the integration model, and then the final output variables is a fusion variables of travel time prediction value.At last, according to the actual city street and traffic flow data in Chengdu and the traffic simulation software VISSIM4.3, the paper got related traffic data.Used the fusion model based on neural network, the urban street travel time prediction value is obtained. The results that fusion value compared with the simulation travel time show that the prediction error of the city street travel time fusion value is within 5%, and the fusion model is more effective than other prediction way,such as Kalman filter way...