The Study Of Dynamic OD Estimation Based On Multi-source Data

With the rapid development of society,economy and technology,the urbanization problems become seriously.Especially,the population explosion braking the balance between urban traffic demand and traffic supply.The most direct manifestation is the traffic jam and the longer travel time.In order to solve the increasingly serious traffic problems,the Intelligent Transportation System(ITS)came into being.It not only can provide effective traffic management and traffic guidance according to the real-time traffic conditions of the road network,but also can provide technical support for alleviating traffic congestion and improving road traffic efficiency.The information of OD demand is the basis for analyzing the traffic status of road network.Scholars from various countries engage the study of OD demand from different aspects in recent years.However,with the influence of the complexity of road network structure and the difficulty of obtaining various data,current studies are most based on the link flow and less from the multi-source detector.In this context,this paper proposes a dynamic OD estimation method based on multi-source data.Fully exploiting the values of various data to achieve the goal of effective estimating the dynamic OD demand,so as to provide basic support for the route guidance system and traffic management system.It is of great significance to alleviate traffic congestion and build a harmonious,livable and convenient urban environment.The purpose of this study is to establish a dynamic OD estimation model for urban network based on different type data and different source data,and taking the relationship between multi-source data and traffic demand as the main structure.Using the traffic information,such as part of link flow,turning flow in each intersection,track flow gaining from Vehicle License Plate Recognition System(VLPR)and trajectory information of floating cars,to achieve traffic demand of road network can simplify the acquisition process,improve the usage of traffic big data and provide the theoretical foundation for the Intelligent Transportation System.The main works of this study are as follow:1.The study of basic theory of dynamic OD estimation.Reviewing the main modeling methods,modeling ideas,and basic data types in existing dynamic OD demand studies.Focus on the problem of making route choice,the issue of dynamic traffic assignment,how to get the travel time and all of their solutions.Organizing the timing-spatial relationships among the traffic parameters in the road network and defining the type of the data,model outputs and constraint condition of the model.2.The study of data processing and data recovery methods of multi-course data.Afterclassify the methods of collecting the multi-course data and the direction of main application of multi-course data,the study pay attention to how to improve the quality of the collecting data.Abnormal data and missing data are two common problems in data base,according to the frequency of problem data arise,this study proposed two methods to recover the data,a single time data recover method and the continuous time data recover method.The case study compare the proposed methods with the traditional method,and set the MAE,MSE,RMSE and MAPE as the evaluation indicators to explain the feasibility and effectiveness of the proposed method.Moreover,the impact of the number of problem data on the data recover results was studied and analyzed.3.The algorithm of traffic assignment matrix based on vehicle trajectory data.The data of vehicle trajectory and the vehicle plate information show the connection between the traffic demand and the flow detection process,so an algorithm was constructed base on it.In the path selection model,the Path-Size Logit model is introduced,and the travel time acquired in real time and the length of each road are taken into consideration as the travel impedance,which avoids the defects of the MNL model and the assumption of having the same space headway in the same vehicle group on the same path.It also avoid the rules of First In First Out(FIFO).4.The study of dynamic OD estimation based on multi-source data.By analyzing the intrinsic relationship between traffic parameters and OD demand in the road network,a dynamic OD estimation model for multi-source data based on the state-space model was constructed from two perspectives: time and space.Taking part of link flow,turning flow of each intersection and part of route flow as detected data can decrease the number of unknown traffic parameters.4 different scenes are designed in the case study to evaluate the proposed model.This study make a conclusion about how the type of detected data and the number of data influence the result of dynamic OD estimation by comparing the 4 error evaluate indicators(MAE,MSE,REMSE,MAPE)in 4 scenes.One of the conclusions is the more diversity of the type of data,the more accurate of the demand result.Another is that when there is only one kind of data,the accuracy of the estimation result is more accurate as the number of data increases.Especially,NO.4 scene describes the situation of problem data in the database.By comparing and analyzing the error indicators of the model calculation results and the actual demand results,the impact of data quality on the OD estimation results is obtained,and the effect of data recovery methods on the error propagation of the dynamic OD estimation model is explained.