Time-dependent OD Estimation Model Based On Multi-source Data

Urban population and the number of vehicles increase rapidly,bringing many urban traffic problems.The spatial and temporal distribution characteristics of traffic demand is the basis of the comprehensive management of urban traffic,which is urgent for urban transportation planners and managers to understand.The traditional researches are mostly based on fixed monitors,with high cost of data acquisition and low processing efficiency.In the context of multi-source data,the theory and method of OD estimation were studied based on multi-source data and related technologies.First,from the perspective of data acquisition and processing technology,the characteristics and processing difficulties of Big Data of Internet of Vehicles were also analyzed.In order to improve the effectiveness of the detector to collect traffic flow data,an algorithm to identify and repair the erroneous data based on S-G filtering and datadriven method were proposed,and the framework of data quality control system based on the multi-source data was established.The algorithm was verified by dealing with data measured on a certain road in Beijing.The results showed that applying S-G filter technique in the time domain to recognize and repair abnormal data could greatly reduce the proportion of abnormal data,and is practical with high computing speed,therefore can be used to support multi-source data fusion and traffic flow data analysis.The method provides a new idea for fault data processing,and lays a good foundation for the following analysis based on multi-source traffic data.Then,on the basis of basic data quality control,an OD estimation method based on split factor was studied,in order to make full use of multi-source traffic data and to improve the precision of OD estimation.The method was divided into two stages.Considering the factors such as OD,link flow and turning ration,the static OD estimation method was established to solve the static control OD total amount,which is the input of the next stage of dynamic OD estimation.The dynamic OD estimation model based on multi-objective optimization was established,genetic algorithm was applied in solving the dynamic split factor to achieve the obtaining of dynamic OD estimation results.Experimental network examples were presented to verify the validity of the model.Research results showed that: with the increase of effective information,static and dynamic OD estimation accuracy were improved,and that the apriori OD information had the greatest influence on the accuracy of the static OD estimation,and supplementary turning ration of information also helped,to a certain extent,to improve the estimation accuracy.Finally,a simulation experiment platform was built to verify the validity of the OD estimation model.Taking a real road network in Beijing as the research object,the based on VISSIM simulation platform architecture was studied,the database was designed,and simulation data was collected by running the simulation model.The precision and the effectiveness of the OD estimation algorithm was analyzed.