Research On OD Estimation Of Road Network And Dynamic Traffic Assignment Based On Vehicel Trajectory

The OD matrix describes the travel characteristics between traffic zones,which is an important basis for making traffic planning and traffic management measures.Most of the existing OD estimation methods are based on the backward deduction of link flows,which fails to take fully advantage of the rich traffic flow information provided by vehicle trajectory,and the assignment matrix is difficult to accurately describe the rules of vehicle travel path selection in the real road network.In recent years,with the rapid development of intelligent transportation,using intelligent transportation data to carry out traffic management and control to alleviate urban traffic problems has been paid more and more attention by traffic managers.License plate recognition data has the advantages of large sample size,wide coverage and high precision.It can provide various traffic information such as link flows,turning flows,link travel time and vehicle trajectory.Therefore,based on the license plate recognition data as the basic data source,this paper extracts vehicle trajectories from it,and conducts road network OD estimation research.This paper first introduces the data preprocessing method of license plate recognition data and road network topology information.Then through data induction,integration and other operations to transform the original data into the form of convenient for data mining.The network topology information is integrated to provide data support for extracting vehicle trajectory,and the quality assessment of the basic data is carried out to verify the overall accuracy of the data used in this paper.In this paper,the method of vehicle trajectory extraction and incomplete travel trajectory completion algorithm based on license plate recognition data is studied.Firstly,the link travel time was extracted and some outliers were cleaned by means iteration method.On this basis,the dynamic travel time threshold was determined to separate the travel chain and obtain the vehicle trajectory.Then,the incomplete travel trajectory completion algorithm is proposed.For the trajectory with missing single point and two points,the incomplete travel trajectory is completed by combining the travel information of the missing starting and ending point and road network topology data.For the trajectory with three or more missing points,multiobjective decision-making algorithm is adopted.The weight of decision-making index is determined by grey correlation method,and the optimal alternative trajectory is determined by TOPSIS algorithm as the complement trajectory.Finally,the validity of the algorithm is verified.When the number of missing points is less than or equal to 4,the completion accuracy is greater than 90%,which is greatly improved compared with other completion algorithms.In this paper,an OD estimation model based on the least square algorithm is established by integrating vehicle trajectory and link traffic flows.The objective function is to minimize the sum of the deviation between the estimated OD and prior OD and the deviation between the estimated OD turning flows and the measured turning flows,in which the prior OD and the assignment matrix are obtained by using the trajectory information of some vehicles.The OD estimation model is solved iteratively by genetic algorithm.The results of OD estimation are analyzed and verified,which show that compared with the link flows,the turning flows can provide more known information,improve the accuracy of OD estimation.When the equipment recognition rate is greater than 50%,a more reliable OD estimation results can be obtained.Finally,a dynamic traffic assignment model based on LSTM neural network is established by using the estimated network OD and measured traffic flows,and the performance of the model is evaluated and analyzed,which confirms the effectiveness of the study on dynamic traffic assignment by neural network.