Research On Travel Time Estimation Using Sparse Track Data

Travel time is a key indicator reflecting the state of traffic operation.This indicator is more intuitive and understandable for both traffic managers and traffic participants,which it is also an important basic data in the intelligent transportation system,and real-time accurate and efficient travel time estimation results have strong practical value.The devel-opment of various data acquisition technologies has made traffic-related data resources more and more abundant,but due to the complexity of the urban transportation system,the information obtained is still sample data rather than a complete set.The previous travel time estimate method usually pays attention to the travel time of each road segment or sub-path.The travel time of the target path is obtained by summation,and the overall consideration of the path is considered.The result obtained by this method is not reliable.In view of the above problems,this thesis proposes a method for estimating the travel time.First of all,this thesis summarizes the existing research results of travel time estimation,and points out the advantages of existing research and the problems to be solved.Secondly,the vehicle trajectory data is basically cleaned and the data prepro-cessing is used to extract the effective information through the coordinate correction map matching,and the dimension of the rich data of the meteorological data and the interest point data is also added.The relationship between the trajectory data and the travel time is then analyzed.On this basis,the multi-dimensional spatiotemporal data is fused by tensor,and the sizing decomposition technique is used to complete the sparse data.The optimal path is proposed for the contradiction between the number of trajectories and the length of the trajectory in the travel time estimation.The segmen-tation model is implemented to realize the travel time estimation based on the link.For different usage scenarios,more sparse data is also proposed.Considering the relationship between adjacent trajectories,the travel time variation of adjacent trajectories is analyzed,and the travel time estimation based on adjacent trajectories with higher computational efficiency is realized.Finally,the estimation method proposed in this thesis is verified on the real dataset of Chengdu,and the difference between the two methods under different conditions is compared.Compared with other forecasting models,the results of this method have obvious advantages,which can provide a reliable reference for travel route selection and commuting efficiency evaluation.