Research On Reliability Of Route Travel Time Based On GPS Data Of Online Ride-hailinge

Travel time reliability(TTR)is an important indicator for the study of travel behavior in the field of transportation.Its accurate quantification is of great significance for mastering the temporal and spatial characteristics of traffic.When applying the large-scale GPS data of online car-hailing to TTR prediction,the traditional road segment division method cannot accurately measure the delay time of the intersection.Based on this,this paper applies a virtual road section division method that fully considered the delays in all directions at the intersection,which is more suitable for large-scale GPS.And then a route travel time reliability model based on the virtual road links and Vine Copula multiple joint distribution was established.The purpose of this paper is to break the limitation of binary copula model in the analysis of two or two links of multi-link path,fully consider the correlation of link travel time and the heterogeneity of distribution,and improve the accuracy and universality of the model.In data processing,in order to ensure that there is a sufficient number of trips between the same OD and improve the accuracy of the model,a Geohash-based trajectory clustering method was proposed,the OD partition was first located quickly and then the trajectories between a Specific OD were clustered,which can greatly reduce the complexity compared with traditional clustering algorithm and improve the clustering rate of large quantities of GPS data.And then the spatial distribution characteristics of OD in different time periods were analyzed and two different types of hot OD paths were extracted for subsequent analysis based on the Didi Express data in the Second Ring Road of Chengdu in November 2016.In terms of model construction,considering the delay in each direction of the intersection and the correlation between the road sections.This article constructed of the route travel time reliability model based on the construction of virtual road sections and the Vine Copula model.Furthermore,for the travel time characteristics of the virtual road section proposed in this paper,the virtual road section is divided into different types according to the types of intersections and the presence or absence of traffic lights,etc.Then the GMM model was used to fit their travel time distribution and was examined,and the better fitting function is selected and tested.At the same time the influencing factors of virtual link travel time distribution were analyzed.In order to verify the applicability of the model,two hot OD paths selected in the previous paper are taken as examples for case analysis.In the verification of the route travel time reliability modeling example,it was found that the prediction results of the R-Vine Gaussian Copula model based on the virtual road links,which considers the correlation,are closer to the empirical value than the independent convolution model based on the actual road section and other Vine Copula models.