Trajectories Estimation Of Passengers In Urban Rail Transit Under The Long And Short Routing Mode

Urban rail transit is the backbone of urban transportation system and the artery of national economy because of its large traffic volume,high punctuality rate and environmental friendliness.The imbalance between supply and demand of transportation capacity caused by the imbalance of spatial and temporal distribution of passenger transport demand widely exists in rail transit system.Long and short routing mode can avoid the waste of capacity in small demand zones while ensuring the supply of large demand zones,but this mode also increases the variability of passengers’ spatial-temporal trajectory which complicates this black box problem.Passengers’ spatial-temporal trajectory is the basis for the operation organization department to monitor the state of the network,optimize the operation organization scheme,and formulate measures for limiting and controlling the passenger flow,while current researches on spatial-temporal trajectory estimation ignored the influence of long and short routing mode.Therefore,on the basis of AFC(Automatic Fare Collection)data,train diagram data and network data,this paper studies the method of passengers’ spatial-temporal trajectory estimation of urban rail transit under the long and short routing mode,and has carried out the following work:1.Based on AFC data of Chengdu Metro Line 2,the characteristics of passenger flow distribution of rail transit under the long and short routing mode are studied.Firstly,the characteristics of passenger flow distribution in working days and non-working days are compared and analyzed in both the temporal and spatial levels.The results show that the passenger flow during morning-evening rush hours on working days is significantly larger than that on non-working days,and there are large passenger flow volume gaps between different stations.Secondly,the characteristics of passenger flow spatial-temporal distribution in long and short routing mode and single routing mode are compared and analyzed.The results show that the spatial and temporal distribution density of passenger flow in long and short routing mode is significantly higher than that in single routing mode.2.An estimation model of passengers’ spatial-temporal trajectory under the long and short routing mode is proposed.Using subprocess analysis method,passenger trip is decomposed into several subprocesses—gathering,delay,on-train,line interchanging,long and short routing transferring,and scattering.Firstly,network topology is constructed based on network data,and KSP(K-Shortest Path)algorithm is used to screen the optional path set between all O-D pairs.Secondly,on the basis of AFC data and train diagram data,the passenger flow is divided into 10 categories according to the number of optional routes and feasible trips,interchange times,and whether there is a long and short routing transfer,and the feasible trip sets of various passenger flow are screened;Finally,based on Bayesian theorem,the whole trip probability is calculated by subprocess probability,and the trip probability model corresponding to passenger flow type is constructed.3.Based on the trip estimation results,a trip subprocess probability estimation method is proposed.Based on the passenger flow with single feasible trip,the sample of scattering time is extracted,and the probability distribution of scattering time is estimated by nonparametric kernel density estimation method.Based on the estimation results of passenger flow trip without interchange,the gathering time samples considering delay are extracted,also the probability distribution function of gathering time and delay are obtained by using Gaussian mixture model and EM algorithm.Based on the estimation results of passenger flow trip with one interchange,the interchanging time samples are extracted,and the probability distribution of interchanging time is fitted by nonparametric kernel density estimation.Based on the estimation results of passenger flow trip with long and short routing transfer,the selective probability of long and short routing transfer stations is calculated by the delay probability of long and short routing stations by using Bayesian theorem.4.Taking Chengdu Metro as an example,the spatial-temporal trajectory of passengers under the long and short routing mode is estimated.Firstly,based on the AFC data and train diagram data of Chengdu Metro on a certain day,the probability distribution of gathering time and scattering time of 136 stations and the probability distribution of interchanging time of 14 transfer stations are estimated and verified by K-S test.Secondly,taking the passengers from Chunxi Road to Longquanyi and Weijianian to Longquanyi as examples,the spatial-temporal trajectory corresponding to AFC data is estimated,and the simulation travel time based on the results is compared with the real travel time,which verifies the accuracy of the estimation model.