Passenger Spatiotemporal Path Estimation And Bottleneck Identification Over Urban Rail Transit Networks

While alleviating road traffic congestion and optimizing urban spatial layout,green and efficient urban rail transit system is also confronted with the problem of passenger traffic congestion within the system due to the huge demand of passenger flow.Passenger flow congestion is becoming a stumbling block affecting the efficient operation and highquality service over urban rail transit system.Only by accurately and comprehensively identifying the bottleneck in the network can we provide scientific and effective guidance for train organization and demand control,so as to prevent further aggravation and spread of passenger flow congestion.Estimating the spatiotemporal distribution of passenger flow is the basis for effectively identifying the bottleneck of the network.The existing researches mianly estimate the passenger flow distribution by analyzing the passenger path choice factors and using aggregate allocation methods.The passenger flow allocation results of this kind of research often exist a big gap with the actual situation for the reason that it fails to describe the dynamic transmission characteristics of passenger flow in the network.In addition,research on bottleneck identification mainly focuses on the identification of capacity bottleneck,and less on passenger service experience and spatiotemporal evolution of network congestion.This paper analyzes the process of passenger-to-path and passenger-to-train from the perspective of the individual passenger to estimate the complete spatiotemporal path of the individual passenger with the aid of Automatic Fare Collection(AFC)data and train timetable data.In addition to the identification of capacity bottleneck,the service bottleneck identification considering the passenger travel service experience is conducted.Further more,the spatiotemporal evolution law of network congestion under the dynamic propagation characteristics of passenger flow is deeply analyzed.To sum up,the research is carried out from the following four aspects:(1)Research on the passenger-to-path estimation method.A mixed distribution model of multi-path Origin-Destination(OD)pairs is constructed by analyzing the travel time distribution of single path OD pairs.After the estimation of travel time components,the optimal selected path set of OD pairs and the corresponding passenger flow distribution allocation ratio can be figured out by combining Bayesian Information Criterion(BIC)and Expectation Maximum(EM)algorithm to calculate the posterior probability of passengers from each selected path.(2)Research on the passenger-to-train estimation method.Based on the estimation result of passenger-to-path,the probability estimation models for through passengers and transfer passengers are established by analyzing the the relationship between passengers and feasible trains.Moreover,the calculation method of passenger flow statistical indicators in different angles is proposed.(3)Research on the bottleneck identification and congestion evolution analysis method.From the perspective of capability and service,the model of bottleneck identification is constructed to identify the areas with insufficient capacity or low operation service level in the network.Based on the theory of spatiotemporal autocorrelation,the analysis method of temporal variation and spatial distribution of network unit spatiotemporal attributes is proposed.(4)Empirical case study.To verify the validity of the methods on passenger-to-path estimation and passenger-to-train estimation,an empirical case study on a domestic urban rail transit network is carried out based on one working day AFC data and train timetable data.The spatiotemporal distribution law of bottlenecks and the spatiotemporal congestion evolution law of passenger flow are analyzed.