Dynamic Modeling For Passenger Flows On Congested Metro Routes

With the development of public transportation in big cities,metro has become the main means of transportation for people to travel.However,the increase in population has brought tremendous pressure on metro operations.Overload operation greatly reduces people’s comfort while also posing a great safety risk to people’s travel.In order to quantitatively characterize the passenger flow on a crowded metro route,this thesis focuses on two aspects: the probability of passengers boarding the train and the probability of being left behind.On the one hand,the probability of passengers boarding the train is an important indicator for studying metro passenger flow.Compared with no-transfer routes,the ridership of passengers under transfer routes is more complex and diverse,which is a major difficulty in the field of metro research.However,the automated data of transfer passengers alone does not enable accurate inferences to be made about the pre-transfer ridership of transfer passengers.Therefore,this thesis introduces the automated data of no-transfer passengers entering the station at the same time as auxiliary data to supplement the ridership information of transfer passengers before transfer.The dynamic model in the no-transfer case is extended to the transfer case to construct a dynamic model of the probability of transfer passengers to board trains.Furthermore,the expectation-maximization(EM)algorithm is used to calculate the unknown parameters in the model.The model realizes a systematic analysis of passenger flow of transfer passengers from the population and individual perspectives.Important indexes including boarding probabilities,transfer time,passenger-to-train assignment probabilities,and total travel time can be inferred.In this thesis,the actual data generated from passenger travel in a section of congested transfer routes of Beijing metro are applied to the developed dynamic model for case study.Eventually,the validity of the model was verified by cross-validation method.On the other hand,the probability of passenger left behind can be a useful description of the level of congestion in the metro system.The probabilities of passengers left behind are often related to the tap-in time.Therefore,this thesis proposes a methodology for inferring these dynamic probabilities on congested metro routes using automated data.First,the EM algorithm is used to compute the maximum likelihood estimators of passengers’ dynamic boarding probabilities.Second,a relationship between the dynamic boarding probabilities,the dynamic left behind probabilities and the behavior probabilities is established through a group of equations,which is further solved to obtain the dynamic probability of passenger left behind.Ultimately,a Monte Carlo simulation method is used to simulate the solved dynamic left behind probabilities,and it is applied to the actual data of passenger travels on a section of congested routes in Beijing metro for case study.