Passengers’ Semi-compensatory Route Choice Behavior Analysis In Urban Rail Transit Network

The operation of new lines enlarges the network scale of the urban rail transit and intensifies the complexity of passenger’s pre-trip route choice decision since many OD(Origin-Destination)pairs have amounts of routes.As an important part of travel behavior analysis,route choice behavior analysis can provide passengers with personalized route planning service and provide the operational departments with transfer,sectional and line flows data which can support train schedule making,passengers flow organization and control,fare clearing approach,network optimization,evaluation of the impacts of new lines on existing lines,etc.Route choice is a process in which in order to achieve the travel purpose,a passenger chooses the best route from the routes set via comprehensively considering all factors in a specific choice context.Focusing on the mechanism of passenger’s route choice and all elements in route choice process,this paper aims at route choice modeling to analyze passengers’ route choice behavior.(1)Analyze network topology,operational,survey and smart card data and extract factors influencing route choice behavior.The difference between official map and actual map improves the impacts of angular cost.This paper revises the traditional angular cost to better describe the impacts of route direction on route choice behavior.And in-vehicle travel time,number of transfers,transfer time and congestion level are the main factors influencing route choice behavior extracted by Pareto chart based on surveyed data.(2)Analyze the limitation of some factors to reveal the non-compensatory behavior in effective routes set generation process and develop effective routes set generation approach to reflect the fact that only the route in the limitation can be considered as an effective route.Comprehensively considering the limitations of travel time and number of transfers,route tolerance degree function is developed to measure the consideration degree of the route.In order to avoid the difficulty in endogenous estimation on all parameters both in the route tolerance degree function and route utility function,a servey plan about the limitation and tolerance is provided,and besed on the surveyed data,the parameters in the route tolerance degree function are estimated and the estimations are reliable.(3)Considering the internal relationship between effective routes set generation and choosing the best route from the set processes,based on CMNL(Constrained Multinomial Logit)model,this paper studies the effects of OD scales and route over-lapping problem in semi-compensatory route choice context.Ultimately,the SCPSL(Scaled Constrained Path Sized Logit)model which belongs to one-step semi-compensatory Logit models is developed together with the exogenous and endogenous estimation approaches.It avoids the estimation bias from the isolation of the two processes and avoids the computational difficulty of two-steps semi-compensatory model facing with large routes set.Semi-compensatory behavior reflects that when the attribute is within the boundary,route utility can keep stable by improving other factors if the attribute goes worse;otherwise,when the attribute is out of the boundary,route utility can’t be kept by improving other factors.The marginal substitution and elasticity analysis also show the semi-compensatory and OD varied characteristics of route choice behavior.Furthermore,the model is applied in travel demand forecasting which has good forecasting performance even when new lines put into operation.(4)Heterogeneity is modeled to reflect the differences among passengers in semi-compensatory route choice process based on SCPSL model.Semi-compensatory Mixed Logit model and semi-compensatory Logit model with latent classes are established respectively.However,Mixed Logit model focuses on individual differences with little consideration of passenger’s group characteristics,referring to the similarity more than the difference among some passengers,which leads to the mixture distribution instead of a pure distribution;while the latent class model assumes that passengers’ behaviors are the same in the same segment which negelects passenger’s heterogeneity in the same segment.Therefore,combining the two types of models,semi-compensatory Mixed Logit model with latent class is proposed by this paper.Meanwhile,according to different psychological tendencies to route direction deviation,three types of angular cost formulations are established which can be regarded as three latent segments.Within each segment,parameters are assumed to follow some specific distributions.In this way,LC-Mixed SCPSL model which is a kind of semi-compensatory Mixed Logit models with latent class is established and the estimations show the superiority of the proposed model over other models.Moreover,the model is applied in personalized route planning service.(5)Travel time reliability,referring to the variation in travel time,has been considered into semi-compensatory route choice modeling and a mean-standard deviation SCPSL model is proposed to reflect the impacts of randomness of travel time on route choice preferences.By smart card data mining,the distribution parameters and the truncated properties of the entry and exit walking,waiting,in-vehicle travel and transfer times are acquired.Then,based on the RP(Revealed Preference)data,smart card data and the combination of the RP and smart card data are used respectively to estimate the proposed mean-standard deviation SCPSL model.The RP data has small sample size without the distribution parameters of various times,but has the actual travel routes and individual characteristics;while smart card data has large sample size with the distribution parameters of the times,but without the actual travel routes and individual characteristics.By combining the two types of times,the estimation results are the best.At last,the model is used in fare clearing approach.