Optimization Of Passenger Route Dynamic Guidance Strategy In Urban Rail Transit Network

With the deepening of the urban rail transit network operation mode,the spatiotemporal imbalance of passenger demand leads to unbalanced distribution of network passenger flow,and the congestion problem in partial network is becoming more and more serious.In addition to increasing capacity supply,operators also need to regulate network passenger flow distribution from demand side,and passengers’ route choice is one key factor affecting network passenger flow distribution.This paper takes passenger route guidance problem in urban rail transit as an entry point and conducts research from three aspects: modeling of passenger route choice behavior provided with guidance information,dynamic deduction of network passenger flow distribution under condition of guidance information provided,and optimization of passenger route dynamic guidance strategy.This research provides theoretical support and auxiliary decision support for improving daily operation efficiency of urban rail transit and travel experience of passengers.Specific research contents are as following:(1)Analysis of passenger route guidance problem in urban rail transit.First,the current situation and future development direction of passenger flow guidance in urban rail transit are summarized and analyzed.Based on the characteristics of passenger route guidance problem,an abstract model of the passenger route dynamic guidance problem is constructed under the Markov decision process framework.By analyzing key elements in the abstract model,it is pointed out that three sub-problems need to be researched to solve the problem,which are modeling of passenger route choice behavior provided with guidance information,dynamic deduction of network passenger flow distribution under the condition of guidance information provided,and optimization of passenger route dynamic guidance strategy.Based on the characteristics of three sub-problems,the research ideas for solving the problems are put forward to provide guiding program for the follow-up research.(2)Modeling of passenger route choice behavior provided with guidance information.By analyzing the characteristics of passenger choice behavior provided with guidance information,the existing Random Regret Minimization(RRM)model is improved.A route choice model based on comparison of route attributes which considering passenger attribute perception,scale effect,and decision inertia is established to describe passenger route choice behavior provided with guidance information.The basic attributes and willingness to subscribe information service are collected by SP(Stated Preference)survey and SP experiment is designed based on optimal orthogonal design method to construct the route choice scenario to collect route choice result data.By conducting Logistic regression between respondents’ basic attributes and willingness to subscribe information service,the factors that influence willingness to subscribe information service are analyzed.The parameters of the improved RRM model are calibrated using maximum likelihood estimation method based on route choice result data.Compared with the classic RRM model and the classic Random Utility Maximization(RUM)model,the results show the route choice model proposed in this paper can describe the passenger route choice behavior provided with guidance information better.(3)Dynamic deduction of network passenger flow distribution under condition of guidance information provided.Considering the characteristics of network passenger flow distribution of urban rail transit network and the requirements of optimization of passenger route dynamic guidance strategy,a discrete event system model of urban rail transit system is constructed.Entities such as infrastructure,trains,passengers,and operator in the urban rail transit system are modeled from mesoscopic level.Key elements such as input,output,state transition,feature tuples,and decision tuples of entity model are given.The coupling relationships among entity models are determined and the coupled model of urban rail transit system is established.Based on the state transition process and interaction of entities in system,the dynamic change of the network passenger flow distribution under condition of guidance information provided can be deduced.A network passenger flow distribution dynamic simulation system is developed based on the discrete event system model.By simulating the train operation process,the passenger decision-making and travel process provided with guidance information,and route guidance activity of operators,network passenger flow distribution under condition of guidance information provided is obtained,which provides a basis for the optimization of passenger route dynamic guidance strategy.The ability to deduce dynamic network passenger distribution is verified by practical case.(4)Optimization of passenger route dynamic guidance strategy.The optimization of passenger route dynamic guidance strategy is researched under the framework of reinforcement learning.Network passenger flow distribution feature is regarded as the state,the route guidance activity of operator is regarded as the action,network passenger flow distribution feature after route guidance is regarded as the state transition,and the reward value for the action of launching route guidance information is calculated based on state transition.A network passenger flow distribution feature extraction method based on Autoencoders,a guidance information launching action generation method based on guidance direction,and a reward value calculation method based on multi-objective weighting are proposed.Based on the Deep Deterministic Policy Gradient(DDPG)algorithm,the passenger route dynamic guidance strategy learning network is constructed,and the interactive data between simulation system and learning network are used for network training to obtain optimized passenger route dynamic guidance strategy.Case studies are carried out respectively under a small-scale hypothetical network and a largescale actual network.The results show that the employment of passenger route dynamic guidance strategy to guide passenger flow can alleviate network passenger flow congestion problem effectively.By changing the relevant parameters in the algorithm,the influence of the proportion of subscribing information services and the weight of regulation goal on the effect of route passenger route guidance is analyzed.