Research On Dynamic Optimization Of Demand-response Bus Routes Considering Short-term Passenger Flow Forecasting

In the recent years,,the attraction of regular bus routes with fixed stations has decreased significantly with the improvement of urban residents’ demand for bus comfort and service level.The Demand Response Transit(DRT)becomes more and more popular due to flexibility and short-term reservation.Compared with the regularbus,DRT can meet the traveller’s needs of "door-to-door" and "real-time/reservation".It is similarto "carpool" bus,but has the advantages of being cheaper than taxis and more flexible and convenient than regular bus.However,DRT operation agencies still mainly rely on human experience without systematic methods and technical guidance when facing a series of decision-making issues,such as bus location selection,demand and supply matching,bus route design and vehicle scheduling.Most existing research cannot be applied to complex real-world public transportation systems.Therefore,this paper studied the travel characteristics of DRT and developed the short-term passenger flow prediction method by analyzing the differences between DRT and regular bus,and a hybrid site selection model is constructed.And then,the DRT route optimization and vehicle scheduling model is constructed by considering the influencing factors,such as passenger demand,vehicle carrying capacity,fleet size and service time.Finally,due to the fact that the vehicle capacity limit too small to serve all reserved passengers at one time,this paper proposed areservation demand splatted method to construct the path optimization model with the solution of a variable neighborhood search algorithm.The main content of this study is listed as follows:(1)Short-term prediction of DRT feeder passengerBy comparing the travel data of regular bus and DRT,one can identify the basic attributes,travel distance,travel time and other travel characteristics of DRT passengers.Combined with the questionnaire survey on the willingness to use DRT,a combination of gray model and artificial neural network model is constructed to predict the DRT short-time feeder passenger flow,which provides the basis for DRT station locating and route optimization.(2)DRT hybrid station locatingTaking into account the operational flexibility and cost control requirements,the DRT stations are divided into fixed stations and random ones.The principle of fixed stations as the bone and random stations as the feeder was proposed for DRT locating design,and correspondently the hybrid station locating model was developed.The results show that the proposed method can effectively save operation and passenger costs compared with the traditional method.(3)Collaborative optimization of DRT route selection and vehicle scheduling considering time-varying travel timeConsidering passenger demand,vehicle capacity,fleet size,service time and other influencing factors,this study proposed a DRT route design and vehicle scheduling cooperative optimization model,and a heuristic solution algorithm based on the gravity model.Meanwhile,considering the real-time changes of passenger demand and travel time between stations,the real-time demand information collection and result refresh mechanism were embedded in the offline model to build a collaborative optimization scheme with route selection and vehicle scheduling by considering time-varying travel time,which can effectively improve the applicability and effectiveness of DRT service.(4)DRT route optimization considering demand splittingThe DRT route optimization problem with demand splitting was investigated in this chapter,and the objective is to minimize passenger travel cost and enterprise operation cost,and the constraints include passenger travel time window,maximum passenger time and maximum vehicle travel time.The results show that demand splitting can effectively improve the usage efficiency of occupied vehicle and reduce the capacity investment and operation cost.The outputs of this paper on DRT can not only enrich the theoretical methods of demand forecasting,mixed station deployment,route optimization and vehicle scheduling in the field of public transportation,but also assist public transport agencies to provide efficient,convenient,and reliable customized travel services to passengers while taking into account the operating costs.