Research On Demand Prediction And Repositioning Problem Of City's Free-floating Bike Sharing System

With the progress of society,the development of economy and the acceleration of urbanization process,the level of urban traffic motorization has been constantly improved,resulting in urban traffic congestion,serious air pollution,parking difficulties and other problems.As a new form of public bicycle,free-floating bike sharing facilitates residents' travel,alleviates urban traffic pressure,solves the "last mile" problem,and has gradually become an indispensable part of the urban transportation system.However,the skyrocketing development of free-floating bike sharing also brought a series of challenges.Especially,the tide phenomenon of rush hours passenger flow leads to the unbalanced distribution of bike sharing system in time and space,resulting in such problems as “difficulty in lending bikes”,“occupying public transportation space” and other issues.The fundamental reason is the lack of scientific and efficient bike repositioning.Based on this,this paper takes free-floating bike sharing as the research object,researches the demand prediction method in the rush hours of weekdays and repositioning optimization problem,which has extremely high theoretical value and practical significance.Firstly,on the basis of summarizing the definition,development,technical characteristics,function orientation and existing problems of free-floating bike sharing,and combining with the statistics and analysis of historical data of Xi'an Mobike,this paper analyses the travel characteristics of free-floating bike sharing from five aspects: user features,basic travel characteristics,travel time distribution characteristics,travel spatial distribution characteristics and the influence of external factors.Significant differences in user travel patterns were discovered during weekdays and weekends,during rush hours and off-rush hours on weekdays.The concept of traffic zone and service node was defined according to the characteristics of random parking.The traditional BP neural network theory is briefly described,and demand prediction method of free-floating bike sharing in the rush hours of weekdays based on self-adaptive BP neural network with additional momentum is proposed.A practical example of Xi'an Mobike is predicted by using SQL server database and Matlab software.The results show that the prediction method is effective and feasible in predicting days with the same external factors.Secondly,the definition,repositioning modes and repositioning present condition of bike repositioning problem of free-floating bike sharing are expounded.A static repositioning model in the rush hours of weekdays is constructed to minimize the weighted sum of total demand loss and total repositioning time which considers the travel time of repositioning vehicles and the service time of loading and unloading bikes.Combined with the demand prediction results,the calculation method of repositioning demand is proposed.Finally,based on the introduction and comparison of common algorithms for solving bike repositioning problems,a self-adaptive simulated annealing ant colony algorithm is designed to optimize the repositioning path,and the principle,process and design ideas of the algorithm are introduced.The loading and unloading strategies of minimizing the demand loss of bike sharing system while reducing the total repositioning time is proposed.The calculation and analysis of Xi'an Mobike as a practical example verify the feasibility and correctness of the model and solving method established in this paper.