Research On Site Demand Prediction And Repositioning Optimization Of Free-floating Bike Sharing

The urban traffic motorization has brought a lot of “urban diseases”,such as the problems of traffic congestion,severe haze and energy shortage and so on.Free-floating bike sharing(FFBS)comes into being as an innovative traffic pattern to alleviate urban traffic pressure and solve the “last mile” problem,which has characteristics of flexibility,green transportation,environmental friendly and good accessibility.However,travelers always are rejected by the lack of bikes or multiple failures of bikes in actual operation process due to its inefficient and unreasonable rebalancing operation.To the best of our knowledge,there isn’t any recent paper about rebalancing of FFBS.Therefore,this paper proposes a new and operator-based bike redistribution methodology that starts from the site demand prediction and ends with repositioning optimization of FFBS according to research theory and method of traditional bike sharing,which has important theoretical and practical significance.Firstly,this paper analyzes travel characteristics and its influence factors of free-floating bike sharing from aspects of time and space travel characteristics,external factors and user travel characteristics according to travel data in Beijing,which is based on the definition of free-floating bike sharing and its function.Secondly,based on time travel characteristics,site demand prediction method of free-floating bike sharing based on BP neural network is proposed.When demand prediction of different periods of one day with the same property is considered to be predicted,SQL Server is used to deal with data and MATLAB is used to be programme.The prediction errors of the whole prediction process are within 5% and results show that the prediction method is effective.Thirdly,based on the prediction result,from the aspects of multi-objective of minimizing the routing length and maxing satisfaction of users,bi-level models are developed,including bike-sharing repositioning optimization model without time window at off-peak period and bike-sharing repositioning optimization model with soft time window at peak period,which considers the importance of station according to different types of sites.Then LINGO10.0 is used to solve the small scale example to verify the accuracy of the model.Finally,hybrid genetic algorithm is designed to solve the massive examples and practical example of MOBIKE in Beijing is calculated to verify robustness and accuracy of the hybrid algorithm.And it is proved that repositioning optimization model considering site importance can improve service level and customer satisfaction.