Usage Patterns And Hierarchical Scheduling Of A Bicycle-sharing System

With a rapid increase in urban motorization level,a series of traffic issues such as congestion,accidents and emissions have become increasingly severe.Developing a multi-mode,multi-level integration of public transport is widely regarded as an effective measure to alleviate those issues.As one type of public transport mode,bicycle-sharing system(BSS)has become a prominent feature of the public transport network in many cities.Along with the boom of BSS,cities face the challenge of bicycle unavailability and dock shortages.This reduces people's enthusiasm to use and also hinders the healthy development of the system.Therefore,this study is developed and it is sponsored by the National High Technology Research and Development Program 863(No.2014AA110303).Based on IC card data mining,the usage patterns,hierarchical scheduling strategy,system architecture and function module,scheduling demand,vehicle routing model and algorithm are explored,aiming at developing a complete set of methodology to support daily operation of real-life bicycle-sharing systems so as to improve users' satisfaction and operating efficiency,reduce operating cost.Firstly,based on the analysis of IC card data,meteorological data,etc.,this paper studies factors influencing the use of bicycle-sharing systems by correlation analysis method.Users' travel patterns and stations' rental patterns are then statistically inducted using Trip-based data and Station-based data from time-space aspect.Secondly,based on usage patterns analysis,the scheduling types and modes are analyzed to determine the hierarchical scheduling strategy.Under the guidance of this strategy,the architecture of the scheduling system is established.Function modules including station monitoring module,data cleaning and analysis module,unusable bicycle recognition module,scheduling demand analysis module and scheduling scheme generation module are proposed.Thirdly,the prediction methodology of static and dynamic scheduling demand is proposed.Based on the analysis of accumulated net value of pickups and returns at each station,the method to deduce the optimal initial bicycle supply is put forward to calculate the static scheduling demand.Based on the rental patterns at each station,the methodology combining cluster analysis,a back-propagation neural network(BPNN),and comparative analysis to predict users' short-term demand is proposed to calculate the dynamic scheduling demand.Finally,under the situation that static and dynamic scheduling demand are determined,a single-objective(minimize the travel cost)Set-Partitioning model with capacity and time constraints,and a multi-objective(minimize the travel cost and maximize users' satisfaction)Set-Partitioning model with capacity constraints are established for the static and dynamic scheduling of a bicycle-sharing system,respectively.The solution algorithms are the designed base on Branch-and-Price,Shortest Path Faster Algorithm and Neighborhood Search Algorithm.A case study is conducted to evaluate the performance of the proposed methodology.The feasibility and validity of the model and algorithm are verified.