Research On Space-Time Partitioning And Evolutionary Fractal Scheduling For Public Bicycle System

As an extension and supplement of urban public transport,public bicycle system(PBS)can solve the "Last Mile" problem of public transport and improve the overall service level of urban transport effectively.At present,the main problem in the PBS operation is the recurring phenomenon of "no bicycle available for renting" or "no vacant stub for bicycle returning" at PBS self-service stations.Therefore,it is of great significance to analyze the service state of each PBS station both in time and space to obtain operation law of PBS stations and then design a reasonable bicycle redistributing scheme to optimize the number of bicycles at each station,so as to solve the "renting/returning difficulty" issue in PBS operation.In addition,Free-floating bike sharing system has developed rapidly in recent years.Although it is more convenient without fixed parking piles,but random parking of shared bicycles has brought great potential hazards to traffic.Therefore,it is necessary to delimit the "electronic fence" for parking of shared bicycles.The amount of bicycles in “electronic fence” is limited,so that the "renting/returning difficulty" issue arises,which need to be solved by scheduling.Therefore,reasonable PBS scheduling scheme can be extended to the Free-floating bike sharing system to improve its service quality.Firstly,in order to abstract the operation law of PBS stations,a spatial analysis method for the self-moving characteristics of public bicycles is proposed.Taking Hangzhou PBS as an example,the five years historical data of PBS operation are mined in time and space.By analyzing the spatial characteristics,time-varying characteristics and OD correlation characteristics of bicycle renting/returning,the travel rules of public bicycle users and the macro/micro operation laws of each station are obtained.These laws are reference data and decision-making basis for follow-up studies such as the acquisition of scheduling time range,the division of scheduling area and the design of scheduling model.Then,for the time and space domain partitioning of PBS scheduling,the acquisition method of scheduling time range based on PBS self-moving model and the partitioning method of self-balanced scheduling region based on fractal tree are proposed.The acquirement of scheduling time range includes calculating dynamic occupancy-capacity ratio threshold,obtaining dynamic time ranges of PBS positive/negative scheduling and generating the macro scheduling time ranges.The partitioning method of self-balanced scheduling region proposes the fractal tree based self-balanced partitioning algorithm(FSPA)and forms multi-level selfbalanced regions by clustering PBS stations layer by layer.The bicycle demand of the nodes being clustered must be complementary to each other.It makes the bicycle redistributing in each region reach self-balance to reduce cross-region scheduling.An empirical study was conducted in Hangzhou PBS,the results show that the obtained dynamic scheduling time ranges are more precise and help reduce the scheduling frequency of PBS,the divided self-balanced scheduling area achieves a fast and low-cost scheduling for PBS and helps improve the operating efficiency and service quality of PBS.Next,the partitioning scheduling problem of PBS is defined as a dynamic cooperative scheduling management problem in complex environment.A new swarm intelligence method named “evolutionary fractal” is proposed which combines the self-similarity of fractal theory with co-evolution mechanism.According to the fractal characteristics of PBS partitioning scheduling,the evolutionary fractal method simplifies the complex PBS partitioning and hierarchical scheduling into the superposition and integration of fractal agents,and controls the PBS scheduling process via scheduling rules designed based on L system.A multi-objective dynamic scheduling model considering time satisfaction and scheduling cost is established and solved by improved adaptive genetic algorithm,so as to obtain the optimal scheduling scheme.Experiments show that the scheduling model of PBS fractal agent simplifies the complexity of PBS scheduling,and the improved adaptive genetic algorithm enhances the optimize performance of the algorithm,so that the final dispatching vehicle route is significantly shortened and the dispatching cost is reduced.Finally,on the basis of the previous data analysis and scheduling method,a GIS-based visualization intelligent scheduling software is designed and developed to support a series of functions such as station state alarm,scheduling route planning and so on.The results show that the bicycle inventory of each station and the route planning of scheduling vehicle can be optimized.It helps reduce the scheduling cost and improve the PBS management efficiency,and then alleviates the "renting/returning difficulty" problem in PBS.