Research On Mean Field Theory And Closed Queueing Network Of Bike Sharing Systems

The development of information technology and the popularization of smart phones provide a platform for the development of sharing economy,and the development of big data and Internet of Things promotes the implementation of sharing economy.At present,the sharing economy has involved many fields such as medical treatment sharing,diet sharing,space sharing,transport sharing,education sharing,etc.among which,transport sharing is the earliest and most mature field.This dissertation takes the bike sharing system in the transport sharing mode as the research object.The bike sharing system is an important part of urban intelligent transportation,which provides a flexible and effective supplement for the first / last kilometer of urban public transportation system,and also an effective way to choose green travel mode and practice the low-carbon life.This dissertation comprehensively applies the theories and methods of stochastic process,mean field theory,queueing network and RG decomposition to establish queueing models of bike sharing systems,calculate the steady-state probabilities of the systems,and evaluate and analyze the performance.The main results of this dissertation are as follows:(1)This dissertation applies the mean field theory to study the large-scale bike sharing system.In view of large-scale bike sharing system,this dissertation proposes a stochastic model analysis method combining the mean field limit theory and the non-homogeneous queueing model.The multi-dimensional and high-order complex bike sharing system is transformed into a low-dimensional and low-order stochastic model problem.The non-homogeneous queueing system is established by using the mean field theory,and then steady-state probability of the system is obtained.Based on the steady-state probability of problematic stations,the probability analysis is carried out to evaluate the performance of large-scale bike sharing system.(2)This dissertation establishes the mean field queueing model of the bike sharing system considering the morning and evening peak.Based on the tidal phenomenon of customers' renting and returning behaviors in the bike sharing system,aiming at the large-scale bike sharing system,a queuing model of the bike sharing system under the Markovian environment is established.Combined with the mean field theory,non-linear queueing and nonlinear birth and death process,the mean field matrix analysis method is proposed,so as to realize the analysis of customers' renting and returning behaviors in the morning and evening peak periods and the performance evaluation of large-scale bike sharing system.(3)This dissertation establishes the closed queueing network model of bike sharing system with Markovian arrival process and irreducible path graph.A closed queueing network model is established by viewing bikes as the objects and stations and roads as nodes.The Markovian arrival process of customers is introduced to express the non-uniformity of bike borrowing and returning in the morning and evening peak period,and the irreducible path graph is introduced to weaken the requirement of network structure of the bike sharing system.The product form solution of closed queuing network is obtained,which provides the calculation basis for the performance evaluation of the system.(4)This dissertation proposes a combined strategy of batch removal of unusable bikes and batch redistribution of repaired bikes.Considering the existence of unusable bikes in the bike sharing system,two batch strategies are proposed: removing unusable bikes in batch from parking regions to the maintenance workshop,and removing repaired bikes in batch from the maintenance workshop and redistributing them to parking regions.Due to the introduction of these two batch strategies and considering the failure process of usable bikes and the maintenance process of unusable bikes,the parking regions and maintenance workshop are two-dimensional Markov processes,which expands the characteristics of nodes in closed queueing networks.The two-dimensional Markov processes of block structure are calculated by RG decomposition,and the nonlinear equations of relative arrival rate and routing matrix are obtained,thus the form of calculation and product solution of closed queueing network are extended.(5)Based on the real data of Hangzhou Bike Sharing System,this dissertation analyzes the operation characteristics of the system.According to the periodicity and inhomogeneity of customers' renting and returning behaviors,this dissertation calculates arrival rates and service rates of customers in multiple periods respectively.At last,this dissertation evaluates the performance of the system and puts forward specific suggestions.