Research On Location Decision And Joint Scheduling Optimization Of Car Sharing System

By the end of 2016,the number of people who has license without car is 200 million.The car sharing in China has huge market potential.Car sharing is not only convenient,but also low-cost due to cost-sharing.In addition,car sharing can reduce car ownership,can be regarded as an effective means to solve traffic congestion and environmental pollution.However,due to the short development time of car sharing in China,there is few scientific and practical guidance in design and operation.This paper mainly aims at optimizing the location of car sharing stations and the joint schedule of vehicles and staff,and provides theoretical suggestions for the development of car sharing.In the aspect of car location,the taxi's GPS data is processed to identify the taxi passengers' boarding and alighting position,which is then clustered by DBSCAN.The clustering of space cluster is regard as the location of car sharing system.The results show that the DBSCAN algorithm has more candidate stations in the central area of the city,and the service of each station is balanced relatively.While K-means clustering algorithm is more concentrated,fewer stations in the city center,resulting in the high service pressure,so this paper selects the DBSCAN clustering results as a site selection scheme.In the aspect of car sharing operation,this paper establishes the joint scheduling optimization model of the car sharing system,and transforms the model into a bi-level programming model.The upper model is an open-loop mTSPTW model with uncertain number of cars.While the upper model is an open-loop mTSPTW model with uncertain number of person.Finally,a genetic algorithm is designed to solve.Finally,Shanghai city is selected as the research object,the location gained by DBSCAN algorithm is used to construct the scheduling network.Then,25 pairs of ODs are randomly selected from the OD database of the taxi.On this basis,the optimization model of car sharing scheduling is solved,and the optimization results,including the configuration of vehicles and person,and the dispatched paths are analyzed.Then for changes in vehicle demand and vehicle cost,this paper made a sensitivity analysis.Finally,this paper use genetic algorithm 10 times.The performance is compared with the simulated annealing,particle swarm optimization algorithm,and the result verify the genetic algorithm is of great stability and effectiveness.