Optimization Of Vehicle Scheduling Problem Under Urban Distribution System Based On Data Driven Methodology

With the rapid development of information technology and social industrial structure,the information construction of urban distribution system has improved a lot.The vehicle scheduling plan in distribution center needs to be formulated according to the different kinds of data analysis to adapt to the trend of "smart city" construction.In this paper,Data driven vehicle scheduling problem(DDVSP)is proposed.We explore the way of modeling and algorithms to provide decision support for daily operation in city distribution center.Firstly,we introduce relevant concepts of data-driven systems.Then subtotals 14 data which are related to the urban distribution system into distribution center data,customer demand data,road network physical data and intersection time data according to different source.After the data analysis and preprocessing,the comprehensive traffic impedance of the urban road sections in the distribution range is obtained.Based on the analysis of DDVSP,two modeling parts can be divided.The first stage modeling considers to single source shortest path finding in urban network,and it is used to calculate elements in time-distance matrix.The Second stage modeling considers two conditions.When in scheduling position,a planning model with multi-type vehicles goes.Alternatively,for analyzing feed-back information,"virtual customers" are drawing to build the dynamic arrangement model.According to the above model,we put forward the IAS-HSA two-stage strategy:(1)An urban-suited value function is founded to drive the new designing,improved A-Star algorithm,in order to find time-distance matrix solution,and its consistency can be proved.(2)design a hybrid simulated annealing algorithm based on multiple heuristic rules to solve the second phase model.We take one distribution center in Dalian Sha Hekou district as an example,set MA-0 benchmark examples and the solution of which is used to verify validity of IAS-HSA,and then found that improved A-star algorithm has lower driving time and steadier solution than Dijkstra signature improved algorithm,with the increasing of crossroads scale in network.Moreover,a vehicle schedule can be came up by the hybrid simulated annealing algorithm has 13.1%cost saver than actual operational schedule.Finally,four types of traffic situation test libraries are constructed,namely the main roads flow adding(test M),the main roads and vise roads flow adding(test MA),the region surroundings flow adding(test MA-C)and the banned road conditions(test MF).In the process of scheduling and real-time dynamic adjustment,the dynamic result of distribution cost and customer service punctuality rate are analyzed.