Research On Scheduling Period Of Dynamic Vehicle Scheduling Problem

How to set scheduling period and scheduling time in the dynamic vehicle scheduling problem is of great significance for the logistics enterprises to improve the vehicle loading rate,reduce logistics cost and increase profits.This paper studies the scheduling period of dynamic vehicle scheduling problem in virtual logistics platform,and establishes an open dynamic pickup and delivery mathematical periodic scheduling model with hard time window based on long-short-term memory neural network(LSTM).A parallel algorithm framework based on Tabu search and cluster splitting is designed to solve this model.The specific contents of this paper include:1.Dynamic adaptive periodic scheduling model based on LSTMIn order to deal with the dynamic characteristics of vehicle scheduling problem in virtual logistics platform and improve the adaptability and flexibility of periodic scheduling,this paper proposes a dynamic adaptive periodic scheduling model based on LSTM,which can predict the number of orders generated in the future and dynamically adjust the scheduling period parameters.Finally,combined with the dynamic vehicle scheduling problem,this paper constructs an open dynamic pickup and delivery periodic scheduling mathematical model with hard time window2.Design of a parallel algorithm frameworkFor solving the model,this paper proposes a parallel algorithm framework based on a Tabu and clustering splitting operator.In the algorithm framework,the insertion algorithm proposed by Solomon is improved to construct the initial feasible solution at scheduling time;The clustering splitting operator based on vehicle path similarity is designed for splitting the initial solution into multiple sub solutions,which Tabu is used for solving the sub solutions to accelerate the solving speed,improve quality of solution;Finally,in order to subject to the dynamic pickup and delivery constraints,six neighborhood search operators such as delivery point relocation,order relocation and order exchange used in Tabu are modified in this paper.3.A new data set for dynamic vehicle periodic scheduling problem based on Li&Lim data setThis paper modifies the Li&Lim data set,which each order record adds "order time".And the number of orders follows from Poisson,uniform and normal distribution,which fits the actual statistical law.Finally,the effectiveness and performance of the parallel algorithm framework designed in this paper is verified under the new data set.4.Verification of the dynamic adaptive scheduling period model in the actual scenarioIn order to verify the dynamic adaptive scheduling period model based on LSTM proposed in this paper,the actual data of the UberEats are used.The experimental results show that the profit of the dynamic adaptive scheduling period model is 6.25%higher than that of the fixed time scheduling period model,and 6.63%higher than that of the dynamic event driven scheduling period model,which verifies the application value of the dynamic adaptive scheduling period model in actual scenarios.