Distributionally Robust Optimization Research For Just In Time Distribution Based On Traffic Flow Prediction

Just-in-time logistics refers to the logistics service of direct end-to-end delivery without warehousing and transit.It matches real-time demand and real-time capacity by means of global scheduling to provide end-to-end small batch and multi-batch delivery service.Timeliness is the core of just-in-time logistics service.In the complex and changeable traffic conditions,the uncertainty of driving time would cause the original distribution scheme unreasonable or even infeasible.Therefore,it is necessary to optimize the distribution scheme and improve the execution of the distribution scheme under the uncertain travel time.It is of great significance for reducing the logistics distribution cost and improving the distribution efficiency.This paper focuses on the robust optimization problem of vehicle routing with uncertain travel time.In order to solve the problem of unreasonable selection of the uncertain set,this paper proposes a method to construct the uncertain set by integrating traffic flow prediction.And then,the distributinally robust optimization model is built.Next,the traffic flow prediction method and solution algorithm are designed.Finally,the effectiveness of traffic flow prediction and robust optimization of vehicle routing distribution are verified.The main contents of this paper are as follows.(1)The distributionally robust optimization method is introduced into the vehicle path optimization to solve the problem of difficult execution of distribution schemes caused by uncertain travel time.This paper proposes a method to construct the uncertain set by integrating traffic flow prediction.And the distributionally robust optimization model is constructed with the objective function of minimizing vehicle driving cost and penalty cost.The model is transformed into the corresponding equivalent form to be solved.(2)In order to improve the traffic flow prediction effect,a new directed spatiotemporal graph is constructed to characterize the relationship of road network structure.The definition of the relative proximity of nodes to represent the influence weight of spatial-temporal dimensions among nodes.The spatial dependence of traffic flow data is obtained by graph convolution,and the spatiotemporal dependence of traffic flow data is obtained by gating cycle unit.A traffic flow prediction model is established based on spatiotemporal graph to predict the traffic flow in the road network.(3)Robust optimization algorithm for vehicle path distribution is designed to solve the distributionally robust optimization model.An improved optimization algorithm with local search strategy which is introduced to node fitness function is designed to solve it.