Integrated Optimization Of E-retailing Order Splitting And Distribution For Front Warehouse Mode

The growing competition in the e-retailing industry has led to the adoption of the front warehouse mode by an increasing number of e-retailers.This mode involves building multiple small-scale warehouses closer to customers to provide timely and accurate distribution services.However,due to the limited inventory of these front warehouses,orders often need to be split into multiple sub-orders containing different types of products and delivered by different warehouses.Therefore,the integrated optimization of order splitting and distribution is crucial to enhance order fulfillment efficiency and reduce associated costs in this mode.The delivery time window promised by e-retailers varies from very tight(hourly or sameday distribution)to relatively loose(next-day distribution).In tight time windows,orders are split and delivered directly by the front warehouse.Conversely,order fulfillment costs can be further reduced for relatively loose time windows,and customer satisfaction can be enhanced through lateral transshipment between warehouses,consolidation,and packing before distribution.Considering this,this dissertation first investigates the integrated optimization problem of order splitting and distribution route;on this basis,the integrated optimization problem of order splitting and distribution route considering lateral transshipment and package consolidation and packing are studied,respectively.An integrated optimization method based on deep reinforcement learning is also studied to address the large-scale problem associated with the e-retailing industry.More specifically:To address the integrated optimization problem of order splitting and distribution route,an arc-flow model is developed which considers practical characteristics such as limited inventory,order splitting constraints based on product type and quantity,heterogeneous vehicle route planning constraints,and time window constraints.To solve this problem,we propose a novel method based on Dantzig-Wolfe decomposition and Benders decomposition that decomposes the arc-flow model by eliminating the coupling between the decision variables of order splitting and vehicle route planning and a branch-price-and-cut algorithm is designed to solve it.The algorithm’s efficacy is evaluated by comparing CPLEX and two heuristic algorithms.The results indicate that the proposed algorithm is superior in tackling relatively larger-scale problems.Furthermore,this dissertation investigates the effects of various factors,including overlapping inventory levels,heterogeneous vehicle combinations,and time window width,on the integrated optimization problem of order splitting and distribution route.To tackle the integrated optimization problem of order splitting and distribution route considering lateral transshipment,an arc-flow model and a set covering model are established.The models consider constraints such as limited inventory,order splitting,lateral transshipment based on product type and quantity,heterogeneous vehicle route planning,and time window constraints considering transshipment time.A heuristic branch-price-and-cut algorithm is proposed along with two logic-based Benders cuts and a heuristic accelerated strategy to address infeasible distribution routes.In the numerical study analysis,the algorithm demonstrates effectiveness in reducing the total fulfillment cost compared with CPLEX and three heuristic algorithms.The effectiveness of the proposed algorithm is evaluated by conducting a comparative analysis with CPLEX and three heuristic algorithms.The results indicate that the lateral transshipment of orders significantly reduces the total fulfillment cost,particularly when the inventory overlap level is low.For the integrated optimization problem of order splitting and distribution route considering parcel consolidation and packing,an arc-flow model and a set covering model are established,taking into account limited inventory,order splitting and lateral transshipment constraints based on product type and quantity,package packaging constraints for customer orders,heterogeneous vehicle routing constraints,and time window constraints considering package consolidation and packing time.A heuristic branch-price-and-cut algorithm is designed to solve the peoblem.An extended bi-directional label setting algorithm is proposed,which can generate feasible parcel consolidation and packing schemes with route planning.The algorithm’s effectiveness is verified in the numerical study analysis compared with CPLEX and two heuristic algorithms.The results indicate that when the inventory overlap level is low,the consolidation and packing of customer parcels considered in the integrated optimization problem can significantly reduce the total order fulfillment cost and the number of packages packed.Finally,for the integrated optimization problem of large-scale order splitting and distribution routes,a Markov decision process model is constructed.An actor-critic algorithm is designed to solve the problem.The strategy network is learned using the node2 vec strategy and attention mechanism to evaluate the quality of generated routes.The numerical study analysis demonstrates that the proposed Actor-critic algorithm outperforms CPLEX and two heuristic algorithms in terms of solution quality and computation time.Furthermore,this dissertation conducts a deeper analysis of the impact of the node2 vec strategy and the utilization of different graphics cards during the training phase on the proposed Actor-critic algorithm.The results reveal that using the node2 vec strategy for node selection significantly enhances the algorithm’s performance in reducing the distribution cost compared to the traditional greedy strategy.This dissertation contributes to integrated optimization methods for multi-warehouse order fulfillment problems by enhancing the order splitting and distribution model for inventory and distribution management.The proposed model reflects the practical complexities of real-world scenarios and offers valuable insights for the collaborative optimization of inventory and distribution management in distribution centers.The findings of this research can benefit decision-makers and operators in the e-tailing industry by providing them with a framework to evaluate order splitting schemes,lateral transshipment strategies,parcel packing,and distribution schemes and make informed decisions.