Spatial and temporal optimization for smart warehouses with fast turnover.
College of Mechanical Engineering, Zhejiang University of Technology, Hangzhou 310023, China
Key Laboratory of Data Analytics and Optimization for Smart Industry (Northeastern University), Ministry of Education, Shenyang 110819, China
School of Business Administration & Institute of Behavioral and Service Operations Management, Northeastern University, Shenyang 110167, China
Smart warehouse;Order splitting;Spatial and temporal heuristic;Order kitting;Column generation
Computers & Operations Research
With the rapid development of e-commerce and the new retail, the same-day or even same-half day delivery service is provided to compete for market share. Recently, an increased number of e-commerce companies implement the unmanned smart warehouses to improve the logistics efficiency. In order to further reduce the demand response time, a novel picking strategy is designed to firstly split the orders, and then assign the partial orders to different pickers. After all the order segments have been collected, it is shipped to the customer. Due to the inherent complexity of the problem, a two stage optimization model is introduced. In the first stage, an order splitting and batching strategy based on spatial measure is proposed. And a MIP model is constructed to minimize the total picking distance, which is then solved via a column generation based algorithm. In the second stage, the newly formed batches are considered as a priori inputs, which are then assigned to automatic pickers. The picking process is modeled as a special parallel machine scheduling problem with multiple due dates for a single item, which could reduce to a customer order scheduling problem on parallel machines, and it is unary NP-complete. A heuristic method is proposed to obtain an approximate solution. Although the order splitting technique is not new in the logistics industry, the split orders are picked according to a vehicle routing problem, which fails to address the kitting issue for fast turnovers. In the numerical analysis section, the proposed algorithm is validated through extensive testing on various scales of instances. It is observed that the optimality gap for our algorithm is within 9%, and the computation time is around 5 min. Also, the average turnover rate increases by approximately 50% in comparison with the no-splitting policy. In most cases, the average order tardiness decreases by 90% compared with order splitting and no-kitting policy. (C) 2020 Elsevier Ltd. All rights reserved.