Order picking heuristics for online order fulfillment warehouses with explosive storage

Abstract

Fulfillment Warehouses (F-Warehouses) convert online customer orders into a delivery package. Key differences between a traditional and an F-Warehouse are (i) a very large number of small storage bins, (iii) an explosive storage policy that disperses inventory of each item to multiple bins, and (iii) commingled bin storage. The paper introduces the explosive storage picking problem, which is a special case of the traditional order picking problem. Since they stock items in multiple locations, there are multiple picking options for each item. A picklist for several items is generated from the current inventory dispersion and the objective is to minimize the picklist bin range. The problem is first formulated as an MIP and then three fast solutions, SpaceX MIP, H1, and H2 are presented. The H1 and H2 seed heuristics use weighted bin order fillability, as a surrogate for the probability a seed bin will generate short picklists. For small problems, SpaceX MIP found the optimal solution. Larger problems compared H1 and H2 with the SpaceX MIP solutions. On average, the H1 solution was 34% longer while the H2 solution was only 18% longer. The H1+2 strategy selects the shorter of the two solutions. For 62% of problems, H1+2 matched the SpaceX solution, and for only 17% of problems, the solution was worse than 20%. We tested H1+2 on an F-Warehouse simulator and it achieved 40% faster fulfillment relative to the benchmark.