Joint Order Selection, Allocation, Batching and Picking for Large Scale Warehouses

TL;DR

This paper introduces a new large-scale batching problem formulation for warehouse order picking, incorporating order selection, and demonstrates that including order selection significantly improves solution quality, though it increases computational complexity.

Contribution

It proposes a novel batching problem formulation for large-scale warehouses that integrates order selection, enhancing realism and optimization potential.

Findings

Basic greedy algorithms are computationally intensive for real-world instances.

Including order selection improves solution quality.

Order selection has large potential for optimization improvements.

Abstract

Order picking is the single most cost-intensive activity in picker-to-parts warehouses, and as such has garnered large interest from the scientific community which led to multiple problem formulations and a plethora of algorithms published. Unfortunately, most of them are not applicable at the scale of really large warehouses like those operated by Zalando, a leading European online fashion retailer. Based on our experience in operating Zalando's batching system, we propose a novel batching problem formulation for mixed-shelves, large scale warehouses with zoning. It brings the selection of orders to be batched into the scope of the problem, making it more realistic while at the same time increasing the optimization potential. We present two baseline algorithms and compare them on a set of generated instances. Our results show that first, even a basic greedy algorithm requires significant runtime to solve real-world instances and second, including order selection in the studied problem shows large potential for improved solution quality.