Walk less, pick more: choosing optimal batches of orders in a warehouse

Abstract

Most warehouses employ a picker-to-parts strategy, where humans (termed pickers) traverse the warehouse to collect items. It is common for pickers to collect items for several orders at once. Such a set of orders is called a batch. Two optimization problems arise from this strategy. The picker routing problem refers to finding routes through the warehouse to minimize the distance traveled. The batching problem refers to selecting a combination of orders that minimizes the distance traveled. These problems are the focus of this thesis. To solve the optimization problems in the context of a real-world warehouse, a graph model representing a warehouse was created. In addition, a model was created for representing orders and batches as sets of nodes in such a graph. Additionally, a collection of algorithms was designed to solve the optimization problems. The models and the algorithms were implemented in code in the form of a library for the C# programming language. The library is accompanied by a suite of tests to help verify the correctness of the implementations. Furthermore, a suite of benchmarks was created based on real-world warehouse data supplied by the company Ongoing Warehouse. These benchmarks were used to evaluate the models and algorithms in terms of quality and runtime. Based on the evaluation, a recommendation was presented to Ongoing Warehouse of algorithms to use for integration into their warehouse management system.