Simulation constitutes a crucial part of designing and operating logistics warehouses. The use of agent-based models (ABM) allows to incorporate human workers specifics in the modeling framework. Warehouse operate in a complex supply chain system and their operation should not be a bottleneck for other processes. Thus, warehouse operation must be predictable, repeatable and finally optimal. Order picking is the main process in any warehouse, being the most time consuming and therefore the most costly. This study addresses the issue of order picking process in the real scale warehouse. Optimization results are compared with the existing picking practice generated by implemented Warehouse Management System (WMS), and with common routing strategies, like Sshape or the largest gap. It is shown that evolutionary algorithm allows to improve the picking routing problem giving clear benefits.

Simulation constitutes a crucial part of designing and operating logistics warehouses. An agent-based modeling (ABM) allows to capture various elements of such a system, which may include human workers, material-handling equipment and different kinds of autonomous subsystems. Warehouse operate in a complex supply chain system and must meet its requirements. Customers expect higher responsiveness, which translates into completion times shortening. Robotic systems are deterministic, while human-based picker-to-parts warehouses are not. This paper presents novel agent-based stochastic distribution center (DC) model, which uses distributional gradient boosting machine learning (ML) to introduce pickers’ uncertainties associated with human behavior. Obtained modeling approach is validated using full-scale DC environment and real-time warehouse data.