A Grid-based goods-to-person Order picking System
Project Type: Research
LoDI Team: Kevin Gue, Mina Shekari
Period: August 2018 – August 2020
Overview
The growth of e-commerce has transformed the role of warehouses from repositories of replenishment stock for retail outlets to high-throughput, rapid response order fulfillment centers for end users. In this research, we propose a new type of automated goods-to-person order picking system, composed of unit-sized conveyors capable of transferring item totes to process individual orders.
Problem Definition
Order processing is a challenging task for e-commerce. This complexity is due to processing individual orders for a large assortment of items with changing popularity. The existing order picking system designs do not give much flexibility to address the order picking problems in such a dynamic environment. Moreover, for the need of accommodating many SKUs, the desired system provides the infrastructure for high space utilization. A decentralized puzzle-based goods-to-person order picking system design is a solution to address all these concerns in one shot. The basic expected functionality from such a system is to bring requested items to any workstation(s), where those items are required to process the individual order(s) under process. Besides, any restrictions on defining the number of workstations and their locations on the perimeter of the grid are not accepted, as they limit the overall utilization out of the system. More specifically, in this research we address the following questions: 1) How to design a decentralized grid-based goods-to-person order picking system to overcome the limitations of existing systems to move the items in any directions? 2) What are the design factors and their impact on the throughput? 3) How the layout choice will be impacted by changes in operational decisions to batch the orders?
Methodology
We develop a multi-agent simulation model to evaluate our proposed design. The routing of the requested totes is executed using an online negotiation-based decentralized procedure in GridHub by Gang Hao. To serve the certain concerns and objectives of an order picking system, we develop the system architecture to preserve and serve the pick requests in a decentralized way, assuring a deadlock-free order picking process. We further design a set of experiments to evaluate the impact of certain design factors, their interactions, and different policies to process the orders, focusing on the commonality of orders processed in the workstations. We define the throughput as the response variable and analyze the impact of all these decisions on layout design. We model the system in AnyLogic 8.3.3, a multi-agent-based simulation package.
Results and Impacts
We demonstrated the proposed design by developing a multi-agent simulation model. We analyzed the impact of different design factors on throughput. One interesting result is that a 77 increase in throughput is possible if the orders processed by different workstations have no SKU commonality. The degree of SKU’s commonality also affects the choice of layout. This design provides a great infrastructure for developing dynamic order picking systems, adding a bulk storage area into consideration.