Order batching in automated warehouses to reduce picking time
Research by Ramzi Hammami (Rennes SB), Yannick Frein (Grenoble INP) and Nicolas Lenoble (PhD student) has been conducted with company KLS Logistic (a developer of Warehouse Management Systems, WMS) in collaboration with several firms. We optimized the batching strategy of customers’ orders in automated warehouses with the objective of minimizing the picking time, that is, the time required to collect a given set of orders. While most companies adopt basic batching methods or simply rely on human experience, we proposed a mathematical optimization approach that exploits the nature of customer’s orders, the characteristics of the storage technology, and the capacity of the system.
Using firms’ real data, we demonstrated that our models outperform the current methods used by companies in practice and lead to considerable savings in picking time, which allows to quote shorter delivery times to customers. Our models have been integrated into the WMS software commercialized by our partner KLS Logistic and have been adopted by several companies in different sectors (healthcare, manufacturing, retailing).
The research was undertaken from 2014 to 2018 and is related to the PhD work of Nicolas Lenoble, co-supervised by Ramzi Hammami (Rennes School of Business, France) and Yannick Frein (Grenoble INP, France). This is an applied PhD (PhD CIFRE, Industrial Agreement in Training through Research) that has been conducted in company KLS Logistic, a developer of Warehouse Management Systems (WMS) (https://www.kls-group.fr/en/kls-group-en-home/). A WMS is a set of tools and processes intended to organize the work of a warehouse or distribution centre and to optimize the main KPIs such as the picking time and the customers’ service level. The research work is therefore a collaboration between a business school (Rennes SB), an engineering school (Grenoble INP), and KLS Logistic, and was co-funded by KLS Logistic and the French National Technological Research Association (ANRT).
The idea of the project stemmed from the need of KLS Logistic’s clients to reduce the picking time in their modern warehouses that use Automated Storage and Retrieval Systems (AS/RS) such as basically Vertical Lift Modules (VLMs) and Carousels. Such automated storage technologies are widely used in practice, especially by online retailers (e.g., Amazon, Leclerc Drive, Carrefour Drive, Cdiscount) and healthcare companies.
Reducing the picking time (i.e., the time required to collect a given set of customers’ orders from the VLMs or/and Carousels) is crucial as it allows the company to offer a shorter delivery time to its customers, which can attract more demand or/and justify quoting higher prices.
We first demonstrated that order batching, namely regrouping customers’ orders into batches and simultaneously collecting the orders belonging to each batch before moving to the next batch, constitutes a critical decision impacting the picking time. We then developed optimization models for customers’ orders batching in automated warehouses equipped with VLMs or carousels, with the objective of minimizing the picking time. While most companies adopt basic batching methods that are generally far from optimal (e.g., First In First Out) or simply rely on human experience (according to the observations of KLS Logistic and the feedbacks collected from the clients), we proposed a mathematical optimization approach that exploits the nature of customer’s orders, the characteristics of the storage technology, and the capacity of the picker. We applied our methods while using real data of different companies (which are operating the WMS software of KLS Logistic) and showed that our models outperform the existing approaches used by these companies with respect to the picking time (time savings range from 20% to more than 50%). Consequently, our models have been integrated into the WMS software commercialized by KLS Logistic, and have been adopted by several companies from different sectors (healthcare, manufacturing, retailing).
References to the research
[3.1]. Nicolas Lenoble, Yannick Frein, Ramzi Hammami (2018). Order batching in an automated warehouse with several vertical lift modules: Optimization and experiments with real data. European Journal of Operational Research 267 (3), 958-976 (https://doi.org/10.1016/j.ejor.2017.12.037)
[3.2]. Nicolas Lenoble, Ramzi Hammami, Yannick Frein (2021). Fixed and rolling batching for order picking from multiple carousels. Production Planning & Control 32 (8), 652-669 (https://doi.org/10.1080/09537287.2020.1751326)
Quality of the research: Outputs [3.1] and [3.2] are published in double-blind peer-reviewed international journals in the fields of operational research and supply chain management.
Paper [3.1] was published in the prestigious 4 stars journal EJOR (ABS, FNEGE, CNRS) under the department “Innovative application of operational research”.
Paper [3.2] was published in the 3 stars journal PP&C (ABS, FNEGE, CNRS), which focuses on real applications of supply chain and production models in practice.
In addition to the above journal papers, the results of this research have been published in the PhD report of Nicolas Lenoble (https://www.theses.fr/2017GREAI052) and in highly reputed international conferences including the IFAC 2017 (20th World Congress of the International Federation of Automatic Control), and the ILS 2016 (International Conference on Information Systems, Logistics and Supply Chain).
Details of the impact
Thanks to the rapid development of automated technologies, different types of Automated Storage and Retrieval Systems (AS/RS) have emerged and are being widely used in modern warehouses. Automated warehouses are increasingly used by industrial and retailers as they offer several advantages, such as space savings and increased productivity. In automated warehouses, the storage machine brings the products to the operator (picker), who stays in front of the machine and does not need to move from one storage location to another, as opposed to manual warehouses where pickers have to retrieve products by walking down aisles until they reach the location where the product is stored. Vertical Lift Modules (VLMs) and carousels are among the mostly used AS/RS in practice. In a VLM, products are stored in the front and rear tray locations. When a tray is requested, an extractor travels vertically between the two columns of trays, pulls the requested tray from its location and brings it to the picker. The carousel consists of a set of shelves (also called trays) mounted on a closed loop rail. Each shelf contains several boxes where products are stored. When a product is requested, the carousel rotates to present the corresponding shelf in the opening zone, from where the picker can retrieve the product. The rotation may be in only one direction (mono-directional carousel) or in both directions (bi-directional carousel).
Reduce the order picking time
One of the main objectives in automated warehouses is to reduce the order picking time. Order picking is the operation with the most impact on a warehouse’s overall performance. It represents approximately 55% of all operating costs (Tompkins et al. 2010). In case of warehouses equipped with a VLM or a Carousel, the picking time depends on the number of trays extracted and the order of extracting them. Since the vertical moving speed of the VLM is much greater than the horizontal speed, it can be assumed that the time required to extract a tray is roughly the same for all trays, which simplifies the problem as the picking time becomes mostly dependent on the number of extracted trays. However, this assumption does not hold for carousels, for which the time required to release a shelf depends on the number of shelves between the initial position and the destination point. In both cases (with VLMs or Carousels), the optimization of picking time is much more complex in case of warehouses with several storage machines since the picker can work in masked time (i.e., for instance, picking products from a given carousel while the others are rotating), which has to be maximized.
This research study aimed to: (1) provide companies with optimization tools to reduce the picking time in automated warehouses equipped with one or several VLMS and one or several Carousels, which represent most situations encountered in practice, and (2) implement the developed tools in the WMS software commercialized by KLS Logistic and make them functional to be used by companies in practice. We demonstrated that the picking time highly depends on the batching strategy employed by the firm (i.e., how to regroup customers’ orders intoLogistics batches to be collected from the VLMs or Carousels) and provided batching methodologies for each type of warehouse to reduce the picking time. This is the first study in the literature to optimize the batching strategy for automated warehouses with the objective of reducing the picking time. We used advanced mathematical optimization techniques and computational methods to develop our models.
Our models were applied and tested in real-world situations for a French hospital and a Swiss company (manufacturer of precision measuring instruments, such as gauges, calibration equipment, calipers, and micrometers). We used real data extracted from the databases of these companies to evaluate the performance of our models compared to the batching strategy currently used by these companies. We showed that our models outperform the current batching strategy in the different settings considered. For instance, in the most representative instance, the average savings in picking time in the case of one VLM were equal to 27% for the French hospital and 22% for the Swiss company. In the case of multiple VLMs, the average savings in completion time were equal to 32% for the French hospital and 26% for the Swiss company. In the case of Carousels, the savings range from 10.49% to 21.50%, for the French hospital, with an average of 17.37%, and from 10.57% to 29.69% for the Swiss company, resulting in an average of 18.60%. Our models have therefore been validated by KLS Logistic and integrated into their WMS software. The PhD student Nicolas Lenoble has been offered a permanent position at KLS Logistic to support the clients of KLS Logistic in adopting the system. [5.1].
Sources to corroborate the impact
[5.1] Nicolas Lenoble, product owner at HARDIS Group (https://www.linkedin.com/in/nicolas-lenoble-49a55847/?originalSubdomain=fr)
[5.2] KLS group (https://www.kls-group.fr/en/kls-group-en-home/)
Period when the impact occurred: 2017-2018
Research Centre: Green, Digital and Demand-Driven supply chain management (G3D research centre)
Dr Ramzi Hammami, Professor in Supply Chain Management, Rennes School of Business