Decision system for smart management of resources in warehouses – AGIRE

In this 4-year project, we aim at developing innovative intelligent human-centered decision support tools for warehouse management. The proposed methodology relies on an activity analysis to identify and to model the main human factors in warehouse operations. Then, different optimization problems will be addressed at operational and tactical levels (e.g. picker routing, scheduling of picking tours, storage location assignment, workload balancing). These decision levels will be integrated subsequently to ensure that decisions at the tactical level are consistent with the human factors at the operational level. In order to evaluate the obtained solutions, simulation tools will be developed, as well as experiments with feedbacks of users inside the warehouse of Jules.

A review of the state of the art dealing with the consideration of human aspects in supply chain optimization was conducted. In addition, interviews were organized with employees at Jules to identify warehouse processes and organization, IT tools used by staff, and internal human resources management. A number of workstations have been identified as requiring heavy lifting, repeated awkward postures, frequent shifting and trunk rotation. In addition to exploratory interviews, visits were made to the warehouse at Jules to conduct semi-structured interviews, visits to different areas of the warehouse and open observations. In addition, research was carried out on optimization problems in warehouses at the operational and tactical levels. In a first work, we studied the consideration of congestion in the aisles of the order-picking area: congestion occurs when several pickers are in the same aisle (or part of an aisle) during a given time interval. The problem studied is the integrated problem of order batching and picker routing. A mathematical model integrating congestion has been developed, as well as a solving method to propose optimal solutions. In a second work, we studied the modeling of discomfort in order-picking routes. The proposed model is based on the following characteristic elements: picking operations have different levels of difficulty (depending on height, type of product, container opening); when moving between two picking points, it is possible to have a recovery phenomenon when the walking distance is acceptable. This model is incorporated into an algorithm that proposes a set of routes with different distances and different levels of discomfort. The results show that it is possible to propose much less uncomfortable tours by slightly increasing the distance to be travelled. In a third work, we investigated the modeling of a product assignment problem in a logistics warehouse. The aim is to assign products to locations with the objective of minimizing the total distance traveled by operators. Several models have been developed to represent different warehouse routing policies. To solve these optimization problems, a Branch-and-Price algorithm was designed. This algorithm gave excellent results on literature instances. Finally, in a last work we studied efficient evaluations of solution modifications in order batching and picker routing problems.

The work carried out as part of the ANR AGIRE project opens up a number of perspectives. On the one hand, from an ergonomic point of view, it seems interesting to study in more detail how order pickers perceive the discomfort of the operations they have to carry out. On the other hand, from an operational research point of view, the initial work conducted during the project leaves plenty of room for improvement in terms of the phenomena taken into account in the models, the robustness of the solutions, and the consideration of aspects linked to fairness between order pickers.

Beyond these mono-disciplinary perspectives, the AGIRE project has highlighted that a central issue for the well-being of operators is the possibility of autonomy in their decision-making. However, in most logistics warehouse management systems, decisions are delegated to an algorithm, and order pickers are obliged to follow the route decided by the algorithm. This obligation to follow decisions is often imposed via digital tools (headset with microphone, digital terminal). Allowing order-pickers to have greater visibility over future operations, or even to deviate from the schedule provided by the algorithm, would bring a significant gain in the autonomy of these operators, and thus in their working conditions. It therefore seems very interesting to consider how to propose tools and algorithms that can offer good solutions to operators, while allowing them to be autonomous by deviating from the solutions, but with a calculation adjusted accordingly to the new solutions.

Prunet, T.; Absi, N.; Borodin, V.; Cattaruzza, D. Storage Location Assignment Problem in Fast Pick Areas: A novel formulation and decomposition method. CPAIOR’2021. Vienna, Austria. July 2021.

Prunet, T., Absi, N., Borodin, V., Cattaruzza, D. Efficient Move Evaluation and Neighborhood Exploration for Integrated Order Picking Planning Problems in Picker-to-Parts Warehouses. Odysseus 2024. Carmona, Spain. May 2024.

Prunet, T.; Absi, N.; Borodin, V.; Cattaruzza, D. Optimization of Human-Aware Manufacturing and Logistics Systems: A Comprehensive Review of Modeling Approaches and Applications. EJTL. 2024.

Torrealba, P.; Feillet, D.; Ogier, M.; Semet, F. A column generation approach to solve the Joint Order Batching and Picker Routing Problem with picker congestion. Odysseus 2024. Carmona, Spain. May 2024.

The project AGIRE addresses the human resources management in warehouses which supply either sale points (B2B) or final consumers (B2C). Such warehouses are under pressure due to the no inventory policy at the sale points and to the constant growth of e-commerce sales. In terms of logistics, this translates into an increasing number of parcels to prepare and to ship to satisfy an order, which is known typically a few hours in advance.

Therefore, designing an efficient, reactive warehouse able to respond to demand variations is a significant challenge to address. A fully automated warehouse cannot be considered as a solution. Indeed, the investments required to satisfy appropriately demand peaks are too large to be considered. A more suitable approach is to combine human operators equipped with high technology devices and partially automated logistic equipments. However, to better manage human resources in such a dynamic environment, innovative decision support methods have to be designed with the objective of maximizing the welfare of workers while guaranteeing the efficiency of the whole logistics system. This is precisely the central objective of this research project.

AGIRE aims at proposing innovative human-centered decision support tools for warehouse management. This will have a tangible impact on the scientific community. Warehouse management problems have been widely studied for decades, but only few works integrate human factors when facing a dynamic demand. This project will explicitly consider these two aspects in the proposed models and algorithms. The main scientific challenges are: (1) the modeling of human and social aspects (e.g. ergonomics, pain/fatigue, congestion); (2) the integration of these human related criteria into optimization models and algorithms, (3) data analysis (e.g. demand forecast) and integration of the uncertainties into the models and solution methods, (4) integration of tactical and operational problems, and (5) solving large-sized instances: a warehouse usually contains hundreds of employees, thousands of storage locations, up to one hundred thousand products to pick each day.

AGIRE is conducted by a consortium of three complementary research teams (INOCS, SFL and SPLOTT) in computer sciences, operations research, and social sciences; and a company in the textile retail: HappyChic. Developed algorithms will be tested on real data collected in the warehouse of the brand Jules of HappyChic.

In this 4-year project, we aim at developing innovative intelligent human-centered decision support tools for warehouse management. The proposed methodology relies on an activity analysis to identify and to model the main human factors in warehouse operations. Then, different optimization problems will be addressed at operational and tactical levels (e.g. picker routing, scheduling of picking tours, storage location assignment, workload balancing). These decision levels will be integrated subsequently to ensure that decisions at the tactical level are consistent with the human factors at the operational level. In order to evaluate the obtained solutions, simulation tools will be developed, as well as experiments with feedbacks of users inside the warehouse of Jules.