EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1723. Learning a Policy for the Real-Time Inventory Rack Storage Assignment and Replenishment Problem
Invited abstract in session MB-4: Recent Methodologies in Explainable AI (XAI) 2, stream Recent Advancements in AI .
Monday, 10:30-12:00Room: 1001 (building: 202)
Authors (first author is the speaker)
1. | Sander Teck
|
Centre for Industrial Management / Traffic & Infrastructure, KU Leuven | |
2. | San Tu Pham
|
Computer Science, KU Leuven | |
3. | Louis-Martin Rousseau
|
Mathematic and Industrial Engineering, École Polytechnique de Montréal | |
4. | Pieter Vansteenwegen
|
Institute for Mobility - CIB, KU Leuven |
Abstract
The e-commerce sector is rapidly automating warehouse operations, necessitating efficient control systems to manage this growing complexity. This study investigates Robotic Mobile Fulfillment Systems (RMFS), employing autonomous mobile robots (AMRs) for inventory rack management. These systems eliminate the need for human operators within the storage area, with AMRs responsible for both storing and retrieving movable inventory racks. Human operators stationed at workstations alongside the storage area fulfill customer orders by picking goods from these racks. RMFS dynamically adjusts rack positions based on usage frequency. Furthermore, effective restocking timing is crucial to prevent operation delays. Real-time scheduling, especially in inventory rack storage and replenishment, is essential for optimal performance.
To minimize cycle time, a deep reinforcement learning (DRL) approach is proposed. The storage area is divided into zones to facilitate efficient decision-making. The learning agent interacts with the environment, observes changes in its state, and learns through trial-and-error from these interactions. Thereby, it is able to construct a policy by mapping environment states to actions. Experimental results demonstrate significant cycle time improvements compared to traditional decision rules. The study underscores the importance of real-time decision-making for the storage assignment and replenishment problem.
Keywords
- E-Commerce
- Warehouse Design, Planning, and Control
- Machine Learning
Status: accepted
Back to the list of papers