2950. Online Container Allocation and Dynamic Rebalancing with Deep Reinforcement Learning
Invited abstract in session MC-59: Freight Transportation and Logistics, stream Transportation.
Monday, 12:30-14:00Room: Liberty 1.14
Authors (first author is the speaker)
| 1. | Emre Kara
|
| IE & IS, Eindhoven University of Technology | |
| 2. | Layla Martin
|
| Operations, Planning, Accounting and Control, Eindhoven University of Technology | |
| 3. | Mehrdad Mohammadi
|
| Industrial Engineering and Innovation Sciences, Eindhoven University of Technology | |
| 4. | Willem van Jaarsveld
|
| Eindhoven University of Technology |
Abstract
Containerized transportation plays a vital role in global trade, and a substantial part of its cost comes from empty container movements. As global trade is imbalanced, regional container shortages and surpluses occur and logistics service providers (LSPs) inevitably reposition empty containers to avoid losing sales. Therefore, efficient container fleet management is needed to balance out this waste and service quality. Container fleet management typically focuses on repositionings between stocks. However, efficient container allocations from right stocks to the right orders solve, or at least alleviate the problem proactively. Additionally, LSPs operate diverse fleets to accommodate varying order requirements. Consequently, a container qualifies for various order types, while an order accommodates different types of container. Hence, allocating the right type of container to the right order is another challenge of the problem. Lastly, customer orders are revealed over time shortly before they must be fulfilled, which makes the problem online optimization. In this study, we model the container allocation problem as an MDP and solve it with DRL. We test the PPO algorithm with simulation experiments based on an industrial partner's use case and benchmark its performance against their heuristic policy. Through this research, we aim to demonstrate the efficiency of DRL algorithms in optimizing container allocation decisions in terms of operational costs and service levels.
Keywords
- Transportation
- Logistics
- Programming, Dynamic
Status: accepted
Back to the list of papers