.png){kind=icon}
.png){kind=icon}
Return
Optimizing container relocation operations by using deep reinforcement learning
.png){kind=icon}
PRE
AI
DOI：10.1080/03088839.2024.2424865.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
With the growing number of containers, it is challenging to improve the efficiency of container retrieval operation due to the container relocation operations. The container relocation problem (CRP) aims to retrieve all containers from a bay with the smallest number of relocations. Most of existing studies attempted to solve the CRP by using exact algorithms, and heuristic algorithms. However, those algorithms could require long and fluctuant computational time for large-scale problems. We proposed a deep reinforcement learning algorithm (DRL) with a beam search method for the CRP (CRP-DRL algorithm). In this algorithm, a deep neural network is trained by the DRL to learn the container relocation strategies. The beam search is integrated into the algorithm to improve the solution quality. Experimental results showed that the solutions obtained from our algorithm are optimal or near-optimal for small-scale problems. For large-scale problems, our algorithm can obtain solutions of high quality with a significant reduction of computational time compared with existing algorithms. This algorithm also provides port operators with the flexibility in the balance between the solution quality and computational time according to their decision-making preference. Hence, it could be efficient in the real-time operation and helpful for the smart port operation systems.
Keywords:
Smart port
deep reinforcement learning
container relocation
beam search
optimization
Journal
IF:
3.6
Papers:
1.0K
Citations:
3.1K