Deep Reinforcement Learning for Dynamic Capacitated Vehicle Routing Problem

Kaiqiang Tang, Huiqiao Fu, Jiasheng Liu, Guizhou Deng, Yuanyang Lu, Yiming Qi, Chunlin Chen

Primary Area: reinforcement learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: attention model, deep reinforcement learning, dynamic capacitated vehicle routing, LKH, OR-tools

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

TL;DR: This paper propose a DRL model based on a designed dynamic attention mechanism network for dynamic CVRP, which extends the attention model from the original static-only CVRP environment to a dynamic CVRP environment.

Abstract: Deep Reinforcement Learning (DRL) has become increasingly popular for solving Capacitated Vehicle Routing Problem (CVRP) due to its great potential. However, the current DRL models are only suitable for static environments where information about customers and orders is provided before the delivery vehicle departs from the depot and does not change during delivery. In reality, delivery tasks are dynamic, and much information about customers and orders is disclosed over time. In this paper, we propose a DRL model based on a designed dynamic attention network for dynamic CVRP, which extends the attention model from the original static-only CVRP environment to a dynamic CVRP environment. With dynamic encoder-decoder architecture, the proposed DRL model can track the changes in customer disclosure status in real-time. For comparison, we develop two methods based on LKH and OR-Tools for dynamic CVRP. Experimental results show that the DRL model outperforms LKH and OR-Tools in computational speed and solution quality. The code is publicly available on https://anonymous.4open.science/r/AM2DCVRP-0D4B.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 1739