The article presents a novel approach to collaborative vehicle routing (CVR), a process where carriers share transportation requests to achieve economies of scale. This method reduces costs, greenhouse gas emissions, and road congestion. However, it presents challenges in deciding which carriers should collaborate and how to fairly compensate each carrier. The researchers propose a solution using deep multi-agent reinforcement learning, which significantly reduces the computational resources required in traditional methods. This approach also addresses both route allocation and gain sharing problems simultaneously, leading to outcomes that correlate well with the Shapley value, deemed a fair profit allocation mechanism. The proposed model managed to reduce run-time by 88%.
Publication date: 27 Oct 2023
Project Page: ?
Paper: https://arxiv.org/pdf/2310.17485
