The article discusses a new method for Collaborative Vehicle Routing (CVR), where delivery companies share information and perform deliveries on behalf of each other to reduce costs and emissions. The authors propose a solution to the problem of deciding which companies should partner and how they should be compensated, which traditionally required complex game theory calculations. Instead, they use a reinforcement learning approach, which is more scalable and accounts for the self-interest of companies. This approach was found to correctly identify optimal partnerships 79% of the time, with a 4.2% optimality gap and a 62% reduction in runtime.
Publication date: 26 Oct 2023
Project Page: https://arxiv.org/abs/2310.17458v1
Paper: https://arxiv.org/pdf/2310.17458
