AIMORL: An Adaptive Integrated Multi-Objective Reinforcement Learning Algorithm for Multi-Modal Logistics Route Scheduling

With the rapid development of intelligent logistics and the increasing complexity of multi-modal transportation systems, achieving efficient and adaptive route scheduling has become a critical challenge. Traditional optimization methods often struggle to balance multiple objectives such as transportation cost, travel time, and environmental impact under dynamic and uncertain conditions. To address this issue, this paper proposes an Adaptive Integrated Multi-Objective Reinforcement Learning (AIMORL) algorithm for optimizing multi-modal logistics route scheduling. The AIMORL framework integrates graph neural networks (GNNs) to model the topological dependencies among logistics nodes and transportation modes, while a multi-objective reinforcement learning (MORL) agent dynamically learns scheduling strategies that minimize overall cost, time, and carbon emissions. Furthermore, a prediction compensation mechanism (PCM) is incorporated to adaptively correct decisions under real-time disturbances such as delays or network congestion. The PCM plays a critical role in adjusting scheduling decisions by compensating for the mismatch between predicted and actual states, effectively mitigating the impact of disruptions such as unexpected delays, traffic congestion, and real-time fluctuations in network conditions. This mechanism ensures the robustness and stability of the logistics network by continuously updating the policy based on real-time feedback. Extensive experiments conducted on synthetic and real-world multimodal logistics datasets demonstrate that AIMORL outperforms existing heuristic and learning-based methods in terms of convergence speed, route efficiency, and robustness under dynamic conditions. The proposed framework provides an intelligent, sustainable, and scalable solution for future logistics scheduling in smart transportation networks.