GAGE-Q: Reinforced Genetic Algorithm using Spatial Neighborhood Graph Embedding for Green Intermodal Transportation

Abstract. Intermodal transportation, using multiple modes in a single journey, promotes sustainable logistics. This paper addresses the Intermodal Vehicle Routing Problem and proposes GAGE-Q, which integrates graph embedding and Reinforcement Learning into a Genetic Algorithm. Our method models cities and their multi-modal connections as a graph, leveraging embedding for spatial dependencies and RL-based crossover for faster convergence and better solutions. Experiments on synthetic and real-world data show that GAGE-Q outperforms state-of-the-art methods, offering improved solution quality and efficiency in intermodal route planning.