Graph Convolutional Networks for logistics optimization: A survey of scheduling and operational applications

Abstract

Graph Convolutional Networks (GCNs) have emerged as pivotal tools in addressing intricate optimization and scheduling challenges within logistics, encompassing canonical problems such as the Vehicle Routing Problem (VRP), Traveling Salesman Problem (TSP), and dynamic job scheduling. This survey presents a comprehensive exploration of GCN applications, emphasizing their capacity to model spatial–temporal dependencies and their seamless integration with advanced paradigms, including reinforcement learning and hybrid optimization techniques. By leveraging these capabilities, GCNs have demonstrated enhanced scalability and interpretability, rendering them indispensable for large-scale, real-time logistics systems. The review extends to real-world implementations, illustrating GCN-driven innovations in resource allocation, traffic management, and supply chain optimization. In addition, the study critically examines persistent challenges—ranging from processing dynamic graphs to ensuring ethical deployment through fairness and sustainability. The paper concludes with forward-looking recommendations, advocating for the evolution of GCN architectures to adeptly manage real-time decision-making and uncertainty in increasingly complex logistical landscapes.