Graph Neural Networks and Deep Reinforcement Learning for Adaptive Warehouse Optimization

Warehouse operations play a critical role in modern supply chains, with order picking representing one of the most resource-intensive processes in both manual and automated warehouses. Improving the efficiency of picking operations has therefore been a long-standing research focus in logistics and operations research. Traditionally, order-picking routing problems have been modeled as variants of the Traveling Salesman Problem (TSP) and addressed using heuristics and metaheuristics. Recent advances in machine learning, particularly deep reinforcement learning (DRL), offer promising opportunities for developing adaptive decision-making systems capable of operating in dynamic environments. At the same time, warehouse systems naturally exhibit graph structures that can be used by graph neural networks (GNNs). This doctoral research proposes an integrated framework combining GNNs and DRL to optimize warehouse processes.