Dynamic Scheduling Based on Two-Layer Deep Reinforcement Learning for Multi-load AGVs

In the rapidly evolving landscape of manufacturing, the efficient scheduling of Automated Guided Vehicles (AGVs) is crucial for optimizing production processes and reducing operational costs. This article delves into the challenges posed by dynamic production tasks and the limitations of existing scheduling methods, particularly for multi-load AGVs. It introduces a groundbreaking two-stage dynamic scheduling framework that leverages two-layer deep reinforcement learning to tackle the complexities of state and action spaces. The framework decomposes the scheduling problem into task assignment and path planning, utilizing reinforcement learning for task assignment and integer programming for path optimization. This approach not only enhances the adaptability of AGV systems to varying task scenarios but also ensures real-time performance and computational efficiency. The article provides a comprehensive mathematical model, detailed implementation steps, and simulation results that demonstrate the superior performance of the proposed method compared to static dispatching rules and single-layer reinforcement learning. By addressing the intricacies of multi-load AGV scheduling, this research offers valuable insights for improving the efficiency and flexibility of modern manufacturing systems.