Simulation and Implementation of an LLM-Assisted Pickup Robot in Unmanned Warehouse

With the rapid growth of e-commerce, the demand for small-scale warehouses has increased significantly to enhance distribution efficiency. However, conventional Automated Guided Vehicles (AGVs) often rely on task-specific algorithms at different operational stages, limiting their generality and hindering widespread adoption. This study proposes a novel approach that employs Large Language Models (LLMs) as a general reasoning tool, which is accessed via the cloud. By offloading high-level reasoning to the cloud, AGVs can utilize LLMs for various task stages while reducing hardware requirements at the edge and improving overall applicability. We designed an AGV to perform item pickup tasks in unmanned warehouse environments. Results from both simulations and real-world experiments demonstrate the effectiveness and practicality of the proposed method.