A Lifelong Conflict-Aware AGV Routing System

Automated Guided Vehicle (AGV) play an increasingly critical role in modern warehouses to cope with the enormous logistic demands that come with the fast developing e-commerce and the growing operational costs. The key component of implementing such a system is planning the paths of a large horde of AGVs to deliver orders from shelves to packing locations. The existing solutions regard it as a Multi Agent PathFinding (MAPF) problem, but they can hardly be applied in practice because none of them could satisfy lifelong (future task unknown and keeps appearing), kinematic (acceleration/deceleration/rotation) and online (fast response to issues) at the same time. Therefore, in this work, we first propose the Conflict-Aware AGV Routing (CAAR) algorithm that can satisfy all these properties. After that, to further improve the system throughput, we propose Flexible Parking Location and Parking Location Buffer to support efficient lifelong task scheduling and put forward several drop-off/pick-up distribution strategies for task organization. Extensive experimental studies verify the superiority of our methods compared with the state-of-the-art.