A Shortest Path Graph Attention Network and Non-traditional Multi-deep Layouts in Robotic Mobile Fulfillment System

Abstract

The rapid development of E-commerce has forced warehouse operations to develop towards a robotics-based system named Robotic Mobile Fulfillment System (RMFS), in which shortest path planning and conflict recognition play a vital role in enhancing the operational efficiency under multiple mobile robots movement. Compared to the traditional double-deep layout in Automatic Guided Vehicle (AGV) system, this paper proposes multi-deep based layouts in RMFS, including the modification of Flying-V, Fishbone and Chevron layouts. Under these circumstances, this paper further adopts the shortest path graph attention network in RMFS. This paper considers the Dijkstra algorithm as a baseline and compares it with Biased Cost Pathfinding methods, Anytime Repairing A-star and Flow-Annotation Re-planning methods. The shortest path graph attention network adoption in RMFS should enhance the overall operational efficiency and effectiveness under different layouts scenarios with different path planning methods.