基于關聯網絡的機器人移動貨架系統貨位分配方法

胡祥培 丁天蓉 張源凱 莊燕玲

摘 要：作為一種新型智能訂單揀選系統，機器人移動貨架系統為解決大型網上超市面臨的一單多品訂單揀選難題提供了前所未有的機遇，而貨位分配是關系該系統運行效率最為核心的關鍵環節之一。針對該系統一品多位貨位分配問題帶來的商品關聯關系復雜、解空間巨大等難題，本文以提高貨架商品間關聯性為目標，基于縮減問題解空間的思想，提出“關聯網絡構建→關聯網絡分析→關聯網絡聚類”的三階段貨位分配方法，將聚類的商品網絡映射成貨架對應的貨位擺放來快速高效生成貨位分配方案。利用某網上超市的實際數據，將所提方法與國內外普遍應用的幾種方法進行對比，結果驗證了所提方法的高效性和實用性，可為網上超市等企業采用機器人移動貨架系統解決訂單履行難題提供理論指導和決策支持。

關鍵詞：機器人移動貨架系統;貨位分配;關聯網絡;聚類;“貨到人”揀選模式

中圖分類號：C934 文獻標識碼：A 文章編號：2097-0145（2022）01-0056-09 doi：10.11847/fj.41.1.56

Abstract：As a new intelligent order picking system， the robotic mobile fulfillment system plays an important role in solving the multi-item orders picking problem faced by online supermarkets. The efficiency of this system hinges on the storage assignment procedure. However， due to the complexity of product correlation and huge solution space， this storage assignment procedure is difficult to optimize. In order to reduce the solution space for the problem of storage assignment in robotic mobile fulfillment system， a three-stage method of “association network construction→association network analysis→association network clustering” is proposed to improve the relevance of products on the rack. This method lays the foundation for solving the storage assignment problem quickly and efficiently by mapping clusters to racks. Using the order data of an online supermarket， this paper compares the proposed method with several widely used methods. Results verify the effectiveness and practicability of the proposed method. This research provides theoretical guidance and decision support for companies who employ this system to solve order fulfillment problems.

Key words：robotic mobile fulfillment system; storage assignment; association network; clustering; parts-to-picker picking mode

1 引言

隨著互聯網的快速發展，網上購物已經成為一種新型購物方式。中國擁有全球最大的網絡零售市場，其中京東商城2020年成交總額達26125億元，訂單量仍在逐年遞增，同時每一個訂單平均包含7～8種商品[1]，傳統“人到貨”揀選模式面對訂單量的爆炸式增長以及高比例的一單多品訂單，導致倉庫運營成本激增，僅訂單揀選這一環節成本支出占倉庫運營成本65%以上[2]。揀貨員需執行高強度、遠距離的重復性揀選操作，揀選效率低下。……