主管：中国科学院
主办：中国优选法统筹法与经济数学研究会
　　　中国科学院科技战略咨询研究院
中国管理科学

中国管理科学 ›› 2022, Vol. 30 ›› Issue (4): 240-251.doi: 10.16381/j.cnki.issn1003-207x.2019.1408

• 论文 • 上一篇    下一篇

“货到人”拣选系统中AGV路径规划与调度研究

李昆鹏1, 刘腾博1, 贺冰倩2, 徐东洋3   

  1. 1.华中科技大学管理学院，湖北 武汉430074； 2.长江存储科技有限责任公司，湖北 武汉430205；3.河南大学商学院，河南 开封475004
  • 收稿日期:2019-09-19 修回日期:2019-12-05 出版日期:2022-04-20 发布日期:2022-04-26
  • 通讯作者: 李昆鹏(1978-)，男(汉族)，湖北枝江人，华中科技大学管理学院,教授，博士，研究方向：物流与供应链管理、生产运作管理，Email：likp@mail.hust.edu.cn. E-mail:likp@mail.hust.edu.cn
  • 基金资助:
    国家自然科学基金资助项目(71902054)

A Study on Routing and Scheduling of Automated Guided Vehicle in “Cargo-to-Picker” System

LI Kun-peng1, LIU Teng-bo1, HE Bing-qian2, XU Dong-yang3   

  1. 1. School of Management，Huazhong University of Science & Technology，Wuhan 430074, China；2. Yangtze Memory Technologies Co., Ltd.，Wuhan 430205, China；3. School of Business，Henan University，Kaifeng 475004, China
  • Received:2019-09-19 Revised:2019-12-05 Online:2022-04-20 Published:2022-04-26
  • Contact: 李昆鹏 E-mail:likp@mail.hust.edu.cn

PDF (PC)

300

摘要： 本文在电商行业快速发展及人工智能不断成熟的背景下，对智能仓库中采用“货到人”拣选系统的AGV(Automated Guided Vehicle)调度优化问题进行研究，此问题需同时考虑单AGV的路径规划及多AGV间的碰撞避免。基于此，本文提出了两阶段优化算法，首先分别采用数学模型求解和基于货架优先级的任务分配算法得到AGV的货架搬运任务序列，并根据AGV行驶规则生成初始路径。然后设计碰撞检测及避免算法对可能发生冲突的路径交叉点进行主动避撞调度。当发生突发情况时(如设备故障)，采取实时重调度措施进行被动路径调整以获得全局AGV无碰路径。最后基于两阶段算法分析了两种AGV搬运任务分配策略的适用情况，并给出了最优的AGV数量配置建议。实验结果表明，本文算法能够调度100台AGV在配备有2015个货架、10个作业平台的仓库中完成100个订单的拣选作业。本研究可为企业采用“货到人”拣选系统实现多AGV 的协同调度提供理论依据和实践指导。

关键词: 智能仓库；AGV调度；动态避撞；“货到人”拣选

Abstract: With the rapid development of E-commerce and artificial intelligence technology, the scheduling optimization problem is studied for AGV(Automated Guided Vehicle)in the intelligent warehouse which uses“cargo-to-picker” picking system. This problem simultaneously considers two sub-problems: the AGV route generating and routes collision avoidance between multiple AGVs. To solve the problem, a two-phase optimization approach is present. First, two methods are developed, i.e., a mathematical model and a rack-priority-based method to allocate the rack-moving tasks to AGVs. Based on the task allocation results, the path generation algorithm is developed to get the AGV initial feasible path. Then, a collision avoidance algorithm is designed to obtain collision-free AGV routes considering possible collision path points. It is further considered unexpected disruptions occur, such as equipment breakdown. Therefore, a real-time rescheduling method is developed to obtain AGV collision-free path in the warehouse. The performance of the two-stage optimization algorithm exploring the two rack allocation methods are evaluated by computational experiments. The results show that this algorithm can help to obtain an optimized AGV schedule in a warehouse with 2015 shelves, 10 operating platforms, 100 AGVs and 100 orders in a batch. The algorithm also can give suggestions of optimal configuration of AGV quantity to give a sized warehouse. The study not only can provide a theoretical model and algorithm for scheduling operation of multiple AGVs, but also can provide practical guidance for enterprises exploring the “cargo-to-picker” picking system.

Key words: intelligent warehouse；AGV schedule；dynamic collision avoidance；“cargo-to-picker” picking

中图分类号: