• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2024, Vol. 60 ›› Issue (2): 323-341.doi: 10.3901/JME.2024.02.323

• 交叉与前沿 • 上一篇    下一篇

扫码分享

增强三维分布估计算法求解分布式生产-运输-装配集成调度问题

张博茹1, 胡蓉1,2, 钱斌1,2, 金怀平1, 向凤红1   

  1. 1. 昆明理工大学信息工程与自动化学院 昆明 650500;
    2. 昆明理工大学云南省人工智能重点实验室 昆明 650500
  • 收稿日期:2023-01-03 修回日期:2023-08-18 出版日期:2024-01-20 发布日期:2024-04-09
  • 通讯作者: 胡蓉(通信作者),女,1974年出生,副教授,硕士研究生导师。主要研究方向为智能优化调度、物流优化。E-mail:ronghu@vip.163.com
  • 作者简介:张博茹,女,1998年出生。主要研究方向为智能算法与优化调度。E-mail:2233289476@qq.com
    钱斌,男,1976年出生,教授,博士研究生导师。主要研究方向为优化调度理论与方法、智能优化方法。E-mail:bin.qian@vip.163.com
    金怀平,男,1987年出生,副教授,硕士研究生导师。主要研究方向为机器学习与数据挖掘。E-mail:jinhuaiping@gmail.com
    向凤红,男,1964年出生,教授,硕士研究生导师。主要研究方向为智能优化方法。E-mail:xiangfh5447@sina.com
  • 基金资助:
    国家自然科学基金(62173169,61963022)和云南省应用基础研究重点(202201AS070030)资助项目。

Enhanced Three-dimensional Estimation of Distribution Algorithm for Solving Integrated Scheduling Problem of Distributed Production, Transportation and Assembly

ZHANG Boru1, HU Rong1,2, QIAN Bin1,2, JIN Huaiping1, XIANG Fenghong1   

  1. 1. Faculty of Information Engineering and Automation, Kunming University of Science and Technology,Kunming 650500;
    2. Yunnan Key Laboratory of Artificial Intelligence, Kunming University of Science and Technology,Kunming 650500
  • Received:2023-01-03 Revised:2023-08-18 Online:2024-01-20 Published:2024-04-09

摘要: 针对一类广泛存在的分布式生产-运输-装配集成调度问题(Integrated scheduling problem of distributed production, transportation and assembly, ISPDPTA),建立问题模型并提出一种增强三维分布估计算法(Enhanced three-dimensional estimation of distribution algorithm, E3DEDA)进行求解。在E3DEDA的全局搜索部分,先根据ISPDPTA的问题特性,对包含工厂信息的工件与产品进行多段编码,并依据最短路径规则获得各车辆路径;再采用双三维概率模型分别学习和积累种群中较优个体的工件块与产品块结构及其位置信息,并采样生成新个体,从而增强算法发现优质解空间区域的能力。在E3DEDA的局部搜索部分,设计自适应变邻域局部搜索来增强算法的局部搜索能力。具体而言,针对问题各阶段特性,设计10种有效的邻域操作组成备选集合,并采用二维概率模型学习由不同邻域操作所构成的优质邻域结构信息,进而采样生成合理的邻域操作排列并依次执行,以实现对全局搜索所发现优质区域的深入搜索。此外,设计块结构概率评价更新机制,可提升算法执行效率。最后,通过仿真试验与算法对比验证E3DEDA可有效求解ISPDPTA。

关键词: 分布式生产, 车辆运输, 集成调度, 三维分布估计算法, 块结构

Abstract: Aiming at a kind of integrated scheduling problem of distributed production, transportation and assembly(ISPDPTA), which widely exists in real-life applications, the model of the problem is established first. And then, an enhanced three-dimensional estimation of distribution algorithm(E3DEDA) is proposed to solve the model. In global search stage, E3DEDA uses a multi-segment encoding mechanism to represent jobs and productions that contain factory information based on the characteristics of ISPDPTA, in which each vehicle path is determined by applying the shortest path rule. Besides, two three-dimensional probability matrices are adopted to enhance the ability of E3DEDA for finding promising search regions. That is, the matrices separately learn and accumulate the information about the structure and location of job blocks and product blocks stemming from elite individuals found during the search process. Thereby, new individuals are generated by sampling the two matrices. In local search stage, E3DEDA adopts adaptive variable neighborhood local search to improve the local search ability of the algorithm. Especially, ten neighborhood operations are used to form the alternative set based on the characteristics of each stage of ISPDPTA, and a two-dimensional probability matrix is used to learn the information about high-quality neighborhood structures regarding different neighborhood operations. By sampling the matrix, permutations are generated and a series of neighborhood operations are in turn performed on the solutions of E3DEDA, so as to intensively examine promising search regions discovered by the global search stage. Moreover, a probability evaluation updating mechanism of block structure is proposed to improve the efficiency of the algorithm. Finally, simulations and comparisons demonstrate that E3DEDA can effectively solve ISPDPTA.

Key words: distributed production, vehicle transportation, integrated scheduling, three-dimensional estimation of distribution algorithm, block structure

中图分类号: