China Mechanical Engineering ›› 2021, Vol. 32 ›› Issue (10): 1213-1221.DOI: 10.3969/j.issn.1004-132X.2021.10.010

Previous Articles     Next Articles

Improved Genetic Programming Algorithm for RCMPSP

CHEN Haojie;DING Guofu;ZHANG Jian;YAN Kaiyin   

  1. Institute of Advanced Design and Manufacturing, School of Mechanical Engineering, Southwest Jiaotong University, Chengdu,610031
  • Online:2021-05-25 Published:2021-06-10



  1. 西南交通大学机械工程学院先进设计与制造技术研究所,成都,610031
  • 通讯作者: 张剑(通信作者),女,1972年生,教授。研究方向为生产调度、智能制造。。
  • 作者简介:陈浩杰,男,1995年生,博士研究生。研究方向为生产调度、智能优化技术。E-mail:。
  • 基金资助:

Abstract: Aiming at the lack of optimization ability for priority rule scheduling, an improved hyper-heuristic genetic programming algorithm for the RCMPSP was proposed to evolve better priority rules. By analyzing the existing priority rules, a normalized attribute set and a top-level discriminant coding method for multi-project scheduling were constructed, and the NSGA-Ⅱ virtual fitness allocation method was applied to evaluate the population for achieving multi-objective optimization. A diversity population updating method was designed to enhance the search ability and avoid the defects that the traditional genetic programming was easy to fall into local optimum. The validity and feasibility of the proposed method were verified by the calculation experiments based on benchmark data set PSPLIB and the aircraft assembly line production instance.

Key words: genetic programming, hyper-heuristic, multi-objective optimization, resource constrained multi-project scheduling problem(RCMPSP), non-dominated sorting genetic algorithm Ⅱ

摘要: 针对优先级规则调度不具备优化能力的缺陷,提出了一种应用于资源受限多项目调度的改进超启发式遗传规划算法以进化出更理想的优先级规则。通过分析现有优先级规则构建出适用多项目调度的归一化属性集和顶层判别编码方式,并结合NSGA-Ⅱ虚拟适应度分配方法对种群进行评估以实现多目标优化。设计了一种多样性种群更新方式,以避免传统遗传规划易陷入局部最优的缺陷和提高搜索能力。通过基于标准数据集PSPLIB所构造的算例和飞机总装装配线的生产实例验证了该方法的有效性和可行性。

关键词: 遗传规划, 超启发式, 多目标优化, 资源受限多项目调度, NSGA-Ⅱ

CLC Number: