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

Chinese Journal of Management Science ›› 2024, Vol. 32 ›› Issue (7): 84-94.doi: 10.16381/j.cnki.issn1003-207x.2021.0481

Previous Articles     Next Articles

Model and Algorithm on Stochastic Scheduling Problem with Activity Overlapping

Zihao Chu1,Zhe Xu2(),Dongning Liu3   

  1. 1.Aviation Industry Development Research Center of China, Beijing 100029, China
    2.School of Economics and Management, Beihang University, Beijing 100191, China
    3.School of Economics and Management, University of Chinese Academy of Sciences, Beijing 100190, China
  • Received:2021-03-10 Revised:2022-03-16 Online:2024-07-25 Published:2024-08-07
  • Contact: Zhe Xu E-mail:xuzhebuaa@163.com

Abstract:

Overlapping can not only effectively reduce the duration of R&D projects, but also decline the risk of R&D failure owe to finding the existing problems in time through the early exchange of information between upstream and downstream activities. In the project scheduling problem with activity overlapping, the uncertainty of activity duration will lead to uncertain overlapping time, uncertain overlapping amount, and uncertain rework time of downstream activities. The stochastic project scheduling is studied with activity overlapping which is NP-hard. Firstly, a multistage decision process model is established to describe the stochastic scheduling process with activities executing in overlapping way and uncertain activity durations. Then, a two-stage GA-rollout algorithm combining open-loop and close-loop policy is designed. A unique chromosome coding and decoding approach is embedded in the GA algorithm to select effective overlapping activities. And the initial solution obtained by GA is used as base policy, and then further optimized by rollout strategy. Finally, through the large-scale experimental study and the comparative analysis of the solution quality of different algorithms, the good performance of our GA-Rollout algorithm is verified.

Key words: stochastic scheduling, overlapping, multistage decision process, approximate dynamic programming, genetic algorithm

CLC Number: