Using genetic programming and simulation to learn how to dynamically adapt the number of cards in reactive pull systems
作者：
Highlights：
• We show how to learn dynamically to adapt the number of cards in real time in token-based pull systems.
• We propose a Simulation-based Genetic Programming approach which does not need training sets.
• We illustrate how the approach can be implemented using Arena and μGP.
• A reactive ConWIP example show the efficiency of the approach and of the knowledge extracted.
• The resulting decision tree can be used online by production managers or for self-adaptation issues.
摘要
•We show how to learn dynamically to adapt the number of cards in real time in token-based pull systems.•We propose a Simulation-based Genetic Programming approach which does not need training sets.•We illustrate how the approach can be implemented using Arena and μGP.•A reactive ConWIP example show the efficiency of the approach and of the knowledge extracted.•The resulting decision tree can be used online by production managers or for self-adaptation issues.
论文关键词：Kanban,ConWIP,Manufacturing systems,Reactive pull systems,Self-adaptive systems,Learning,Simulation,Genetic programming
论文评审过程：Available online 29 November 2014.
论文官网地址：https://doi.org/10.1016/j.eswa.2014.11.052
