Ordinal optimization of G/G/1/K polling systems with k-limited service discipline

S. C. Horng, S. Y. Lin*

*Corresponding author for this work

Research output: Contribution to journalJournal Article peer-review

25 Scopus citations


In this paper, we propose an ordinal optimization theory based algorithm to solve the optimization problem of G/G/1/K polling system with k-limited service discipline for a good enough solution using limited computation time. We assume that the arrival rates do not deteriorate visibly within a very short period. Our approach consists of two stages. In the first stage, we employ a typical genetic algorithm to select N=1024 roughly good solutions from the huge discrete solution space Ω using an offline trained artificial neural network as a surrogate model for fitness evaluation. The second stage consists of several substages to select estimated good enough solutions from the previous N, and the solution obtained in the last substage is the good enough solution that we seek. Using numerous tests, we demonstrate: (i) the computational efficiency of our algorithm in the aspect that we can apply our algorithm in real-time based on the arrival rate assumption; (ii) the superiority of the good enough solution, which achieves drastic objective value reduction in comparison with other existing service disciplines. We provide a performance analysis for our algorithm based on the derived models. The results show that the good enough solution that we obtained is among the best 3.31×10-6% in the solution space with probability 0.99.

Original languageEnglish
Pages (from-to)213-231
Number of pages19
JournalJournal of Optimization Theory and Applications
Issue number2
StatePublished - 02 2009
Externally publishedYes


  • K-Limited service disciplines
  • Ordinal optimization
  • Performance analysis
  • Polling systems
  • Stochastic simulation optimization


Dive into the research topics of 'Ordinal optimization of G/G/1/K polling systems with k-limited service discipline'. Together they form a unique fingerprint.

Cite this