• DocumentCode
    478576
  • Title

    Improving the Performance of Interactive Configuration with Regular String Constraints

  • Author

    Hansen, Esben Rune ; Tiedemann, Peter

  • Author_Institution
    IT Univ. of Copenhagen, Copenhagen
  • Volume
    1
  • fYear
    2008
  • fDate
    3-5 Nov. 2008
  • Firstpage
    3
  • Lastpage
    10
  • Abstract
    A generalization of the problem of interactive configuration has previously been presented in [1]. This generalization utilized decomposition to extend the standard finite domain interactive configuration framework to deal with unbounded string variables and provided features such as prefix auto-completion. In this paper we present several significant improvements to the core data structures and algorithms in [1] as well as the first implementation of an interactive configurator on string variables. The primary improvement is obtained by replacing the binary decomposition model with a finite domain model. We then describe an optimization for this model which allows us to replace the use of costly conjunctions with simple restrict operations during synchronization between decomposed constraints. In addition we describe how to improve the performance of the auto-complete operation, by using projection on relevant variables to significantly reduce the size of the data structures involved. We empirically verify the critical significance of these improvements using our own implementation of a string variable based configurator on real-world example data.
  • Keywords
    data structures; decision support systems; optimisation; binary decomposition model; core data structures; finite domain model; interactive configuration; regular string constraints; Artificial intelligence; Binary decision diagrams; Boolean functions; Constraint optimization; Constraint theory; Data structures; NP-hard problem; Polynomials; Constraint Programming; Decision Support; Regular Languages;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Tools with Artificial Intelligence, 2008. ICTAI '08. 20th IEEE International Conference on
  • Conference_Location
    Dayton, OH
  • ISSN
    1082-3409
  • Print_ISBN
    978-0-7695-3440-4
  • Type

    conf

  • DOI
    10.1109/ICTAI.2008.131
  • Filename
    4669665