Simultaneous solution of Lagrangean dual problems interleaved with preprocessing for the weight constrained shortest path problem

R. Muhandiramge, N. Boland

    Research output: Contribution to journalArticle

    17 Citations (Scopus)

    Abstract

    Conventional Lagrangean preprocessing for the network Weight Constrained Shortest Path Problem (WCSPP), for example Beasley and Christofides (Beasley and Christofides, Networks 19 (1989), 379–394), calculates lower bounds on the cost of using each node and edge in a feasible path using a single optimal Lagrange multiplier for the relaxation of the WCSPP. These lower bounds are used in conjunction with an upper bound to eliminate nodes and edges. However, for each node and edge, a Lagrangean dual problem exists whose solution may differ from the relaxation of the full problem. Thus, using one Lagrange multiplier does not offer the best possible network reduction. Furthermore, eliminating nodes and edges from the network may change the Lagrangean dual solutions in the remaining reduced network, warranting an iterative solution and reduction procedure. We develop a method for solving the related Lagrangean dual problems for each edge simultaneously which is iterated with eliminating nodes and edges. We demonstrate the effectiveness of our method computationally: we test it against several others and show that it both reduces solve time and the number of intractable problems encountered. We use a modified version of Carlyle and Wood's (38th Annual ORSNZ Conference, Hamilton, New Zealand, November, 2003) enumeration algorithm in the gap closing stage. We also make improvements to this algorithm and test them computationally. © 2009 Wiley Periodicals, Inc. NETWORKS, 2009
    Original languageEnglish
    Pages (from-to)N/A
    JournalNetworks: an international journal
    Volume53
    Issue number4
    DOIs
    Publication statusPublished - 2009

    Fingerprint Dive into the research topics of 'Simultaneous solution of Lagrangean dual problems interleaved with preprocessing for the weight constrained shortest path problem'. Together they form a unique fingerprint.

  • Cite this