Strong cost-based filtering for lagrange decomposition applied to network design

Abstract

This paper describes a new hybrid algorithm for a multicast network design application. The problem consists of finding a minimum-cost topology and link dimensions subject to capacity and multicast routing constraints. The algorithm exploits Lagrange Decomposition (LD) to yield separable subproblems. Each subproblem has a Constraint Programming relaxation. We derive a sharper cost bound and stronger cost-based filtering rules. The results indicate that our solver outperforms the pure LD approach. © Springer-Verlag Berlin Heidelberg 2004.