INTEGER PROGRAMMING APPROACHES FOR MINIMUM STABBING PROBLEMS

Autor(es)

Piva, Breno ^{[1, 2]} ; de Souza, Cid C. ^[1] ; Frota, Yuri ^[3] ; Simonetti, Luidi ^[3]

Afiliação do(s) autor(es)

^[1] Univ Estadual Campinas, Inst Comp, BR-13083852 Campinas, SP - Brazil

^[2] Univ Fed Sergipe, Dept Comp, BR-49100000 Sao Cristovao, SE - Brazil

^[3] Univ Fed Fluminense, Inst Comp, BR-24210240 Niteroi, RJ - Brazil

• Tipo de documento: Artigo Científico
• Fonte: RAIRO-OPERATIONS RESEARCH; v. 48, n. 2, p. 211-233, APR-JUN 2014.
• Citações Web of Science: 1

Resumo

The problem of finding structures with minimum stabbing number has received considerable attention from researchers. Particularly, {[}10] study the minimum stabbing number of perfect matchings (MSPM), spanning trees (MSST) and triangulations (MSTR) associated to set of points in the plane. The complexity of the MSTR remains open whilst the other two are known to be NP-hard. This paper presents integer programming (in) formulations for these three problems, that allowed us to solve them to optimality through IP branch-and-bound (B\&B) or branch-and-cut (B\&C) algorithms. Moreover, these models are the basis for the development of Lagrangian heuristics. Computational tests were conducted with instances taken from the literature where the performance of the Lagrangian heuristics were compared with that of the exact B\&B and B\&C algorithms. The results reveal that the Lagrangian heuristics yield solutions with minute, and often null, duality gaps for instances with several hundreds of points in small computation times. To our knowledge, this is the first computational study ever reported in which these three stabbing problems are considered and where provably optimal solutions are given. (AU)

• Processo FAPESP: 07/52015-0 - Métodos de aproximação para computação visual
• Beneficiário: Jorge Stolfi
• Modalidade de apoio: Auxílio à Pesquisa - Temático