Methodological aspects of ring network design
- 98 Downloads
We analyse ring network design problems, with the aim of satisfying at a minimum cost a given demand matrix. The network model considers link capacities ranging over a discrete set of values, and both fixed and linear costs, which pose severe limitations to the possibility of finding global optimum solutions. An approximated version of the problem — which neglects the discrete nature of link capacities — is here formulated as a multicommodity flow problem with linear cost function and fixed costs on a hypergraph. Such a problem is NP-hard. A greedy algorithm, which extends the one proposed by Minoux, is devised. A more general solution approach is also developed, which consists of a decomposition of the general problem into two major steps. Aim of the first is to design a partial network which satisfies a given percentage of the overall demand. This task is formulated as a pure combinatorial problem, in terms of 0–1 linear programming. The second step consists of finding a completion of the partial network, and can be formulated as a classical multicommodity graph-flow problem with fixed costs. Prior to both approaches, effective cluster analysis techniques are suggested for reducing the input size, according to demand, and/or to geographical, logical, and economic criteria.
KeywordsLink Capacity Network Design Problem Internal Demand Partial Network Linear Cost Function
- M. Jambu, M-O. Lebeaux (1983): Cluster Analysis and Data Analysis, North Holland.Google Scholar
- M. Minoux (1976): "Optimisation et Planification des Résaux de Telecommunications", in G. Goos, J. Hartmanis, J. Cea (eds.): Optimization Techniques, Lecture Notes in Computer Science, Vol. 40, Springer-Verlag, 419–430.Google Scholar
- C. Arbib, U. Mocci, C. Scoglio: "Optimization of Ring Network Design: an Overview", FUB Int. Rep. 4B00491Google Scholar
- P. Nobili, A. Sassano: "A Branch and Cut Algorithm for the Set Covering Problem", to appear.Google Scholar