L'optimisation du déploiement des réseaux optiques. Considérations sur l'incertitude de la demande.

Abstract : The recent increase in bandwidth requirements in telecommunication networks drives operators to deploy new infrastructures. For the fixed access network, optical fibers is the chosen technology. Due to financial stakes and the inherent complexity that goes with such deployment, it is crucial to optimize its cost while ensuring both the quality of service expected level and the deployment engineering rules. This thesis is the continuation of a former study on this issue which proposed a modeling of the problem under the form of a mixed integer linear program. An extensive effort has been put into the improvement of the solving algorithm, and several research ways had been considered to follow up on this work. Among the major issues that were identified, the tackling of demand uncertainty has a big part in this study. Indeed, as future clients are not known yet, it is not possible anymore to dimension the network according to fixed and precisely known data. In that case, the problem becomes a combinatory uncertain optimization problem. The choice that was made is to deal with it under the scope of robust optimization. This approach enables one to protect itself against uncertainty by ensuring feasibility in every case while optimizing the "worst case" scenario. The resulting formalism often leads to problems that are complex to model and to solve. Indeed, they use multi level formulations where decisions are taken sequentially, before or after the revealing of the uncertain scenario. Adapted algorithms have been developed to enable the application of robustness to optical fiber network deployment. These algorithms, exact or not, helped to gather real strategic information for future deployments. From these investigations, focused on optical deployment, some results have been extended and generalized to other robust optimization problems. Probability bounds on the relevance of uncertainty sets and probabilistic estimation of future costs in two-stage robust optimization problems are among them. In parallel of the work on uncertainty that is a major part of this study, other fields have been investigated. Indeed, in order to improve the taking into consideration of future costs of the network (due to maintenance, administration, etc.) that are, on the long run, the most important, an approach has been proposed that enables one to take "best practices" into account within the optimization process. The integration of these considerations, grouped under the name OA&M (for Organization, Administration and Maintenance), has been validated by developed costs macro-models that are able to estimate the future gains that can be expected, due to these new constraints. Finally, our efforts focused on the solving of a specific version of the problem in tree-graphs, with taking cabling constraints into account. For this problem which was already studied, a new dynamic programming algorithm has been proposed. It heavily relies on the problem's properties and uses them to explore a very limited number of solutions, while remaining optimal. The algorithm performances showed a clear improvement of the computing time compared to mixed-integer linear programming based approaches. In the end, this work uncovered new areas of inquiries, especially on alternative tackling of uncertainty, as well as a better taking into account of cabling constraints within the optimization.
Document type :
Theses
Complete list of metadatas

Cited literature [51 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00960733
Contributor : Cédric Hervet <>
Submitted on : Tuesday, March 18, 2014 - 4:11:35 PM
Last modification on : Wednesday, July 3, 2019 - 10:48:04 AM
Long-term archiving on : Wednesday, June 18, 2014 - 1:31:04 PM

Identifiers

  • HAL Id : pastel-00960733, version 1

Collections

Citation

Cédric Hervet. L'optimisation du déploiement des réseaux optiques. Considérations sur l'incertitude de la demande.. Optimisation et contrôle [math.OC]. ENSTA ParisTech, 2013. Français. ⟨pastel-00960733⟩

Share

Metrics

Record views

437

Files downloads

700