Flexibilité et performance de codes de calcul en optimisation et simulation

Abstract : This manuscript presents the synthesis of research works in optimization and simulation with the aim of developing flexible and performant software tools. The contributions are centered on three main domains:

(i) combinatorial optimization, with the resolution of problems of hypermedia documents synchronization and resources optimization in cloud computing;

(ii) numerical simulation, with an original outline of optimization-simulation coupling for solving a bus routing design problem in an urban transportation network, and a methodology to minimize the vulnerability of a grassland ecosystem facing climate changes;

(iii) generic software development, with design patterns for developing optimization algorithms, and a library to extend generic programming to "concepts" and thus to allow improving the quality of computer codes.

This multidisciplinary approach is explained by the complementarity of these domains when tackling the studied problems. Notably, combinatorial optimization allows designing very efficient algorithms to solve operational research problems, but is based on a usually too simplistic modeling of the studied system. Numerical simulation, which generally allows representing more precisely the system, can be coupled with the optimization process to identify better applicable solutions.

The studied systems can also be very complex and can need extremely time-consuming optimization algorithms and simulation models. Software development becomes thereby a crucial element where usual software engineering methods are not always enough to meet high performance computing needs. It seems thus relevant to study the possibilities offered by generic programming and metaprogramming to combine, with few compromises, flexibility and performance in computer codes.
Complete list of metadatas

Cited literature [121 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/tel-01509550
Contributor : Bruno Bachelet <>
Submitted on : Tuesday, April 18, 2017 - 10:41:15 AM
Last modification on : Monday, January 20, 2020 - 12:14:05 PM
Long-term archiving on: Wednesday, July 19, 2017 - 12:25:38 PM

File

Identifiers

  • HAL Id : tel-01509550, version 1

Collections

Citation

Bruno Bachelet. Flexibilité et performance de codes de calcul en optimisation et simulation. Génie logiciel [cs.SE]. Université Blaise Pascal - Clermont-Ferrand II, 2016. ⟨tel-01509550⟩

Share

Metrics

Record views

602

Files downloads

659