You will find here some of my current or past academic projects and industrial contracts.

Stable project

The Stable project funded by STIC-AMSUD is entitled Solving Combinatorial Optimization Problems with Stable Sets Constraints. It aims to study optimization problems where some of the constraints are related to stable sets, such as the Maximum Stable Set problem, the Coloring Problem, etc. This class of problems is interesting both for its practical applications and for the numerical and algorithmical challenge that it represents.

Our contributors come from research teams from Brazil (PARGO - Federal University of Ceará, Fortaleza), France (LIA, EA 4128 - University of Avignon and PriSM UMR CNRS 8144 - University of Versailles-Saint Quentin), Chile (Universities of Santiago de Chile and Concepción) and Argentina (National University General Sarmiento).

Pajero projet

Le projet PAJERO a pour objectif de développer une plateforme logicielle permettant, pour la première fois, la gestion complexe de ressources multiples (ressources humaines, planification, gestion des temps et activités associées…).

Il réunit Horizontal Software, en tant que chef de file, deux PME – EquiTime et Caps Entreprise – ainsi que trois partenaires académiques : le CRIL (Université d'Artois), le PRiSM (Équipe Caro, Université de Versailles-Saint-Quentin-en-Yvelines) et l'I3S (Pôle MDSC, Université Nice-Sophia-Antipolis).

Ce projet représente un investissement total de 16,5 millions d'euros sur une période de 48 mois.

Horizontal Software et ses partenaires ont reçu, le 18 octobre 2011, une aide de 7,6 millions d’euros d’OSEO dans le cadre du programme ISI, Innovation Stratégique Industrielle.

Le projet Pajero a été labellisé par le pôle de compétitivité System@tic.

The Pajero Project founded by OSEO is head by the HSW compagny. Two compagnies are envolved in the project EquiTime and Caps Entreprise

From an academic point of view PRiSM , the CRIL and I3S are the laboratories which cooperate for the project.

The CARO Team, and more specially me are envolved in the projet to parallelize the different solvers used in the project. We actually port the Google Constraint Solver Or-Tools on top of the Bob++ framework.

Horus projet

Bus, tramway and train scheduling and the corresponding workforce planning are major issues for public and private transportation companies. The complexity of this task usually justifies its decomposition in sequential steps : timetabling of the bus, tramways or trains, assignment of the vehicles to the transportation tasks, building shifts, and finally, individual rostering. Currently, only the major companies have access to optimisation modules. In the last decade, EURODECISION integrated its optimization component, LP-EasyDriver, to the activities of RATP and, more recently, VEOLIA. This offer, as well as similar ones from competitors, is based on column generation techniques. In the railway sector, the use of optimization modules is less expanded because of the higher complexity of the problems.

The main goal of HORUS project is to overcome the following challenges :

  • extension of the bus optimisation systems to the railway sector (higher combinatorial complexity and higher business complexity),
  • global optimization of the whole problem (time tabling and crew scheduling),
  • access to multi-core computing through algorithms parallelization,
  • access to the optimization algorithms through free software and shared web services.

From an academic point of view PRiSM and CReSTIC laboratories, which are well known for their specialization in combinatorial optimization and parallel computing, propose in this project significant contributions in the resolution methods dedicated to these kinds of problems: arcs-states modeling, column generation algorithms, innovative metaheuristic (Trust Branching Path) and hybridization techniques including metaheuristics and exact approaches.

Choc projet

The CHOC project was funded by the ANR-CICG (2005-2009), in one sens this project is the following of the Doc-G project.
  • the Opale Team of the PRiSM laboratory (UVSQ)
  • the MOAIS Project (INRIA Grenoble)
  • the GILCO Laboratory (Grenoble)
  • the Dolphin-INRIA Futur Project
Many extensions have been developed for Bob++, ParadisEO and Kaapi. We have the best sequential and parallel solving algorithms for the Q3AP. In exact solving, we have proposed algorithmic and programming improvements (like the use of the SSE instruction set) for the computation of low bounds. Furthermore, the management of symmetrical configurations has allowed us to solve optimally for the first time an problem instance of size 16. In approximated solving using metaheuristics, this project has allowed to develop an hybrid genetic/simulated annealing algorithm. It has shown that using a suitable initialisation procedure to generate the initial population, we have been able to reach for the first time solutions for problem instances of size 18 and 22. We have also mixed exact and approximated methods in two ways: (1) an exact method is used as intensification inside a genetic algorithm, and (2) a ParadisEO metaheuristic is used by an exact method to improve dynamically the solutions generated. The three softwares frameworks of this project:
  • Bob++ : framework to develop exact solvers for Combinatorial Optimization problems, bobpp
  • ParadisEO : framework to develop approximated solvers for Combinatorial Optimization problems,
  • Kaapi : multithreaded runtime environment to run parallel applications on large heterogeneous machines (until around 3600 cores on Grid5000 in 2007 and 2008)

e-Toile projet

The e-Toile project was funded by the RNTL (réseau national de recherche en logiciel) an agency head by the French Ministry of Research. It starts at the begining of 2001 and ends at the end of 2003. The goals was :
  • To Develop a GRID Testbed based on
    • VTHD and RENATER backbones
    • for testing computing intensive and data-intensive applications
  • To Develop a Middleware prototype
    • that integrates the most recent and relevant works of the french computer science laboratories (INRIA, CNRS)
    • Focused on enhanced communication services
  • To Provide a high performance network and software support for the ACI-GRID initiative
I have been involved in the development of the Middleware prototype, and also to perform some application tests the deploy a QAP solver based Bob++ on top of kaapi.

DOC-G projet

The project DOC-G was funded by the ACI-GRID. The goal was to solve large instance of combinatorial optimization problems on grid.
  • Collaboration with the MOAIS Project (INRIA Grenoble)
  • Collaboration with the Dolphin-INRIA Futur Project