|
The Group
With the recruitments of Xavier Gandibleux in 2004 as full
professor in Computer Science by the University of Nantes,
followed by Anthony Przybylski in 2007 as senior lecturers in
Computer Science, the core of the group consists of two
permanent members. From 2006 to 2008, Matthias Ehrgott has been
permanent member of the group as director of research at the
CNRS.
Non-permanent members are involved in the group for some
periods. Sana Belmokhtar (from Ecole des Mines de Saint-Etienne)
has joined us as researcher in 2006-2007. She is now assistant
professor at ESIEE PARIS. Hadrien Hugot (from LAMSADE,
University of Paris-Dauphine) got in 2007-2008 a post-doctoral
position funded by the CNRS. Benoit Guédas and Guillaume Pinot
(both from IRCCyN, Ecole Centrale de Nantes) have been
associated to the group while they served as lecturer in the
department. Zohra Aoudia (from Université de Bejaia) has joined
us for a period of 18 months.
Master and PhD students contribute to the works of the team.
Recently Quentin Delmée has defended its PhD thesis and another
PhD student will joint us soon. Former PhD students who prepared
their thesis under our supervision are Xavier Delorme (now
professor, Ecole des Mines de Saint Etienne), Anthony Przybylski
(now assistant professor, University of Nantes), Fabien
Degoutin, Julien Jorge, Benoit Guédas, Hugo Fouchal, Aurélien
Mérel, Thomas Vincent, and Audrey Cerqueus (now assistant
professor, Ecole des Mines de Saint Etienne).
Research Activities
Our work, mainly based on discrete optimisation in Operations
Research, focuses on the accumulation of knowledge towards the
development of advanced optimization methods that are capable of
solving complex optimization problems in reasonable time. The
optimization problems of interest are reference problems in
discrete optimization and their application in socio-economic
contexts, such as railway transportation (capacity of railway
infrastructure), and communication networks (routing policies,
deployment of new infrastructure).
In this context, the motivation characterizing the research
direction of the group is to study, model, and solve large scale
multiobjective (mixed) integer optimization problems. Procedures
for these problems are essentially problem dependent and employ,
among others, efficient enumerative methods (two phase method;
branch and bound/cut) or hybrid optimization techniques
(multiobjective metaheuristics and exact algorithms). Our
research directions are:
- Fundamental:
Study, characterization, and understanding of (mixed)
integer and combinatorial multiobjective optimization
problems.
- Methodological and algorthmical:
New techniques and methods for the solution of large scale
(mixed) integer and combinatorial multiobjective
optimization problems; Development of algorithms to improve
the efficient solution of NP-hard single and multiobjective
problems.
- Validation and verification:
Application to real world multiobjective optimization
problems with the ultimate goal of being able to solve
concrete problems in complex real world environments
(production systems, transport, communication, environmental
policy). Most applications are collaborations with
industrial partners.
Some Results of Our Work
- State of the Art Annotated Bibliographic Survey.
For many years we collected and summarized the literature on
multi-objective combinatorial optimization (MOCO) problems.
In 2000 and in 2002, papers reporting our synthesis have
been published. Later we did a similar work about
multi-objective metaheuristics (MOMH).
M. Ehrgott, X. Gandibleux (2000). A Survey and annotated
bibliography of multiobjective combinatorial optimization. OR
Spektrum, 22(4): 425-460.
- Path-relinking for multi-objective optimization.
Approximation methods for MCDM problems have received a lot
of attention in recent years. With two Japanese colleagues
we introduced the path-relinking concept for MOMH with
success for many MOCO problems.
X. Gandibleux, H. Morita, and N. Katoh (2004). Evolutionary
operators based on elite solutions for bi-objective
combinatorial optimization. Chapter 23 in Applications
of Multi-Objective Evolutionary Algorithms (C. Coello
Coello and G. Lamont Eds.), pp. 555-579. Advances in Natural
Computation Vol. 1, World Scientific, Singapore.
- Two phase method for MOCO problems.
Introduced in the nineties by Ulungu and Teghem, this method
has been considered as a generic method for bi-objective
optimization problems. One of the major contributions
Anthony Przybylski's PhD thesis has been the generalisation
of this method for dealing with problems with more than two
objectives.
A. Przybylski (2006) Méthode en deux phases pour la
résolution exacte de problèmes d'optimisation combinatoire
comportant plusieurs objectifs : nouveaux développements
et application au probléme d'affectation linéaire. PhD
thesis, University of Nantes, December 2006 (In French).
- Exact and efficient procedures for solving the linear
assigment problem with two and three objectives.
Considered as a fundament optimization problem, we proposed
algorithms for the exact solution. They have been
demonstrated to be the most efficient algorithms considering
the literature available.
A. Przybylski, X. Gandibleux and M. Ehrgott (2008).
Two-phase algorithms for the bi-objective assignment
problem. European Journal of Operational Research
185(2):509-533
- Railway infrastructure capacity.
The question investigated here can be stated as follows:
«How many trains can go through a junction or a station?».
With the cooperation of partners we developed methodologies,
algorithms and software dealing with this question. The case
studies are real situations from the SNCF (France) and the
DB (Germany) networks.
J. Rodriguez, X. Delorme, X. Gandibleux, Gr. Marlière, R.
Bartusiak, F. Degoutin, and S. Sobieraj (2007). RECIFE:
models and tools for analyzing rail capacity. Recherche
Transports Sécurité, 95:19-36.
Some Major Events Involving the Group Members
The members of the group have been involved in several
international scientific events, four of which are immediately
related to the MCDM field.
- MOMH 2002:
Multiple Objective Metaheuristics International Workshop,
November 4-5, 2002, Paris, France
- MOPGP 2006:
7th International Conference on Multi-Objective Programming
and Goal Programming, June 12-14, 2006, Loire Valley
(Tours), France
- EMO 2009:
5th International Conference on Evolutionary Multi-Criterion
Optimization. April 07-10, 2009, Nantes, France
- RAMOO 2015 and 2018:
Recent Advances in Multi-Objective Optimization, June 19, 2015 and November 15, 2018, Nantes, France
- Juliaday Nantes'2019:
Juliaday on "Julia and Optimization", June 17, 2019, Nantes, France
At the national level, the French Working Group dedicated to
Multiple-Objective Programming (PM2O) has been co-founded on
1999 by Xavier Gandibleux. He has served as the coordinator of
this group for four years.
Visitors and Collaborators
Invited professors who visited us these last years for a period
of one month were e.g. Kathrin Klamroth in 2005 (University of
Erlangen-Nuremberg, Germany), Eric Taillard in 2006 (HEIG-VD,
Switzerland), Margaret Wiecek in 2007 (Clemson University, USA),
and Eckart Zitzler (ETH Zürich, Switzerland). The group also
hosts visiting PhD students: Daniel Salazar Aponte from
University Las Palmas de Gran Canaria (6 months from Sept 2005)
and Andrea Raith from Auckland University (3 months from August
2007). If you are interested in visiting us, please contact us.
We have a long tradition of working with colleagues in OR and
MCDM. Several collaborations are on-going with Matthias Ehrgott
(Lancaster University, UK), Stefan Ruzika (University of
Koblenz, Germany), Kathrin Klamroth (Bergische Universität
Wuppertal, Germany), Sophie Parragh (Johannes Kepler Universität
Linz, Austria), Karl Doerner (University of Vienna, Austria),
Dario Da Silva (University of Nottingham, UK), Naoki Katoh
(Kyoto University, Japan) and Hiroyuki Morita (Osaka Prefecture
University, Japan) to name a few.
Since 1999 we are involved in research works related to railway
transportation. Joaquin Rodriguez (from IFSTTAR, the French
National Research Institute on Transportation and Security) is
one of our collaborators on this topic.
To conclude this section, we are collaborating also with
colleagues of regional institutions, e.g. Saïd Hanafi
(Université Polytechnique Hauts-de-France), Frédéric Saubion
(University of Angers), and Marc Sevaux (University of South
Brittany-Lorient).
Projects
The group is regularly involved in regional (e.g. LigeRO),
national (e.g. ANR GUEPARD) and international (e.g. ANR-DFG
vOpt) research projects.
Software
RECIFE is a decision support system specifically designed
for the analysis of railway infrastructure capacity. For a given
station or node of the network, various functionalities such as
verifying the feasibility of expected traffic, studying
infrastructure saturation and stability of resulting timetables
are offered to a decision maker. Two geographical situations
have already been studied: The Pierrefitte-Gonesse node located
north of Paris and the Lille-Flandres station.
vOptSolver is an open-source
ecosystem developed in Julia, for modeling and solving
multi-objective linear optimization problems (MOCO, MOIP, MOMIP,
MOLP).
It integrates several exact algorithms for computing a complete
set of non-dominated points for structured and non-structured
optimization problems with at least two objectives. vOptSolver
is composed of two packages integrated to the collection of
Julia packages: vOptGeneric and vOptSpecific.
To Contact Us:
XAVIER GANDIBLEUX
Université de Nantes
UFR Sciences / Département informatique
2, rue de la Houssinière BP 92208
F-44322 Nantes Cedex 03 - FRANCE
http://www.univ-nantes.fr/gandibleux-x
|
