The Marseille Constraint based Configuration Group

The web resource of the Configuration Team, member of the InCA Team, from the LSIS Laboratory, Marseille, France

Constraint Based

Configuration

Configuration refers to the tools and techniques that can be used to simulate the construction of complex products out of elementary components. Constraint based configuration uses constraint programming techniques to address the problem.

People

Laurent Henocque (Associate Pr.)

Nicolas Prcovic (Assistant Pr.)

Mathieu Estratat (Post doc.)

Mathias Kleiner (Doctor)

and formerly

Stephane Grandcolas (Assistant Pr.)

Group Objectives

The group wishes to explore the possibilities offered by finite model search, applied to a range of first order problems called "Constrained Object Models". More precisely, we wish to assess the usefulness of this approach to address problems commonly viewed as cognitive, or artificial intelligence related.

Algorithms

First order generalizations of constraint programming to address configuration, but raise a number of complexity issues. Among them are the problems raised by isomorphisms. One field of our activity relates to tractable symmetry elimination algorithms for configuration.

Alternatively, incomplete finite search target scalability by improving response times in the presence of satisfiable problems

Modelling in higher order languages

We explore the possibility of using higher order relational set theoretic languages to model configuration problems.

Applications

Natural Language Processing

Building semantic representations from natural language texts can be viewed as a finite model search problem, and thus offers a large field of AI applications of configuration

Semantic Web Service Composition

Again, composing workflows and Web Services can be addressed as a finite model search problem

Technologies

Constraint Programming

We use the Ilog JConfigurator program as a complete finite model search engine for configuration problems.

Linear Time Symmetry Elimination

In order to address scalability issues in this highly combinatorial context, we investigate tractable symmetry elimination algorithms for configuration problems.

Stochastic Search

We use "ant colony optimization" as a stochastic/reinforcement learning incomplete search procedure for configuration problems.

Particle Swarm Optimization

We use PSO to explore the problem space of stochastic algorithm parameters.

Acknowledgements

This work is mainly funded by the french CNRS and Marseille Universities.

It has also greatly benefitted from the financial and technical support of the Ilog Company