ATHENA

The optimization of complex decision-making process is the subject of studies in operations research and mathematical optimization for many years. Generally, the origin of the problems comes from an implicit cutting of the problematic, which was made taking into account the nature of the problems and of the obvious complexity of trying to treat them jointly. These problems are addressed in a very precise way and the difficulty lies in how to make them communicate. For example, in the context of the supply chain, production decisions are strongly related to inventory management decisions as well as distribution policy. However, the problems are most often discussed separately. Recently, thanks to the advances in optimization techniques, thanks to performing software available for researchers and to advanced technology, it is now possible to tackle problems of great complexity. For example, addressing some decisional problems in an integrated manner rather than separately is possible and it is established that it provides solutions of a great interest. The goal of the project is to address complex optimization problems at an operational level (for instance production scheduling and vehicle routing problems). So that the study is complete, multiple contexts are considered. First, we consider that the decision environment is centralized and static, it means that the data are known with certainty and do not vary in time. The techniques of combinatorial optimization based on decomposition are then preferred to seek optimal or near optimal solutions. Secondly, we consider that the environment is centralized and static but that the integrated decision problems have distinct objective functions, possibly conflicting. So, multi-agent approaches (n the sense of “competing agents”) are preferred and compromise solutions are searched. Thirdly, still in a static environment, we consider now that the decision is distributed and that each of the integrated problems is under the responsibility of an autonomous decision-maker. We have to consider the cooperation mechanisms between decision-makers and find balanced solutions. Finally, in a last part, we consider that the environment is dynamic; it means that the problem parameters may change over time. The proposed solution is disturbed and the problem is to find, according to the definition of uncertainties, to develop methods that find robust solutions. Four partners are involved in the project: the “Laboratoire d’Informatique” of the University of Tours, the laboratory called Heudiasyc of the Université de Technologie de Compiègne, the LAAS-CNRS of Toulouse and the LIMOS of the University of Clermont-Ferrand. The project will last 48 months and the main funding request is composed of two co-supervised thesis (LI-LAAS and LIMOS-Heudiasyc).