Aller au menu Aller au contenu
Laboratoire des Sciences pour la Conception, l'Optimisation et la Production de Grenoble

In short
The objective of this chair is to develop novel and self-adaptive decision making methods for
data driven and ultimately self-reconfigurable network of production systems (supply chains)
in the context of industry 4.0. Optimization, simulation and AI techniques will be coupled to
this end.

Chair holder(s): Gülgün Alpan (G-SCOP/Grenoble INP)
Chair title: AI for data-driven and self-configurable supply chains
Other chair participants: Hadrien Cambazard (G-SCOP/Grenoble INP), Iragaël Joly
(GAEL/Grenoble INP), Pierre Lemaire (G-SCOP/Grenoble INP), Bernard Penz (GSCOP/
Grenoble INP), Zakaria Yahouni (G-SCOP/Grenoble INP)
International collaborators: Uwe Clausen, Sören Kerner, and Michael ten Hompel
(Fraunhofer IML / TU Dortmund)
Industrial partners: Renault Group

Scientific objectives and context
Operations and supply chain management (O&SCM) involves in general some form of
explanatory and predictive analytics (forecasting, simulation) and prescriptive analytics
(optimization) that are often used together. Despite their adequacy to solve O&SCM
problems, simulation and optimization may lack precision or speed as system complexity
increases, may fail to handle complex objectives and/or constraints, and lack reactivity to
cope with fast evolving environment of industrial and service systems. To overcome these
downsides, we propose to develop advanced decision making tools by coupling optimization
and simulation with AI techniques.

Research directions
• O&SC applications often require (or generate) domain-specific knowledge resulting in
significant amount of customization of the algorithms used. We will explore data analytics
and machine learning to extract generic knowledge, which can be used to better
characterize the behavior of an algorithm and to anticipate its performances.
• Techniques based on reasoning such as constraint programming could be better suited to
handle fundamental production planning problems (e.g. lot sizing), in the presence of
complex objectives and additional combinatorial constraints. We will explore reasoning
mechanisms to design new solution techniques that deliver fast and high quality solutions
for such problems, with a quality guarantee (an optimality gap) to the decision maker. A
valuable contribution for practical use is to incorporate these mechanisms into industrial
information systems (e.g. a demonstrator using an open source ERP such as Odoo)
• When a decision aid tool is deployed in industry, the initial fine-tuned parameters are
rarely updated during utilization. Data analytics and learning mechanisms will be included
to develop self-readjusting decision tools, which can adapt to an evolving environment.
• Simulation is commonly used in industry for “what-if scenario analysis”. Today, these
scenarios are mostly defined based on manager’s own judgment and expertise, and may
not include all potentially promising scenarios. AI techniques can help in defining the
scenario bases for high quality solutions and fast analyses.

Motivation and expected impact
Today’s markets are characterized by high product diversity, uncertain and volatile demand,
shortened product-life-cycles and production lead times. To stay competitive, the companies
need to deploy flexible manufacturing systems, robust and traceable supply chain networks,
and agile, self-adapting operations planning processes. While “Industry 4.0” provides the
technological background (e.g. IoT, Io(Thinking)T, smart sensors,…), the methods developed
in this chair will provide the necessary decision-making tools, to this end.

mise à jour le 10 octobre 2019


Université Grenoble Alpes