Project Manager : Jean-claude HENNET




Keywords :cognitive engineering, knowledge engineering, Game Theory, Multicriteria Analysis, Monitoring, Risk Management, Human Machine Interfaces, holonic systems, supply chains.


The project aims at the development, analysis and use of knowledge models of human behavior in different environments where human beings perform tasks of exploration, decision-making (actors) and execution (operators). In a complementary manner, the project attempts to design and organize systems for control, monitoring and management including the allocation of tasks between humans and machines and their interfaces.


Methodology: The research project focuses on systems that involve human or decision-making bodies developed by humans in performing tasks in a socio-technical world. Therefore, the models developed, validated and used are based on knowledge engineering, cognitive engineering, discrete event systems and operations research. In particular, the modeling of structure and behavior of complex organizations with human and technological dimensions is performed with the aim of solving problems related to their security, autonomy and performance. The solutions to these problems are in the form of computer tools for decision support, assistance in solving these problems, or assistance in the validation via simulation / emulation. In the first step of the methodology, we develop methods for extracting knowledge from a corpus that can be made of various traces: textual traces, traces of activity (interaction with the device during the execution of tasks), interviews of experts, etc.. The domain of knowledge extracted is structured either on a semantic basis (objects, actions, reasoning), or, in the case of knowledge with poor semantics, by the definition of distance or similarity functions. The construction of a conceptual modeling or behavioral models from the field of cognitive engineering can then help understanding the reasoning mechanisms of human operators engaged in the tasks studied. Problem solving is then derived from the use of formal operations, analytical methods or tools that necessarily apply to formal models: discrete event models (Petri Nets, DEVS, Markov models), mathematical models, optimization problems. The transition from a conceptual modeling to one or several formal models is therefore a key step in our methodology. Our ambition is not to automate this transformation, but rather to improve and validate by knowledge engineering the formal models proposed in specific application areas: risk management, re-engineering of monitoring devices, design of adaptive Man-Machine Interfaces, management of production and flows in supply chains, information seeking on a website, learning environment, etc.. Interactions between players or between groups of autonomous actors often take the form of negotiations, consisting of sequences of proposals to arrive at mutually acceptable decisions. The main factors for building actions and decisions are the goals, knowledge and procedures. Game theory formalizes these factors and put them in interaction. It improves understanding of the mechanisms of negotiation and is the basis for the development of powerful tools for decision support. At the algorithmic level, Operations Research provides both support for modeling problems (SED models, mathematical programming, multicriteria formulations) and for solving them using powerful mathematical and computing tools. Finally, some of the studied systems may be represented by models operating without any hierarchical control, developed on the basis of ad hoc integration of elements from the theory of priorities and the theory of open hierarchies. In this context, we are interested in distributed decision-making structures, for a holonic, isoarchical and multi-objective control, based in particular on the AHP (Analytic Hierarchy Process) and ANP (Analytic Network Process) methods. Validation experiments will be done by using a distributed simulation environment HLA (High Level Architecture), and eventually through interaction with a real system ’in the loop’: the Versatile Production Platform of LSIS. From a general standpoint, simulation allows a detailed description of the system and computing its performance. It also helps to validate the analytical models developed for solving problems. More specifically, simulation of behavioral models is a goal of research in that it provides: • a means for verification of models; • a means of reflection and learning for system actors; • a way forward to consider new forms of organization

Application sectors:

For the management of critical situations, the project focuses on the design of safer Human Machine Interface to avoid errors and improve decision making in supervision tasks. In this context, we use an approach based on simulation: simulation of the controlled equipment and the Human Machine communication interface using discrete event simulation and simulation of human behavior using cognitive models. A privileged application domain is to assist the task of an operator in a computer-mediated activity. For example in the field of information retrieval, e-learning or sharing experiences, the goal is to understand and analyze the path of the user, model the decision process to be able to offer dynamic online help or restructure the documents to improve the effectiveness of environments (depending on the models built). Networks of production, supply and demand for goods and services are studied mainly under distributed decision-making structures and mechanisms of negotiation, by two complementary approaches: the strategic and cooperative approach. On such complex decision environments in production, the research also involves the piloting of flexible automated production systems (Intelligent Manufacturing System) or partnership-self-organized logistics networks (Intelligent Supply Chain). The exploration of these approaches will be strengthened by studying the use of the infotronic technology, with a vision of the type ’ambient computing’. Methodological convergences have been developed between project members on problems of analysis and risk management in supply chains. This problem, yet little discussed in the literature, offers vast fields of scientific investigations and industrial applications.