Know more

Members of the jury :

• Reviewers before defense :
  • Fabien MICHEL, Associate Professor, IUT, Montpellier, France
  • Emmanuelle GILOT-FROMONT, Professor, VetAgro - Sup, Lyon, France
• President :
  • Catherine BELLOC, Professor, Oniris, Nantes, France
• Examiners :
  • Christine LARGOUET, Associate Professor, Agrocampus Ouest, Rennes, France
  • Nicolas SABOURET, Professor, Centrale Supélec, Gif-sur-Yvette, France
  • Samuel SOUBEYRAND, Research Director, INRAE, Avignon, France
• PhD supervisor :
  • Sébastien PICAULT, Researcher, INRAE, Nantes, France
• PhD co-supervisor :
  • Mathieu ANDRAUD, Researcher, ANSES, Ploufragan, France

Abstract :

Epidemiological modelling has proven to be an essential tool for health crises. Understanding, anticipating and controlling diseases are major public health and veterinary medicine challenges, and need a detailed representation of the processes involved. The spatio-temporal organization of the host population is one of the factors influencing infection dynamics and control measures, but remains difficult to represent explicitly in epidemiological modelling. Artificial intelligence computing methodologies (agent-based system) make it possible to develop solutions for explicitly representing highly structured complex systems in mechanistic epidemiological models.
This PhD thesis aims to propose a generic, flexible and modular solution, based on artificial intelligence, to address these challenges. We have developed a multi-level agent-based organizational design pattern for representing the coupling
between agents, environments, and organizational levels. The pattern has been implemented into the EMULSION framework and applied to model, explicitly with fine granularity, the spread of a swine influenza virus and of the porcine reproductive and respiratory syndrome virus in farms practising batch management.
These models confirmed the significant role played by farm management, and identified the key elements of transmission. The design pattern and its applications open up new perspectives for multilevel agent-based systems and modelling.

Keywords :

multi-level multi-agent system, artificial intelligence, design pattern, epidemiological modelling, porcine reproductive and respiratory syndrome, swine influenza A virus.