Skip to Main content Skip to Navigation

Modélisation et Gestion de Flux par Systèmes Multiagents: Application à un système d'aide à la décision en épidémiologie

Abstract : Our work consists in studying complex systems. Some complex problems have no analytical solution or have a too complicated solution to be easily implemented. The traditional analytical methods are declined either from a global point of view (holistic) where the complexity is separated into distinct elements to represent the behaviour of only by only one entity (reductionism). So, the various levels of complexity of the system's organisation and their overlap are underestimated. In fact, the evolution of this type of systems is almost impossible to anticipate because of the important number of present entities and their interactions.
Another approach is based on the modeling of the behavior of each element which participates in this evolution likewise on the modeling of their interactions with the other elements and the environment. During these interactions, different kinds of data can be exchanged (information, money, food, virus, etc...). Each exchange can be interpreted as a flow. The principle is to connect entities which have simple characteristics. This allows them to interact together in a specific environment in order to obtain a general higher level behaviour. As we can observe in social animal societies, the collective performance emerges from these direct or indirect interactions: this is the result of a self-organization process during which the environment and the community structure themselves. The stochastic evolution of such systems does not allow them to be characterized completely. Consequently, to model and understand the circulation of flow in this type of systems, we need to make use of simulation.

In order to take into account the multiple specifications of complex systems in their modeling, Multi-Agent Systems (MAS) represent a particularly suitable method. In this way, environmental phenomena are represented as the consequences of the interaction of agents acting in parallel, each agent being a reactive and autonomous entity.
Our work focuses on the study of flow circulation in complex systems by the development of Agent-Oriented Simulation (AOS). A particular application of this flow management method is to simulate the circulation of a parasite (Cryptosporidium parvum) in an ecosystem. The principal objective is to better understand the various episodes of infection within different host populations (animal or human) under specific constraints. Because of the fact that this parasite is particularly resistant to the traditional disinfection methods, it is necessary to enhance the prevalent factors acting in the contamination and the propagation of C. parvum.
To this end, the data, received from biologists, enabled us to design a tool of experimentation in epidemiology. This AOS allows the exploration of the system behaviour where Cryptosporidium spp. circulates. Thus, different scenarios can be simulated. On the one hand, the software enables us to propose new assumptions on the parasite dissemination, and, on the other hand, the results show the accuracy of our modeling.
With an aim of bringing an autonomous decision support system to the biologists, we developed a system of higher level (meta-system) able to carry out the monitoring of an AOS. This meta-system, based on the concept of metaheuristic, tries to optimize the behavior of the system, according to precise problems, by evaluating the impact of preset factors on the evolution of the AOS. Thus, it is capable to interpret the simulation's results in order to allow causes to emerge which influence the parasite propagation by self-generation of scenarios.
Document type :
Complete list of metadatas
Contributor : Alexandre Weber <>
Submitted on : Friday, February 29, 2008 - 5:46:09 PM
Last modification on : Tuesday, November 24, 2020 - 2:18:22 PM
Long-term archiving on: : Friday, November 25, 2016 - 10:30:08 PM


  • HAL Id : tel-00259941, version 1



Alexandre Weber. Modélisation et Gestion de Flux par Systèmes Multiagents: Application à un système d'aide à la décision en épidémiologie. Modélisation et simulation. Ecole Centrale de Lille, 2007. Français. ⟨tel-00259941⟩



Record views


Files downloads