Modélisation Multi Agents Pour Systèmes émergents Et Auto Organisés
Resumen del Libro
In this work a multi-agent architecture for self-organizing and emergent systems (MASOES) is defined. This architecture allows the possibility of modeling a self-organizing and emergent system through a society of agents (homogenous or heterogeneous), who work in a decentralized way, with different types of behavior: reactive, imitative or cognitive. Also they are able to dynamically change their behavior according to their emotional state, so that the agents can adapt dynamically to their environment, favoring the emergence of structures. For it, a two-dimensional affective model with positive and negative emotions is proposed. The importance of this affective model is that there are not emotional models for studying and understanding how to model and simulate emergent and self-organizing processes in a multi-agent environment and also, its usefulness to study some aspects of social interaction multi-agent (e.g. the influence of emotions in individual and collective behavior of agents). On the other hand, a methodology for modeling with MASOES is specified, it explains how to describe the elements, relations and mechanisms at individual and collective level of the society of agents, that favor the analysis of the self-organizing and emergent phenomenon without modeling the system mathematically. It is also proposed a verification method for MASOES based on the paradigm of wisdom of crowds and fuzzy cognitive maps (FCMs), for testing the design specifications and verification criteria established such as: density, diversity, independence, emotiveness, self-organization and emergence, among others. It also shows the applicability of MASOES for modeling diverse case studies (in a diversity of contexts) such as: Wikipedia, Free Software Development and collective behavior of pedestrians through the Social Force Model. Finally, the two models proposed in MASOES: the initial multi-agent model and the model with FCMs based on that initial multi-agent model complement…
