Mise en oeuvre d'un modèle de données à base d’agents pour optimiser le processus de la généralisation automatique des données géographiques

Abstract

The generalization process of geographic data consists to derive from detailed geographic data, less detailed data adapted to the users needs (for example level of detail, application context). Several approaches are proposed to automate this process, the Multiagent system and optimization approaches are usually used to solve the problems in this domain but they had some disadvantages. The purpose of this thesis is to hybrid these two approaches to address their inconvenients and benefits from their advantages. In this approach, the geographical objects (roads, buildings, etc...) are modeled as agents which selfgeneralize basing on the results of the genetic algorithm that they run. Agents are given powers of perception of their environment and communication capabilities and are composed of three main modules, module of generalization, optimization module and module of transformation of topological relationships between regions and the new concept of ribbons. Each agent performs its genetic algorithm to find solutions to meet the cartographic constraints in satisfaction situation. The supervisor agent synchronizes the transformations applied by the agents to avoid as much as possible conflicts between neighboring agents and also it can apply a global genetic algorithm in the case in which there are conflicts are not resolved at the local level of the agent.