The multi-agent cooperative navigation system Mawa: a model of dynamic knowledge specialization for a user-centric analyse of the Web

More and more, multi-agent models are exploited to build complex applications or applications in which no entity has a global vision of the situation. As a result, the Internet is becoming an adequate environment for multi-agent systems. Such systems are open; agents can be added, removed or modified at any moment. But, this characteristic is rarely illustrated. In this paper, we present the deployment of the multi-agent cooperative navigation system Mawa. At the beginning it is only composed of an assistant and a connector; the number of agents grows because agents are able to reorganize and restructure information. Firstly, we present functionalities of Mawa and our model of agents: we considered the Web as a cooperative environment for information retrieval. Then, we describe the agents' behaviours and the system dynamic