Abstract
The reflexions lead around the idea of collective (or collaborative) intelligence suggest, that those emerge within a system in interaction of which technical, social and semiotic factors are strongly imbricated ( PIERRE LEVY , 1994). In order to leverage this challenge, this thesis propose to mix two innovative concepts. Firstly, the collaborative ubiquituous space is used to catalyse the interactions between groups of agents and services. Secondly, an infrastructure with distributed resources is used to dynamically coordinate the resources and the security according to quality of service criteria. A collaborative ubiquituous space integrates artificial processes to achieve massive computation (simulations, data mining, etc.) and multimodal interfaces including shared visualisation for interactions between humans. In terms of infrastructure, this thesis explains why the Web does not respond suitably to the problem of collective intelligence and why we looked at GRID. In particular, a singular property of GRID is the stateful service that allows to coordinate distributed ressources in a dynamic and secured manner. To show it, instead of developping an analytical model, we have chosen a systemic model. The architecture model Agora proposes to use GRID to deploy collaborative ubiquituous spaces. Validated by a empirical method, this model has been evaluated iteratively based on returns from experimentations with communities of users. According to the progress of our investigations, a GRID architecture coupled concepts linked to multi-agents systems show caracteristics required for scalability with a rationnal use of resources, while offering garanties of sécurity and high availibility.