Valérie Camps

Designing Agents’ Behaviors and Interactions within the Framework of ADELFE Methodology

ADELFE is a methodology devoted to software engineering of adaptive multi-agent systems. Adaptive software is used in situations in which the environment is unpredictable or the system is open; in these cases designers cannot implement a global control on the system and cannot list all situations that the system has to be faced with. To solve this problem ADELFE guarantees that the software is developed according to the AMAS (Adaptive Multi-Agent System) theory2. This theory, based on self-organizing multi-agent systems, enables to build systems in which agents only pursue a local goal while trying to keep cooperative relations with other agents embedded in the system. ADELFE is linked with OpenTool, a commercialized graphical tool which supports UML notation. The paper focuses on the extension of OpenTool to take into account AMAS theory in designing agents’ behaviors. The modifications concern static aspects, by adding specific stereotypes, and dynamic aspects, with the automatic transformations from Agent Interaction Protocols into state machines. Then state machines simulate agent behaviors and enable testing and validating them.

Self-Adaptation of a Learnt Behaviour by Detecting and by Managing User's Implicit Contradictions

This paper tackles the issue of ambient systems adaptation to users needs while the environment and users preferences evolve continuously. We propose the adaptive multi-agent system Amadeus whose goal is to learn from users actions and contexts how to perform actions on behalf of the users in similar contexts. However, considering the possible changes of users preferences, a previously learnt behaviour may become misfit. So, Amadeus must be able to observe if its actions on the system are contradicted by the users or not, without requiring any explicit feedback. The aim of this paper is to present the introspection capabilities of Amadeus in order to detect users contradictions and to self-adapt its behaviour at runtime. These mechanisms are then evaluated through a case study.

Apprentissage comportemental en continu de dispositifs ambiants par systèmes multi-agents adaptatifs

Nous proposons detendre la resolution emergente de probleme au cadre des systemes ambiants, au travers de la conception du systeme multi-agent adaptatif Amadeus. Lobjectif dAmadeus est dapprendre le comportement correct a attribuer a un systeme ambiant. Cet apprentissage passe par lobservation des differentes actions des utilisateurs, aen detre capable de progressivement realiser ces actions a leur place. Pour etre reellement adaptatif, Amadeus realise un apprentissage en continu sans connaissance a priori ; il sadapte a lapparition/disparition de dispositifs et est capable de modieer son comportement en cours de fonctionnement en cas devolution du comportement utilisateur. Amadeus a ete evalue par simulation sur plusieurs scenarios, et a montre sa capacite a apprendre localement le comportement correct a attribuer a un dispositif par lobservation des actions utilisateur.

Transformation de modèles d'agents dans la méthode ADELFE : Des stéréotypes de conception à l'implémentation

In this paper, we present a cooperative agent model used by the ADELFE method to design adaptive multi-agent systems. This method is based on object-oriented and agent-oriented process (Rational unified process) and notations (UML and AUML). From the static viewpoint, agents are described as classes, the structure of which is constrained by using stereotypes with a strong semantic on the accessibility to the different modules composing a cooperative agent. From the dynamic viewpoint, communications between agents are specified using AUML protocol diagrams. The functional and structural agent model can therefore easily lead to code generation, and then fill the gap between design and implementation phases, as proposed in the MDA (model driven architecture) framework.