In many key domains like: semantic web (1990->), web services and service oriented software (2000->), peer to peer computing (1999->), grid computing (1998->), ambient intelligence (1998->), self-* systems (2003->), complex systems, etc... it has been proven that agent and multi-agent technologies are vital [Luck, 2005].
Application complexity in terms of number of agents, role and data distribution, non-linearities, dynamics of the environment, creates the need for new models, languages, techniques, tools, ... to design autonomous and robust multi-agent systems. This is the context of the works of SMAC (Cooperative Multi-Agent Systems) team.
Challenge and issues
Our main challenge concerns the design of complex systems for which the global behaviour emerges from the behaviours and interactions of the agents composing the system. Each agent has an individual goal and a behaviour based on a cooperative attitude. Since 1995, we studied an approach for the design of adaptive complex systems based on adaptive multi-agent systems and emergence. For this we elaborated a theory called AMAS (Adaptive Multi-Agent Systems) [Camps, 98][Georgé, 04]. This theory gives local criteria to design agents so as to enable the emergence of an organisation in the system and thus of its global function. The adaptation of the system enables this function to change and is realised by self-organisation of the agents. The cooperative attitude is the engine of this self-organisation since it guides, locally, the agents in its decision making.
Our approach is original compared to other multi-agent design approaches [Weiss, 99], [Wooldridge, 02]. Indeed, we are using emergence as a response to the complexity of current and future applications. As a designer, we have to define: the agents, the environment (if needed) and the means for interaction; the organisation is emerging. In other approaches like those based on the AGR (Agent, Groups, Roles) model [Ferber, 98], the designer has to specify the organisation.
We currently have:
- an emergent functionality software design method, ADELFE [Picard, 04], with operational needs/requirements analysis, analysis and design phases, and freely available on www.irit.fr/ADELFE. This method stands apart from other agent oriented methods [Bergenti, 04][Henderson-Sellers, 05] by its specialisation on adaptive systems. It is also one of the firsts to take into account the environment as well as helping to identify the agents.
- a formal model of the AMAS theory using extended automata products [Capera, 05].
- numerous experimentations: ARCADIA, STAFF, ABROSE, FORSIC, ANTS, SYNAMEC, ETTO, EPE, ...
Research Challenges for 2006-2010
The central issue for the next 4 years remains the design of complex systems using adaptive multi-agent systems. Our work will aim at complex problem solving, modelling for property checking, engineering self-organising systems and extensions towards other domains.
The common ground to all these researches is the AMAS theory and/or the related ADELFE method. All works done by researchers inside SMAC team will fit that framework. The main research topics which will be explored during the next 4 years and further along are:
-* The study of the properties of complex systems.
This topic consists in the study and extension of our understanding of the emergent behaviour of these systems, as well as of the underlying self-organisation mechanisms. The AMAS theory and its reinforcement/extension is at the core of this study.
- The study of the local cooperative behaviour of an agent.
The notion of cooperation which appears both in natural and artificial systems will be explored here to enhance the behaviours of the agents.
- The engineering of self-organising systems.
This topic focuses on the development of tools and methods to enable engineers to design these systems.
These three topics will be applied to several different domains:
- Software engineering and modelling
- User modelling
All these research focuses will be experimented in different application domains, among them:
- Simulation in computational biology (MICROMEGA project)
- Data mining, web services, semantic web and profiling
- Design and autonomous maintenance of ontologies (DYNAMO project)
- Aeronautics (MASCODE and ATOCA projects)
- Ambient systems (Neocoputing and A2C projects)
- Production management.
- AMAS theory (Adaptive Multi-Agent Systems)
- Formalisation of the AMAS theory to show the following property: if all agents currently involved in the computation reach a cooperative state then the system satisfies all environment and/or functional constraints.
- Define constraint satisfaction problem solving algorithms and multi-criteria multi-disciplinary optimisation algorithms by extending the concepts manipulated inside the AMAS theory.
- Engineering self-organising systems
- Complete the ADELFE method (Toolkit to develop software with emergent functionality) with design and deployment phases.
- Provide tools based on the AMAS theory to adjust the parameters during the design of any complex system simulation.
- Engineering cooperative agents
- Design an infrastructure, notably intended for ambient systems, to encapsulate software entities with cooperative behaviours, representations of their environment and of themselves (creation and maintenance of profiles), interaction and inter-comprehension capabilities (creation and maintenance of ontologies).
- [Bergenti, 04] F. Bergenti, M-P. Gleizes, and F. Zambonelli, editors. Methodologies and Software Engineering for Agent Systems. The Agent-Oriented Software Engineering handbook. Kluwer Publishing, 2004.
- [Camps, 98] CAMPS Valérie, GLEIZES Marie-Pierre, GLIZE Pierre - Une théorie des phénomènes globaux fondée sur des interactions locales - Actes des Sixième journées francophones IAD&SMA, Pont-à-Mousson, Editions Hermès, Novembre 1998.
- [Capera, 05] CAPERA Davy, FANCHON Jean, GEORGÉ Jean-Pierre, CAMPS Valérie - A Generic Model Based on Automata for Multi-Agent Systems- The Third European Workshop on Multi-Agent Systems (EUMAS’05), Brussels, Belgium, 7-8 December 2005, KVAB, Brussels, pp. 79-90.
- [Ferber, 98] J. Ferber and O. Gutknecht. Aalaadin: a meta-model for the analysis and design of organizations in multi-agent systems. In Y. Demazeau, editor, Proceedings of the Third International Conference on Multi-Agent Systems, ICMAS’98, pages 128-135, Paris, France, July 1998. IEEE Computer Society.
- [Georgé, 04] Georgé Jean-Pierre, Gleizes Marie-Pierre, Glize Pierre - Conception de systèmes adaptatifs à fonctionnalité émergente : la théorie Amas - Revue d’Intelligence Artificielle, RSTI série RIA, Vol. 17, N. 4, pp. 591-626, 2003.
- [Henderson-Sellers, 05] B. Henderson-Sellers and P. Giorgini, editors. Agent Oriented Methodologies. Idea Group, 2005.
- [Luck, 05] Luck M., Mc Burney P., Shehory O., and Willmott S. - Agent Technology : Computing as Interaction, A Roadmap for Agent Based Computing, ISBN 085432 845 9, 2005.
- [Picard, 04]G. Picard and M.-P. Gleizes. The ADELFE Methodology - Designing Adaptive Cooperative Multi-Agent Systems, chapter 8, pages 157-176. Kluwer Publishing, 2004.
- [Weiss, 99] G. Weiss. Multiagent systems. A modern approach to distributed artificial intelligence. The MIT Press, 1999.
- [Wooldridge, 02] M. Wooldridge. An introduction to multi-agent systems. John Wiley & Sons, 2002.