Analysis, Design and Validation of Multi-Agent Systems (MAS)
Keywords: Agents, Coordination, Formal specification, Distributed Observation, Distributed Planning, Petri Nets, Interaction, Modal Logics, Multi-Agent Systems, True Concurrency Semantics , Verification.
Our approach for MAS Design
We are interested in MAS as a cognitive approach that helps developing distributed and cooperative systems based on intelligent agents. Our goal is the formal specification and verification of such systems. MAS, from our point of view, have three main features:
Concurrency: allows agents to distribute tasks and to achieve their goals, both in cooperative and competitive cases.Interaction: allows agents, viewed as intelligent components, to exchange their knowledge (or beliefs), to share goals, and to cooperate in order to achieve a common goal. In the case of competitive agents, interaction enables to coordinate activities, to negotiate or to reach a consensus between agents.Cognition: refers to intelligent skills of agents (autonomy, learning, reasoning, etc.). Cognition endows agents with reflexive behaviour, i.e. agents act and control autonomously their actions.
In our research, we are interested in MAS where intelligent agents interact and are physically distributed (e.g. web, network, etc.. applications). Consequently, the models, algorithms and protocols we developed take into account the following aspects:
- The distribution as an intrinsic property of our models.
- Formal specification to make possible our models’ validation.
- Most of our theoretical models have been implemented in the context of industrial projects.
I. Cognitive models for several kinds of agents (autonomous, rational, cooperative or competitive agents)
- BDI-agents’ learning
- Coordination of competitive agents based on coalition formation
- Rationality in MAS
- Models of negotiation based on multi-criteria decision-making
II. Distributed models of coordination for cognitive agents
- Coordination based on distributed planning
- Cooperation protocols
- Distributed and reactive planning
III. Interaction in MAS
- Engineering of interaction protocols
- Learning interaction in MAS (for load-balancing)
- Intentional semantics for agents interaction
- Cognitive modelling of interaction
IV. MAS engineering
- Operational semantics and programming languages for cognitive, mobile and intelligent agents.
- Platform for mobile agents design (MASIF compliant)
- Design of MAS based on agents and MAS frameworks.
- Web services (planning and coordination)