A Multi-Agent Approach for Complex System Design

Multi-agent systems that arose from research in Distributed Artificial Intelligence are now considered as a new paradigm to design and model complex systems. The design of complex systems seems to be more intuitive with cognitive agents because the designer works at a high level of abstraction where the agent is an important granularity entity with aspects of calculation, reasoning and control which make it quasi autonomous. After a brief overview of the required notions to understand the multi-agent systems' domain, we describe the JACK agent-oriented environment on top of which we are developing the SCALA environment that provides a multiagent-based methodology and tool for the design of complex systems. In SCALA, the global behaviour (i.e. its goal) of the system is modelled through a functional approach based on the definition of a graph of dependencies between the basic behaviours (i.e. tasks to accomplish to achieve a goal). The definition of this graph provides the necessary knowledge to manage cooperation between the agents and to plan reactively their activities when new events occur and when they have to reorganise themselves.