A multi-level generic multi-agent architecture for supervision of collective cyber-physical systems

Cyber-physical systems like networked embedded systems or ambient sensors networks are called large-scale artificial complex systems. They are difficult to supervise because of their numerous components in interaction relying upon physical devices, their extension and their openness. A multi-scale organisation can be a solution to make them more accessible. Multi-agent systems (MAS) are well suited for modeling large complex systems, as multi-agent organisational capabilities allow introducing multi-level observation. This paper proposes a multi-level multi-agent mechanism based on recursion for supervision and observation of large-scale artificial complex systems. It is developed as a free framework that is a decentralised application allowing truly physically decentralised MAS to communicate with abstract multi-agent layers. As an illustration, the framework is applied to a wireless sensor network supervision system.