Multi-agent Self-Organization and Reorganization to Adapt M2M Infrastructures

Machine-to-Machine (M2M) systems are based on networks of sensors and actuators immersed in the physical world, like in Smart Cities. To manage such networks, classical so-called vertical approaches (where one stakeholder is involved from services to sensors) are very expensive. Thus, a growing trend emerges for horizontal architectures, where several stakeholders collaborate within the same infrastructure. Moreover, the growing number of connected devices raises the key issue of properly controlling and adapting capabilities so that human intervention is minimized. Based on a Multi-Agent Oriented Programming approach, the Adaptive Governance MAS for M2M systems (AGaMeMnon) system installs such capabilities by combining self-organization and reorganization mechanisms. This will be the focus of this paper. AGaMeMnon has been implemented and linked to the SensCity project middleware and evaluated experimentally by the means of simulations.