Multiagent and network systems
Multiagent and network systems are one of the most effective and modern occurrences of 'complex systems'. This class of systems result from the interconnection of a possibly very large number of elementary subsystems, with the typically complex pattern of local interactions modeled as a graph. Applications include: dynamic network flows (e.g., of vehicular traffic or of power); interactions in social and economic networks; epidemic diffusion models; mobile agents networks (both natural and man-made); and distributed and cooperative algorithms in sensor networks. The analysis and synthesis of such network systems poses fundamental mathematical challenges. From the analysis viewpoint, a crucial point is concerned with understanding the relationship among the interconnection graph topology, the local interaction laws, and the emerging global system behavior (e.g., synchronization, correlation decay, propagation of instabilities through cascading mechanisms, fragility and resilience). As far as control synthesis is concerned, the complexity of such network systems makes the classical centralized control paradigm (whereby information is collected in a single point in order to compute the control action) unfeasible because of its lack of scalability and the inherent assumption that perfect information is available in a single point. Modern decentralized and distributed control approaches are in need, relying on the use of local information and cooperation strategies, as is evaluating the fundamental limitations implied by the communication constraints among the different control units.