Plenary talk III

The rapid pace of innovation in the areas of control, computation, and communication is leading the way for the class of networked multi-agent systems that are characterized by their complex interconnections, diversity of components, and the interactions with the physical world and possibly humans. These systems offer a vision of increased automation and benefit for society from environmental, economic, and social perspectives. Examples of such systems include automated transportation networks, distributed power generation (“smart-grids”), groups of autonomous vehicles, or interacting groups of robots, to name just a few. These growing application areas with their challenging performance specifications do require a solid theoretical foundation in order to understand, influence and design the dynamical behavior of these complex systems. Te constituent parts of interconnected systems are the individual dynamical subsystems, the interconnection topology between the subsystems, and the individual links used to realize the interconnections. All three parts can be of different complexity yielding three dimensions of complexity in interconnected systems that are often summarized in the so-called complexity cube with its dimensions system complexity, topological complexity and link complexity. System complexity is the classical focus of systems and control theory and refers to the complexity of the individual subsystems. Topological complexity is commonly described using interconnection graphs. Finally the link complexity takes into account that individual communication links are imperfect, e.g., because of transmission delays or effect of packed switched networks. It turns out that the three dimensions of complexity cannot be addressed independently. Elevated complexity along one dimension usually yields constraints along the other dimensions. In this talk, the tradeoff between system complexity, topological complexity and link complexity in networked multi-agent systems will be explored. We will give an overview over the state of the art and will show that many of today's challenges in applications can be addressed with the methods developed over the past few years.