Consensus in Networks: Coordination and Control of CyberPhysical Systems, from Unmanned Vehicles to Energy-Efficient Buildings

Summary form only given. In this talk we discuss the consensus variable approach to the coordination and control of cyberphysical systems (those with a tight integration of physical dynamics, sensors and actuators, and computing infrastructure). We begin with an overview of motivating problems and a summary of key results related to the consensus (or agreement) paradigm. We illustrate the application of this idea to several problems, including simulation of swarms, experimental demonstration of formation control of mote-based robots, and experimental demonstration of robotic autonomous mobile radio nodes for wireless tethering between a base station and a leader in a tunnel exploration scenario. We note that a consensus protocol can be represented as a graph with (static) weighted edges and nodes that are integrators. Generalizing this idea, we next present what we call dynamic consensus networks. Such networks are graphs whose nodes are integrators and whose edges are real rational functions representing dynamical systems that couple the nodes. We show that the modeling of thermal processes in buildings motivates such a system and from this motivation we generalize the notions of interconnection matrices and Laplacians to the case of graphs with integrating nodes and dynamic edges. We give conditions under which such graphs admit consensus, meaning that in the steady-state the node variables converge to a common value. Finally, we consider the collective description and properties of the interconnection of one dynamic graph (the plant) with another (the controller). We conclude with a discussion of research questions related to these ideas and to their application to energy-efficient control of buildings and other systems, such as the power grid.