Cooperative observers and regulators for discrete-time multiagent systems


Correspondence to: Kristian Hengster-Movric, Automation and Robotics Research Institute, University of Texas at Arlington, 7300 Jack Newell Blvd., S, Ft. Worth, TX 76118, USA.



This paper investigates the design of distributed observers for agents with identical linear discrete-time state-space dynamics networked on a directed graph interaction topology. The digraph is assumed to have fixed topology and contain a spanning tree. Cooperative observer design guaranteeing convergence of the estimates of all agents to their actual states is proposed. The notion of convergence region for distributed observers on graphs is introduced. It is shown that the proposed cooperative observer design has a robustness property. Application of cooperative observers is made to the synchronization problem. A command trajectory generator and pinning control are employed for synchronizing all the agents to a desired trajectory. Complete knowledge about the agent's state is not assumed. A duality principle is shown for observers and state feedback for distributed discrete-time systems on graph topologies. Three different observer/controller architectures are proposed for dynamic output feedback regulator design, and they are shown to guarantee convergence of the estimate to the true state and synchronization of all the agents' states to the command state trajectory. This provides design methods for cooperative regulators based on a separation principle. It is shown that the observer convergence region and feedback control synchronizing region for discrete-time systems are inherently bounded, so that the conditions for observer convergence and state synchronization are stricter than the results for the continuous-time counterparts. This is in part remedied by using weighting of different feedback coupling gains for every agent. Copyright © 2012 John Wiley & Sons, Ltd.