2013/01/21 by Soummya Kar, José M. F. Moura, Kar, Soummya +1
Physics and Astronomy · Computer Science · #Opinion Dynamics and Social Influence #Distributed Control Multi-Agent Systems #Distributed Sensor Networks and Detection Algorithms
paper · pdf · doi:10.48550/arxiv.1301.5047
The paper studies the problem of distributed parameter estimation in\nmulti-agent networks with exponential family observation statistics. A\ncertainty-equivalence type distributed estimator of the consensus + innovations\nform is proposed in which, at each each observation sampling epoch agents\nupdate their local parameter estimates by appropriately combining the data\nreceived from their neighbors and the locally sensed new information\n(innovation). Under global observability of the networked sensing model, i.e.,\nthe ability to distinguish between different instances of the parameter value\nbased on the joint observation statistics, and mean connectivity of the\ninter-agent communication network, the proposed estimator is shown to yield\nconsistent parameter estimates at each network agent. Further, it is shown that\nthe distributed estimator is asymptotically efficient, in that, the asymptotic\ncovariances of the agent estimates coincide with that of the optimal\ncentralized estimator, i.e., the inverse of the centralized Fisher information\nrate. From a technical viewpoint, the proposed distributed estimator leads to\nnon-Markovian mixed timescale stochastic recursions and the analytical methods\ndeveloped in the paper contribute to the general theory of distributed\nstochastic approximation.\n