Communications in Information and Systems
Volume 15 (2015)
Distributed estimation for parameter in heterogeneous linear time-varying models with observations at network sensors
Pages: 423 – 451
In this paper, a distributed stochastic approximation based estimation algorithm is proposed to estimate the parameter in heterogeneous linear time-varying models associated with sensors from a network. At any time, each agent updates its estimate using the local observations and the information derived from its neighboring agents. The estimates are shown to converge to the one that minimizes the long run average of the square residuals. Switch of the communication graphs is assumed to be deterministic, and the regressors of the linear models are assumed to satisfy some ergodic property, rather than the conditional independence or strict stationarity. Numerical simulations are given to illustrate the obtained theoretic result.