Convergence of a Distributed Least Squares

In this paper, we consider a least-squares (LS)-based distributed algorithm build on sensor network to estimate an unknown parameter vector of dynamical system, where each in the has partial information only but is allowed communicate with its neighbors. Our main task generalize well-known theoretical results traditional LS current case by establishing both upper bound accumulated regrets adaptive predictor and convergence estimator, following key features compared existing literature estimation: Firstly, our theory does not need previously imposed independence, stationarity or Gaussian property system signals, hence applicable stochastic systems feedback. Secondly, cooperative excitation condition introduced used paper for weakest possible one, which shows that even if any individual cannot LS, whole can still fulfill estimation LS.