Distributed State Estimation for Uncertain Linear Systems With a Recursive Architecture

This paper investigates distributed state estimation for a class of discrete-time linear systems, with dynamics subject to parameter perturbation and Gaussian noise simultaneously. A sensor network is adopted, where each sensor estimates the system state by fusing its own measurements and its neighbors' information. A novel distributed state estimator is proposed, where the estimator gains are designed by using a state-error augmented technique. To check whether the estimator is effective on the infinite horizon, a very simple criterion only on system parameters is developed. Moreover, a relation between the estimation error covariance and the system state covariance is revealed, which shows that a performance index similar to the signal-to-noise ratio is ensured by the designed estimator. Compared with the existing methods in the literature, such as augmented methods and linear matrix inequalities-based methods, the estimation method proposed in this paper is much more computationally efficient. Finally, several numerical simulation results are demonstrated to illustrate the effectiveness of the new estimator.