Estimation for Networked Random Sampling Systems With Packet Losses

Abstract

The state estimation problem is investigated in this article for networked random sampling linear stochastic systems. In the system, the system state uniformly updates and the measurement is randomly sampled. Packet losses induced by unreliable networks from a controller to an actuator and from a sensor to an estimator under the TCP protocol are tackled by employing two independent Bernoulli distributed stochastic variables. A state space model (SSM) at successfully received measurement sampling (SRMS) points is developed under the condition of known sampling time. Using an innovation analysis approach, a recursive nonaugmented optimal estimator is proposed in the linear minimum variance (LMV) sense. It can obtain state estimates at state update (SU) points and SRMS points. In addition, for multisensor systems, a centralized fusion estimator by reordering measurement data from sensors and a suboptimal distributed covariance intersection fusion estimator are proposed, respectively. The effectiveness of the proposed algorithms is verified through an example.