Distributed filtering in sensor networks with randomly occurring saturations and successive packet dropouts

Abstract

SUMMARYThis paper is concerned with the distributed scriptHMathClass-rel∞ filtering problem for a class of nonlinear systems with randomly occurring sensor saturations (ROSS) and successive packet dropouts in sensor networks. The issue of ROSS is brought up to account for the random nature of sensor saturations in a networked environment of sensors, and accordingly, a novel sensor model is proposed to describe both the ROSS and successive packet dropouts within a unified framework. Two sets of Bernoulli distributed white sequences are introduced to govern the random occurrences of the sensor saturations and successive packet dropouts. Through available output measurements from not only the individual sensor but also its neighboring sensors, a sufficient condition is established for the desired distributed filter to ensure that the filtering dynamics is exponentially mean‐square stable and the prescribed scriptHMathClass-rel∞ performance constraint is satisfied. The solution of the distributed filter gains is characterized by solving an auxiliary convex optimization problem. Finally, a simulation example is provided to show the effectiveness of the proposed filtering scheme. Copyright © 2013 John Wiley & Sons, Ltd.

Estimation, filtering and fusion for networked systems with network-induced phenomena: New progress and prospects

“…Hence, much work has been done on the topics of estimation, fusion, and distributed H ∞ filtering for networked systems over sensor networks in [134]- [137] and the references therein. For example, the estimation and fusion problems have been studied for networked systems over sensor networks in [36], [136], [138], [139] with missing measurements, in [136], [139]- [141] with time-delays, in [142] with sensor saturations, in [143] with signal quantization, and in [144] with channel errors. We will return to the topics of estimation and fusion for complex networks/sensor networks later, and more details concerning the recent advances will be presented in the following section.…”

“…Accordingly, several techniques have been proposed including linear matrix inequality method, recursive/parameter-dependent linear matrix inequality approach, and backward/forward Riccati difference equation method and so on. For example, by using the linear matrix inequality approach, a stochastic sampled-data scheme has been proposed in [154] to address the distributed filtering problem for time-invariant nonlinear systems over sensor networks, a distributed state estimator has been designed in [155] for discrete-time systems over sensor networks with randomly varying nonlinearities and missing measurements, and the distributed filters have been constructed in [25], [142] for nonlinear systems over sensor networks with randomly occurring saturations, quantization errors and successive packet dropouts.…”

Multi-sensor information fusion estimators for stochastic uncertain systems with correlated noises

“…For example, the parameter uncertainties of the systems can be described by multiplicative noises. The systems with missing measurements, quantization effects and randomly occurring sensor saturations can be converted into the model with multiplicative noises [10,11]. Nonlinear polynomial filters are presented for systems with multiplicative noises [12].…”

Robust distributedH∞filtering over an uncertain sensor network with multiple fading measurements and varying sensor delays

“…In networked systems, the limited bandwidth of communication channel leads to several inevitable network‐induced phenomena, which could seriously deteriorate the system performances such as communication delays, missing measurements or packet dropouts, and randomly occurring nonlinearities . The communication delay has been often considered as a deterministic phenomenon that is not realistic.…”