CMStatistics 2016: Start Registration
View Submission - CMStatistics
B1476
Title: Distributed filtering in connected sensor networks from measurements with random matrices and transmission failures Authors:  Josefa Linares-Perez - Universidad de Granada (Spain) [presenting]
Raquel Caballero-Aguila - Universidad de Jaen (Spain)
Aurora Hermoso-Carazo - Universidad de Granada (Spain)
Abstract: Recently, the research on fusion estimation problems in networked systems is especially being focused on sensor networks with a given topology, to consider the possibility of coordination among the sensors to perform global tasks by the exchange of information with neighboring sensor nodes. Also, the use of random matrices is gaining a great interest as they provide a unified framework to model different random network-induced phenomena, such as stochastic sensor gain degradation, measurement multiplicative noises or missing measurements. We consider the estimation problem of discrete-time signals from measurements perturbed by random parameter matrices and white additive noises. These measurements are obtained by multiple sensors, located at the nodes of a network, which are connected according to a specified network topology. This is modeled by a directed graph and both one-step delays and packet dropouts are assumed to occur randomly during the data transmission through the network communication channels. We address the so-called distributed estimation problem, in which the signal is estimated at each node using its own measurements and those from its neighbors. Using covariance information, without requiring the evolution model generating the signal process, a recursive algorithm for the distributed least-squares linear filter is derived by an innovation approach.