CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Distributed fixed-point smoothing estimators in sensor networks connected by a directed graph Authors:  Raquel Caballero-Aguila - Universidad de Jaen (Spain)
Aurora Hermoso-Carazo - Universidad de Granada (Spain)
Josefa Linares-Perez - Universidad de Granada (Spain) [presenting]
Abstract: The focus is on the distributed fixed-point smoothing problem of discrete-time stochastic signals from measurements provided by a sensor network with a topology represented by a directed graph. The signal evolution model is assumed to be unknown and only the mean and covariance functions of the processes involved in the sensor measurement equations are available. The sensor measurements present random failures modelled by random parameter matrices and additive noises, which are assumed to be cross-correlated at the same time and correlated with the signal at the same and subsequent time steps (situation that occurs, for example, in state-space models when the sensor noises at each instant are correlated with the system noise at the previous time). The distributed estimation is performed in two stages; in the first one, every sensor node collects its own measurements and also, according to the network topology, those from the sensors within its neighbourhood in order to generate, through an innovation approach, recursive intermediate least-squares linear fixed-point smoothing estimators. In the second stage, the proposed distributed fixed-point smoothing estimators are generated in each sensor node as the optimal matrix-weighted linear combination of the intermediate estimators provided by its neighbours, according to the mean squared error criterion.