CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Signal filtering over sensor networks with random delays and loss compensations: Distributed and centralized framework Authors:  Raquel Caballero-Aguila - Universidad de Jaen (Spain) [presenting]
Aurora Hermoso-Carazo - Universidad de Granada (Spain)
Josefa Linares-Perez - Universidad de Granada (Spain)
Abstract: Recently, the estimation problem in sensor networks has received considerable attention due to its great number of application fields. In a network environment, the inaccuracy of the measurement devices and the limitations of the communication resources usually cause random phenomena that impair the performance of the estimators; consequently, many interesting challenges arise in the research of the fusion estimation problem over sensor networks featuring random imperfections. When the measurements are subject to transmission losses, a hot topic is how to compensate them; the more common compensation mechanisms consist of processing either nothing (non-compensation) or the latest successfully transmitted measurement. The aim is to study the distributed and centralized fusion filtering problems from measured outputs with uncertainties modelled by random matrices and correlated noises. At each sampling time, every sensor output is sent just once to a local processor and, due to random transmission failures, one-step delays and, consequently, packet losses may occur; it is assumed that the losses are non-consecutive and compensated with the last measurement received at the local processor. Modelling the random delays and losses by different sequences of Bernoulli random variables, the estimation is addressed by a covariance-based approach. In a simulation example, the results are compared with those obtained when the packet losses are not compensated.