Title: Sampling design schemes for distributed parameter estimation networks

Authors: Ming-Fong Hsu; Tsang-Yi Wang; Chao-Tang Yu; Tsai-Cheng Wu

Addresses: Institute of Communications Engineering, National Sun Yat-sen University, 80424, Taiwan ' Institute of Communications Engineering, National Sun Yat-sen University, 80424, Taiwan ' Department of Electronic Engineering, Southern Taiwan University of Science and Technology, 71005, Taiwan ' Vivotek Inc., 23553, Taiwan

Abstract: This paper proposes two sampling schemes for distributed parameter estimation networks. The estimation network comprises a number of remotely located sensors which process the observed signal locally and then convey the processed data to a data fusion centre to make the final estimate of the parameter of interest. In the first sampling scheme, all sensors utilise same number of sampling points, and the distribution of sampling points at each sensor is determined to maximise the estimation accuracy. By contrast, in the second one, the sensors are assigned a different number of sampling points, and the objective is to determine the sampling point assignment which maximises the overall estimation performance given a constraint on the total number of sampling points available. The two sampling design problems are solved by minimising the Fisher information loss. The validity of the proposed sampling design schemes is demonstrated by means of two numerical examples.

Keywords: distributed estimation; sampling design; optimal allocation; sensor networks; maximum likelihood estimation; person by person optimisation; parameter estimation networks; data fusion; Fisher information loss.

DOI: 10.1504/IJAHUC.2016.076595

International Journal of Ad Hoc and Ubiquitous Computing, 2016 Vol.22 No.1, pp.62 - 69

Received: 30 Jun 2014
Accepted: 17 Nov 2014

Published online: 17 May 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article