Decentralized parameter estimation with fuzzy information

Abstract

The use of fuzzy sets in representing uncertainty in signal detection and estimation problems has been shown to complement conventional approaches using probabilistic modeling. We concentrate on an approach where the sample information available from the physical phenomenon of interest is assumed to be vague. We study and analyze a parameter estimation scheme in a decentralized system when the data available at each sensor is vague. The vagueness of the data is represented by means of 'fuzzy events' defined over the real line. The optimum global estimator (in a minimum mean square error sense) is obtained, and the corresponding optimum partitioning of the fuzzy information space is presented. We also discuss a suboptimum data partitioning method using the Fisher information measure.