DYNAMIC ESTIMATION OF EVIDENCE DISCOUNTING RATES BASED ON INFORMATION CREDIBILITY

Information quality is crucial to any information fusionsystem as combining unreliable or partially credible pieces of informa-tion may lead to erroneous results. In this paper, Dempster-Shafertheory of evidence is being used as a framework for representing andcombining uncertain pieces of information. We propose a method ofdynamic estimation of evidence discounting rates based on the credi-bility of pieces of information. The credibility of a piece of informationCre( I n ) is evaluated through a measure of consensus (corroborationdegree) between a set of belief functions, and this measure serves asa basis for quantifying the credibility of the source (sensor or fusionnode) itself, Cre( S k), used then as a discounting factor for all fur-ther belief functions provided by S k. The process is dynamic in thesense that the credibility of the source is revisited in the light of newincoming piece of information. The method proposed relies on a hy-brid fusion topology in which the sensors are grouped according to thefeature they measure (similar and dissimilar sensors), allowing to se-lect different kinds of measure for estimating the corroboration degrees.Through simulations, we compare (a) the hybrid-combination using thesource credibility and the robust combination rule (RCR-L) account-ing automatically for sensors’s credibility; (b) the hybrid-combination,with different membership degrees and corroboration degrees used toestimate the sources credibility. We show that the new hybrid topol-ogy together with the credibility-based evidence discounting estimationalgorithm provide a faster identification of the observed object