Valuation Networks for Implementing Fusion Algorithms

PDF

Authors
  1. Jousselme, A-L.
  2. Bossé, É.
Corporate Authors
Defence R&D Canada - Valcartier, Valcartier QUE (CAN)
Abstract
In this report, we explore the ability of valuation networks to implement information fusion algorithms for the particular problem of target identification. Valuation networks are the graphical representation of valuation algebras, an abstract framework for implementing local computation in different kinds of graphs. Then, the benefit sought is twofold since, beside allowing a general representation of uncertainty so that different formalisms can be considered, they provide an efficient computational scheme for the updating and aggregation of information in different kinds of graphs. Indeed, in valuation algebras, at least probabilities, belief functions, possibilities can be used for reasoning. Moreover, valuation networks, which are the graphical structure associated to valuation algebras, implement in a generalized approach graphs such as Bayesian networks or diagnosis trees for hierarchical evidence. After providing some theoretical background on both representation of uncertainty (mainly probability theory and evidence theory), we apply valuation algebras to a problem of target identification. We compare in this framework these two kinds of uncertainty representation and their ability to identify targets.

Il y a un résumé en français ici.

Report Number
DRDC-VALCARTIER-TR-2006-786 — Technical Report
Date of publication
01 Jul 2008
Number of Pages
104
DSTKIM No
CA031281
CANDIS No
530183
Format(s):
CD ROM

Permanent link

Document 1 of 1

Date modified: