Evidence combination using fuzzy linguistic terms in a dynamic, multisensor environment

There have been few procedures that effectively manage certainty for real-time multi-sensor environments such as battlefield decision making. In these environments inferences are rarely certain due to: unreliable data, inappropriate inference rules, and indeterminate temporal nature of data. Thus, there is a vital need for an effective certainty management scheme for these real-world applications. This paper presents extensions to our earlier paper (1989) and presents a formalism for computing membership functions as a mechanism for combining evidence. It proposes a "unified" methodology that combines certainties associated with evidence and rules for a given proposition, and systematically propagates these certainties down the (rule-based) decision tree. The methodology takes into account the relative importance of the propositions as well as the rules. The proposed methodology supports both numeric certainty values and linguistic variables that model human cognition. In addition, the methodology supports "confirmation" and "disconfirmation" constructs that are very useful for knowledge engineering.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>