A method for dynamic, multi-sensor, evidence combination using fuzzy linguistic terms

There have been few procedures that manage certainty effectively for battlefield decision-making and other real-time, multi-sensor environments. In these environments inferences are rarely certain; lack of certainty can derive from unreliable data, from inappropriate inference rules, and from the indeterminate temporal nature of the data. Thus, an effective certainty management scheme is vital for real-world applications. Such a scheme must support knowledge engineering by simplifying the process of modeling certainty and arriving at results that mimic the way experts reason. This paper proposes a unified methodology to combine certainties associated with evidence and rules for a given proposition, and to propagate this certainty systematically down the (rule-based) decision tree. The relative importance of the propositions as well as the rules themselves have been considered. The proposed methodology supports both numeric certainty values and linguistic variables that model human cognitive models. In addition, the methodology supports “confirmation” and “disconfirmation” constructs which are very useful for knowledge engineering