Evidential Reasoning: An Implementation for Multisensor Integration

Abstract

One common feature of most knowledge-based expert systems is that they must draw conclusions on the basis of evidential information, Yet there is very little agreement on how this should be done. In this paper, the authors present their view of this problem and its solution for multisensor integration. They begin by characterizing evidence as information that is uncertain, incomplete, and sometimes inaccurate. On the basis of this characterization, they conclude that evidential reasoning requires both a method for pooling multiple bodies of evidence to arrive at a consensus and some means of drawing the appropriate conclusions from that consensus. They contrast their approach, which is based on a relatively new mathematical theory of evidence, with those that have their basis in Bayesian probability models. They believe that their method has significant advantages over Bayesian methods in its ability to represent and reason from bounded ignorance. They describe an implementation of these techniques by means of two kinds of memory: long-term memory and short-term memory. This implementation provides for automated reasoning from evidential information at multiple levels of abstraction over time and space.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 55 citations exist upstream.