We consider the problem of fusion of multiple information sources, par
ticularly in environments when the sensor observations are imprecise.
The concept of a combinability relationship is introduced to enable th
e inclusion in the fusion process of information about the appropriate
ness of fusing different elements from the observation space. This ide
a allows for the use of an expert knowledge base, containing informati
on about the domain of the particular problem, in the fusion process a
nd leads to a more intelligent aggregation. We show that if we use a c
ombinability relationship that only allows fusion of identical element
s then the only idempotent fusion of any collection of fuzzy observati
ons is their intersection. Using the idea of the fuzzy measure we cons
idered situations in which we allow a partial collection of the observ
ations determine the fused value. (C) 1997 John Wiley & Sons, Inc.