Abstract
For solving large complex problems, it is essential to construct problem solving methods which incorporate the human ability to cope with fuzziness. This paper proposes a new method of representing and treating“fuzziness”, by noting that fuzziness can be reduced to a coherent structure to the order constraints involved in the object space (universe of discourse). Seen from this perspective, fuzzy inference can be regarded as a kind of constraint propagation. Furthermore, the new methods of fuzzy set operations and defuzzification introduced are effectively utilized for problem solving. Through them, symbolic (hard) inference is related to fuzzy (soft) inference on the same ground, in such a manner that the qualitative part of fuzziness, its“type”, is set at the initial stage of problem solving, and the quantitative part of fuzziness, its“shape”, is progressively set in accordance with the advances in problem solving, thus resulting in a “distributed network system structure”of fuzziness.
