A generic method for computing the response of a numerically known dynamic system to qualitative inputs

This paper proposes a method for computing the response of a numerically-known dynamic system to inputs given in a qualitative form. Such a problem occurs when one wants to modelize a complex device by separating its numerically-known parts from those for which qualitative reasoning techniques are expected to be better suited than classical ones. The presented methodology, which rests upon the well-known theory of optimal control, allows a rigorous treatment of continuous linear time-invariant systems, pointing out the desirable property of "normality" and stating the additional necessary hypotheses to be made upon a signal provided by a qualitative simulation. Its extension to nonlinear systems is then outlined, and various problems it suggests are discussed in conclusion.<>