Using partial Diagnoses for Sequential ...

In Model-Based Diagnosis settings, sequential diagnosistechniques can be applied in case thenumber of possible explanations for an unexpectedbehavior of a system is too large to beinspected manually. Such sequential approachesrely on additional measurements to isolate the truecause of the observed problem, usually in an iterativeprocess. In order to speed up the typicallycomputationally costly process of determiningthe next measurement point, we propose thenew concept of “partial” diagnoses, which can beefficiently determined given a small number ofminimal conflicts. We evaluate our proposal bycomparing it with other domain-independent sequentialdiagnosis approaches on different benchmarkproblems. The results show that our newmethod helps to significantly reduce the requiredcomputation times