Harnessing the Power of Temporal Abstractions in Model-Based Diagnosis of Dynamic Systems

In this paper we extend our previous work in the area of temporal diagnostic reasoning. Based on a logical framework ex- tended by qualitative temporal constraints we show how to describe behavioral models (both consistency- and abductive-based), discuss how to use abstract observations and show how abstract temporal diagnoses are computed. This yields an expressive framework which makes computation and representation independent of the number of observations and timepoints in a temporal setting. An example of hepatitis diagnosis is used throughout the paper. Previous work in (6) described a formalism for diagnosing dynamic systems by using qualitative temporal relations (a subset o f Allens interval algebra) to describe dynamic behavior. In this paper we extend this diagnostic framework in several ways. First, we introduce two different behavioral models: the ab ductive model is used to generate explanation as covering, while the consis- tency constraint model must be satisfied by a diagnosis and is used to reduce the number of possible diagnoses and/or to strengthen the constraints used in the representation of diagnoses. Secon d, we ex- tend the framework by the concept of a mode constraint graph which allows the representation of knowledge about different behavioral modes. Third, we introduce the concept of abstract observations as a temporal abstraction from time points into time intervals. Finally, we give a declarative definition of an abstract temporal diagnosis and we show on an example, how to compute it.