Hierarchical diagnosis guided by observations
Luca Chittaro, Roberto Ranon · Institutional Research Information System (University of Udine) · 2001
We propose a technique to improve the performance of hierarchical model-based diagnosis, based on structural abstraction. Given a hierarchical representation and the set of currently available observations, the technique is able to dynamically derive a tailored hierarchical representation to diagnose the current situation. We implement our strategy as an extension to the well-known Mozetic's approach [Mozetic, 1992], and illustrate the obtained performance improvements. Our approach is more efficient than Mozetic's one when, due to abstraction, fewer observations are available at the coarsest hierarchical levels.