Model based diagnosis is concerned with finding defects in technical devices. The research in this field has concentrated on two main problems, a modeling problem and an efficiency problem. The modeling problem is to find a useful model to perform model based diagnosis. The efficiency problem is concerned with finding ways to reason about the obtained models in an efficient way. The requirement of putting model based diagnosis into practive produces an additional problem. In general, the diagnosis program to be built will not be able to use unlimited resources. The most important limitation is the costs, for example the cost of time, space, and computational power. But also, the costs of repairing a component in the system and the cost a breakdown of the system will cause. The study focuses on the cost of time. The most important question addresses how a system that provides an answer before a certain deadline can be built. Some literature from the field of real time systems and the field of metareasoning is discussed and preliminary areas about applying the theory to model based diagnosis are given. Anytime algorithms will turn out to be possible solution to the problem of providing an answer within a particular time. In combination with metareasoning techniques, these algorithms could be used to construct a resource limited diagnostic system.
|Publisher||University of Twente|