Adding temporal reasoning to expert-system-building environments

Abstract

The authors discuss the incorporation of temporal-reasoning capability in an expert-system-shell environment, and the implication of this approach by storing attribute values with time tags. In essence, this adds an extra dimension (time) to the frame database but, due to the implementation method chosen, the storage space increase tends to be small. A time-tag capability for point times and intervals was implemented to make temporal reasoning as general as possible. Since values are often stored at point times or unconnected intervals, the meaning of the values at nonspecified times must be covered. Four different methods to cover those times were implemented, and knowledge engineers can select the one appropriate for each attribute. To focus the research, the authors applied temporal reasoning to a monitoring-and-diagnosis application in which time-dependent data had been difficult to handle. They developed Wheels, an expert system that diagnoses problems with the Hubble space telescope's reaction wheels.