KNOWLEDGE-LEVEL ANALYSIS FOR ELICITING COMPOSABLE SCHEDULING KNOWLEDGE

This paper shows how composable scheduling knowledge can be elicited f rom existing scheduling systems at the knowledge level, which is a mor e abstract level than that of actual system implementation. We begin b y defining the job assignment task as a class of target scheduling pro blems. As a result of a preliminary analysis of four scheduling expert systems described in the literature, three problem-solving patterns a re obtained as abstract templates for component elicitation. We then d escribe how we investigated another two scheduling systems in collabor ation with the engineers actually involved in the development, so that the components obtained can be validated not only by knowledge engine ers but also by the system developers. Consequently, a total of eleven problem-solving components are identified, and their reusability is d emonstrated.