Terminological knowledge representation systems based on KL-ONE are widely used in artificial intelligence to represent and reason with taxonomies of concept descriptions. A major limitation of terminological approaches, however, is an inability to represent and reason with plans. Plans, complex compositions of actions that achieve given goals, play a central role in many areas that have provided applications for terminological representation systems.
This research concerns the development of knowledge representation and reasoning systems that extend the notions of terminological systems to plans. CLASP provides a knowledge representation language to describe plans by integrating work in automata theory with work in knowledge representation. CLASP was motivated by the needs of a knowledge-based information retrieval system in the domain of telephone switching software (e.g., plans represent telephone features such as call forwarding). In T-REX, plans are represented as networks of actions related by a rich variety of constraints: qualitative temporal constraints, metric temporal constraints, and equality constraints. T-REX also introduces a new view of plan recognition as a process which dynamically partitions a plan library by modalities, e.g., necessary, possible and impossible, while actions are being observed.