Exploiting Temporal Knowledge to Organize Constraints.

This paper examines the role of temporal knowledge in regard to the representation and utilization of constraints within time dependent problem domains. The author first considers the representation of constraints whose interpretations may vary in different temporal regions of the solution space. A representation that organizes constraint variants via the temporal relationships among them is presented and seen to support a simple mechanism for determining the applicable variant at any point in time. He then turns his attention to the management of temporal constraints that are dynamically imposed as various commitments are made by the reasoning system e.g. as resources are allocated to activities in a plan. Constraint propagation techniques which collectively insure consistency in the hypotheses under development are presented and discussed. These techniques are driven by the temporal relationships present in the domain model. This work is motivated by ongoing research with ISIS, an intelligent scheduling and information system currently being applied to the problem of scheduling job shops, and examples throughout the paper are drawn from this domain. Author