IQE: An Incremental Qualitative Envisioner

John W. Collins Qualitative Reasoning Group Beckman Institute University of Illinois 605 N. Mathews St. Urbana, Illinois 61801 email: jcollins@cs .uiuc.edu The finite representation of infinite behaviors provided by total envisionments has been shown to be useful for tasks such as model development, explanation, and monitoring . Unfortunately, generating total envisionments of complex systems can be intractable, or at least excessive for highly-focused tasks . Such shortcomings have encouraged some researchers to favor alternative schemes, such as attainable envisioning and history generation . However, we argue in this paper for an incremental means of envisioning that can realize many of the practical advantages of total envisioning while supporting more-focused search to address issues of tractability and relevance . In this paper, we describe our theory of incremental envisioning and compare it with QPC [3] and QPE [9] . We emphasize how IQE attempts to reason about states at low (i.e ., abstract levels of detail when that is sufficient for the overall task . Such reasoning helps avoid the inefficiency of making state distinctions that are irrelevant for the overall task a problem common among most other current qualitative simulators, including QPC and QPE .