We’ve examined various thinkable ways in which plan formation, execution, monitoring and replanning could take advantage from Truth Maintenance techniques. The creation of a plan is based on a set of assumptions about the external environment and task goals. The inconsistency of these sets or the falsehood of some among their elements can cause failures both in plan generation or execution. On failure it becomes important to understand which assumptions were responsible for the flaw in order to get a more updated knowledge base or a consistent goal. Consistency restoration, multisensor information integration and updating knowledge bases, all need knowledge revision techniques. In this paper we report some possible schemes for the interleaving between the activities of an Assumption Based Truth Maintenance System (ATMS) and one or more (STRIPS-like) Planners. We distinguish four cases of useful adoption of an ATMS: to check and eventually resolve inconsistencies during a single planner planning, to update the knowledge base of a single planner from multisensor information integration or on failure during the execution, case 1 but in a multi-planner architecture, case 2 but in a multi-planner architecture.
Integrating Planning and Knowledge Revision: preliminary report / Dragoni, Aldo Franco. - STAMPA. - Series Workshops in Computing:(1993), pp. 323-327. (Intervento presentato al convegno AI and Cognitive Science ’92 tenutosi a University of Limerick nel 10–11 September 1992) [10.1007/978-1-4471-3207-3_31].