Planning Under Uncertainty: A Qualitative Approach

Hierarchical planners create descriptions of abstract states and divide their planning task into subproblems for refining these states. In spite of their success in reducing the search space, they classically assume the existence of certain and complete information. In real world planning instances, one has to select among alternative strategies at each abstract state, observing both incomplete knowledge of the attributes that each strategy may pose, and partial ordering of these attributes. In addition, reasoning is defeasible: further information may cause an-state, observing both incomplete knowledge of the attributes that each strategy may pose, and partial ordering of these attributes. In addition, reasoning is defeasible: further information may cause another alternative to be more preferable than what seems optimal at the moment. This work presents a planning framework based on qualitative value decision making formalisms. Sketching the appropriate strategy operator schemata for hierarchical planning, it focuses on aspects of uncertainty handling by combining abilities of constraint programming languages with the introduced concepts of credulous and skeptical conclusions of an issue. The topics of argumentation among the planning agents and common conflicts in the ordering of defaults are discussed by means of a real planning example.