State-Based Regression with Sensing and Knowledge

Richard B. Scherl, Tran Cao Son, Chitta Baral

Abstract: This paper develops a state-based regression method for planning domains with sensing operators and a representation of the knowledge of the planning agent. The language includes primitive actions, sensing actions, and conditional plans. We prove the soundness and completeness of the regression formulation with respect to the definition of progression and the semantics of a propositional modal logic of knowledge. It is our expectation that this work will serve as the foundation for the extension of recently successful work on state-based regression planning to include sensing and knowledge as well.