Reasoning about Non-Deterministic Observability and Hypothetical Action Occurrences in Multi-Agent Domains

Loc Pham, Tran Cao Son, Enrico Pontelli

Abstract: This paper proposes a method based on high-level action description languages for reasoning about hypothetical action occurrences in a multi-agent environments. In order to accommodate this type of reasoning, one needs to consider non-deterministic observability of action occurrences of agents. The paper presents an extension of the language mA*, called mA*e, to allow for non-deterministic observability and hypothetical actions. The paper defines the semantics of the new language using edge-conditioned update models and discusses properties of the new language, including differences from approaches that use Dynamic Epistemic Logic. The new definitions are illustrated using the well-known story of two stockbrokers from the literature.