A production rule-based framework for causal and epistemic reasoning

Abstract

Action theories are an important field of knowledge representation for reasoning about change and causality in dynamic domains. In practical implementations agents often have incomplete knowledge about the environment and need to acquire information at runtime through sensing, the basic ontology of action theories needs to be extended with epistemic notions. This paper presents a production system that can perform online causal, temporal and epistemic reasoning based on the Event Calculus and on an epistemic extension of the latter. The framework implements the declarative semantics of the underlying logic theories in a forward-chaining rule-based system. This way, it combines the capacity of highly expressive formalisms to represent a multitude of commonsense phenomena with the efficiency of rule-based reasoning systems, which typically lack real semantics and high-level structures. © 2012 Springer-Verlag.