Generalized plan recognition

Abstract

This paper outlines a new theory of plan recognition that is significantly more powerful than previous approaches. Concurrent actions, shared steps between actions, and disjunctive information are all handled. The theory allows one to draw conclusions based on the class of possible plans being performed, rather than having to prematurely commit to a single interpretation. The theory employs circumscription to transform a first-order theory of action into an action taxonomy, which can be used to logically deduce the complex action(s) an agent is performing.