A Proper Ontology for Reasoning about Knowledge and Planning

Research on the knowledge preconditions problems for actions and plans has sought to answer the following questions: (1) When does an agent know enough to perform an action? (2) When can an agent execute a multi-agent plan? It has been assumed that the choice of temporal ontology is not crucial. This paper shows that this assumption is wrong and that it is very diicult to develop within existing ontologies theories that can answer both questions (1) and (2). A theory of linear time does not support a solution to the knowledge preconditions problem for action sequences. A theory of branching time solves this problem, but does not support a solution to the knowledge preconditions problem for multi-agent plan sequences. Linear time supports prediction, but does not support hypothetical reasoning; branching time supports hypothetical reasoning, but does not support prediction. Since both prediction and hypothetical reasoning are essential components of the solution to the knowledge preconditions problems, no comprehensive solution has yet been proposed. To solve this problem, we introduce a new temporal ontology, based on the concept of an occurrence that is real relative to a particular action. We show that this ontology supports both hypothetical reasoning and prediction. Using this ontology, we deene the predicates needed for the proper axiomatization for both knowledge preconditions problems.