Detail

Activity Recognition with Intended Actions, Answer Set Programming Approach

We consider an activity recognition problem as follows. We are given a description of the action capabilities of an agent being observed. This description includes a description of the preconditions and effects of atomic actions the agent may execute and a description of activities (sequences of actions). Given this description and a set of propositions about action occurrences and intended actions, called “history”, the problem is to determine what has already happened, what the intentions of the agent are, and what may happen as a result of the agent acting on those intentions. Our formalization is based on a recent approach to reasoning about intended actions in A-like action languages with answer set programming implementations.