Fast and Complete Symbolic Plan Recognition: Allowing for Duration, Interleaved Execution, and Lossy Observations

It is important for agents to model other agents’ unobserved plans and goals, based on their observable actions. This process of modeling others based on observations is known as plan-recognition. Plan recognition has been studied for many years. It often takes the form of matching observations of an agent’s actions to a plan-library, a model of possible plans selected by the agent. However, there are several open key challenges in modern plan recognition: (i) handling lossy observations (where an observation or a component of an observation is intermittently lost); (ii) dealing with plan execution duration constraints; and (iii) interleaved plans (where an agent interrupts a plan for another, only to return to the first later). In this paper, we present efficient algorithms that address these challenges, in the context of symbolic plan recognition. The algorithms allow (i) efficient matching of (possibly lossy) observations to a plan library; (ii) efficient computation of all recognition hypotheses consistent with the observations, subject to interleaving and duration constraints.