Abstrakt
Plan recognition is the problem of recognizing a goal task and an agent's plan based on the observed actions. Plan recognition techniques can be employed in multiagent systems, behaviour recognition, computer security, and other fields related to artificial intelligence.
Hierarchical task networks (HTN) describe the decomposition hierarchy of tasks in planning problems. In HTN plan recognition, a prefix of the plan (actions observed so far) is given as an input, and the aim is to find a task that decomposes into a sequence of actions with the given prefix.
In this paper, we show how the performance of parsing-based HTN plan recognition can be improved by restricting possible suffixes of the given prefix based on generalized arc consistency of a corresponding context-free grammar.