Summary information can be used to find abstract solutions that are guaranteed to succeed no matter how they are refined because the information describes all potential conditions of the underlying decomposition. Thus, some commitments to particular plan choices, whether for a single agent or between agents, can be made based on summary information without worrying that deeper details lurk beneath that will doom the commitments. While HTN planners have used abstract conditions to guide search <e.g.,>sacerdoti:74,tsuneto:98, they rely on a user-defined subset of constraints that can only help detect some potential conflicts. In contrast, summary information can be used to identify all potential conflicts.
Having the formalisms of the previous section, we can now define summary information and describe a method for computing it for non-primitive plans (in Section 3.4). Because there are many detailed definitions and algorithms in this section, we follow the same structure here as in the previous section, where we first give a more informal overview of the key concepts and notation, into which we then subsequently delve more systematically.