...plan
For a more formal development of the refinement search semantics of partial plans, we refer the reader to the work of Kambhampati, Knoblock, and Yang [25].
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...tree
In practice, this limit is actually a bound placed on the number of steps contained in the plan.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...failure
DERSNLP+EBL's EBL component explains only analytical failures. Depth limit failures are ignored. This means that the failure explanations that are formed are not sound in the case of a depth limit failure, and that the retriever may reject a case when it is applicable. Rejecting an applicable case may lead to the storage of duplicate cases and a larger library size. However, our empirical work has not shown this to be of practical importance for reasons outlined in Section 3.2.2.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...airport
For a more comprehensive evaluation over an unbiased problem set see Section 4.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...plan
An analogous decrease in plan quality occurs in state-space plan reuse, when sequencing macro-operators results state loops [29].
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...plan
Consider, for example, a domain in which the plane may transport two packages in one trip, or not, depending on its capacity.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...problems
There is more opportunity for interaction in larger problems. For example, a 6-goal problem could contain 6 goals that mutually interact, whereas a 5-goal problem has a maximum of 5 interacting goals.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
...sec
DERSNLP+EBL in from-scratch mode used a best-first strategy. In replay, this best-first strategy is biased so that the subtree under the replayed path is explored first, before the siblings of this path.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

11/5/1997