3.4 Deriving Summary Conditions

Now that we have algorithms that determine interactions of abstract plans based on their summary conditions, we can create an algorithm that derives summary conditions according to their definitions in Section 3.2. Figure 9 shows pseudocode for the algorithm. The method for deriving the summary conditions of a plan $p$ is recursive. First, summary information is derived for each of $p$'s subplans. Then conditions are added based on $p$'s own conditions. Most of the rest of the algorithm derives summary conditions from those of $p$'s subplans. Whether $p$ is $consistent$ depends on the consistency of its subplans and whether its own summary conditions and resource usages are in conflict. The braces '{' '}' used here have slightly different semantics than used before with the brackets. An expression {$x$,$y$} can be interpreted simply as ($x$ or $y$, respectively).

Figure 9: Algorithm for deriving summary information

Definitions and algorithms for temporal relationships such as $always$-$first$ and covers are in Appendix A. When the algorithm adds or copies a condition to a set, only one condition can exist for any literal, so a condition's information may be overwritten if it has the same literal. In all cases, $must$ overwrites $may$; and $first$, $last$, and $always$ overwrite $sometimes$; but, not vice-versa. Further, because it uses recursion, this procedure is assumed to work on plans whose expansion is finite.