Conditional Planning with External Functions

Abstract

We introduce the logic-based planning language ${\mathcal{K}^c}$ as an extension of $\mathcal{K}$[5]. $\mathcal{K}^c$ has two advantages upon $\mathcal{K}$. First, the introduction of external function calls in the rules of a planning description allows the knowledge engineer to describe certain planning domains, e.g. involving complex action effects, in a more intuitive fashion then is possible in $\mathcal{K}$. Secondly, in contrast to the conformant planning framework $\mathcal{K}$, $\mathcal{K}^c$ is formalized as a conditional planning system, which enables $\mathcal{K}^c$ to solve planning problems that are impossible to express in $\mathcal{K}$, e.g. involving sensing actions. A prototype implementation of conditional planning with $\mathcal{K}^c$ is build on top of the $\textsc{DLV}^\mathcal{K}$ system, and we illustrate its use by some small examples.