An Extensive Empirical Evaluation of Inferring Preconditions and Effects of Compound Tasks in Ground HTN Planning Problems
Year: 2025
Type: article
Abstract: HTN planning requires the decomposition of compound tasks into primitive and executable actions. In the currently most frequently used formalism, compound tasks lack explicit preconditions and effects. Those are, however, useful, e.g., for pruning techniques, heuristics, or the comprehension of domains. Previously, we introduced and formalized different kinds of inferred preconditions and effects of compound tasks based on their decomposition methods together with a complexity analysis. In this paper, we present an empirical evaluation of computing these inferred preconditions and effects using the IPC benchmark sets. Specifically, we analyze their frequency of occurrence and compare the performance of an approximation to the exact preconditions and effects. Our goal is to provide a comprehensive overview of the proposed techniques, enabling researchers to determine the extent to which they can be utilized in their given application.
DOI: 10.1613/jair.1.17279
Institutions: Universität Ulm, Australian National University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
Funder: Australian Government
OpenAlex ID: W4408478829