Hierarchical Task Network (HTN) planning offers powerful expressiveness through task decomposition. Most prior work concentrates on generating plans, while post‑generation optimization, especially plan deordering, has received far less attention in the HTN context. Plan deordering aims to eliminate unnecessary ordering constraints without breaking plan validity. In this paper we adapt two well‑known deordering techniques from classical planning to HTN by extending them to respect hierarchical decomposition constraints. We evaluate the approaches on the IPC 2023 Partial‑Order HTN benchmarks and compare them with Optiplan, an HTN planner that directly produces partially ordered plans. Results show a substantial reduction in the number of ordering constraints for both implementations; the critical path length also decreases, though the improvement is less pronounced. The study demonstrates that deordering can simplify HTN plans while preserving their hierarchical structure, offering greater flexibility for downstream scheduling.
Review: The paper skillfully bridges classical deordering methods to the HTN domain, providing solid empirical evidence of its benefits and opening a practical avenue for HTN plan post‑processing.