Original title: Lose the Order, Keep the Hierarchy: Deordering HTN Plans
Takudzwa Togarepi, Gaspard Quenard, Damien Pellier, Humbert Fiorino · September 2026
Hierarchical Task Network (HTN) planningA planning method where tasks are broken down into subtasks in a hierarchy. is an AI method that solves big goals by breaking them down into smaller steps. Most work focuses on making these plans, not improving them afterward. Often, a plan forces actions into a strict sequence when some could happen in any order. The researchers tackled this by adapting plan deorderingRemoving unnecessary order rules between actions in a plan., a technique that drops unneeded order rules while keeping the plan workable. They took two existing methods from simpler planning systems and updated them to respect the breakdown of high-level tasks. Testing showed that both updated methods caused a substantial reduction in unnecessary ordering rules. They also cut the critical path lengthThe longest sequence of dependent actions in a plan, determining the minimum time to complete it., meaning the shortest total time to finish, though those gains were less pronounced.
Imagine packing for a trip. You might list all tasks like pack clothes, then shoes, then toiletries in a fixed order. Some tasks don't depend on each other, so you could pack shoes before clothes or vice versa, but the fixed order slows you down if something changes. In this paper, the researchers take a plan with strict orders and remove the unnecessary ones, making it more flexible like being able to pack items in any order without messing up the plan.
Making plans more flexible means robots or automated systems can adapt to changes without redoing everything. This could save time and resources in real-world tasks like manufacturing or logistics.
The study is limited to benchmarks from IPC 2023, so results may not generalize to all HTN planning problems. Reductions in critical path length are less pronounced, indicating that not all aspects of plan optimization are equally improved. The adaptation is based on two classical techniques, which might not cover all deordering methods.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Lose the Order, Keep the Hierarchy: Deordering HTN Plans — Takudzwa Togarepi, Gaspard Quenard, Damien Pellier et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.