Research map / Probabilistic and scientific modeling / Molecules and proteins
Learning Hierarchical Task Knowledge for Planning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
A review (AAAI 2025 invited-style paper) of learning hierarchical task networks for planning, covering motivation, learning paradigms, HTN learning subproblems, and seven open challenges.
Problem
Acquiring hierarchical task knowledge (HTNs) for planning automatically rather than through costly, error-prone manual construction.
Approach
Reviews HTN learning by specifying its inputs and decomposing it into three subproblems (identifying hierarchical structure, unifying method heads, finding method conditions), then proposes seven research challenges.
Key results
- Identifies three subproblems of HTN learning: identifying hierarchical structure, unifying method heads, and finding method conditions.
- Poses seven challenges for HTN learning, including improving condition finding, acquiring binding constraints, and integration with action learning.
- Notes that no experimental comparisons of analytical versus statistical HTN learning approaches exist to date.
Surprising findings
- Uninformed search through a space of decompositions can increase complex compared to search in the flat space.
Limitations
- Limited to learning from sample solutions; interactive task learning and learning from written instructions are not covered.
- No experimental comparison among HTN learning approaches is provided.
Experimental setup
Builds on: hierarchical task networks macro-operators PDDL
Closest papers by meaning
- Refining HTN Methods via Task Insertion with Preferences (AAAI 2020 · similarity 0.89)
- Learning Neuro-Symbolic Abstractions for Robot Planning and Learning (AAAI 2024 · similarity 0.84)
- Learning Domain-Independent Heuristics for Grounded and Lifted Planning (AAAI 2024 · similarity 0.83)
- SayCanPay: Heuristic Planning with Large Language Models Using Learnable Domain Knowledge (AAAI 2024 · similarity 0.83)
- Hierarchical Planning for Complex Tasks with Knowledge Graph-RAG and Symbolic Verification (ICML 2025 · similarity 0.83)
- Emergent Hierarchical Reasoning in LLMs through Reinforcement Learning (ICLR 2026 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.