Automatically Learning HTN Methods from Landmarks (2404.06325v1)
Abstract: Hierarchical Task Network (HTN) planning usually requires a domain engineer to provide manual input about how to decompose a planning problem. Even HTN-MAKER, a well-known method-learning algorithm, requires a domain engineer to annotate the tasks with information about what to learn. We introduce CURRICULAMA, an HTN method learning algorithm that completely automates the learning process. It uses landmark analysis to compose annotated tasks and leverages curriculum learning to order the learning of methods from simpler to more complex. This eliminates the need for manual input, resolving a core issue with HTN-MAKER. We prove CURRICULAMA's soundness, and show experimentally that it has a substantially similar convergence rate in learning a complete set of methods to HTN-MAKER.
- 2009. Curriculum learning. In ICML, 41–48.
- 2021. Learning Hierarchical Task Networks with Preferences from Unannotated Demonstrations. In CoRL, 1572–1581.
- 2005. Learning teleoreactive logic programs from problem solving. In ILP, 51–68.
- 2020. Learning hierarchical task networks with landmarks and numeric fluents by combining symbolic and numeric regression. In ACS.
- 2018. Learning combinatory categorial grammars for plan recognition. In AAAI, 3007–3014.
- Geib, C. W. 2009. Delaying commitment in plan recognition using combinatory categorial grammars. In IJCAI, 1702–1707.
- 2016. Automated planning and acting. Cambridge University Press.
- 2016. Word2HTN: Learning task hierarchies using statistical semantics and goal reasoning. In IJCAI Workshop on Goal Reasoning.
- 2016. Autonomously constructing hierarchical task networks for planning and human-robot collaboration. In ICRA, 5469–5476.
- 2004. Ordered landmarks in planning. JAIR 22:215–278.
- 2008. HTN-MAKER: Learning HTNs with minimal additional knowledge engineering required. In AAAI, 950–956.
- 2016. Learning hierarchical task models from input traces. Computational Intelligence 32(1):3–48.
- 2023. Leveraging Demonstrations for Learning the Structure and Parameters of Hierarchical Task Networks. In FLAIRS, volume 36.
- 2005. Learning approximate preconditions for methods in hierarchical plans. In ICML, 337–344.
- Kantharaju, P. 2021. Learning Decomposition Models for Hierarchical Planning and Plan Recognition. Ph.D. Dissertation, Drexel University, Philadelphia, Pennsylvania, USA.
- 2010. Sound and complete landmarks for and/or graphs. In ECAI, 335–340.
- 2006. Learning recursive control programs from problem solving. Journal of Machine Learning Research 7(3).
- Langley, P. 2022. Learning hierarchical problem networks for knowledge-based planning. In ILP.
- 2009. Learning probabilistic hierarchical task networks to capture user preferences. In IJCAI, 1754–1759.
- 2016. Constructing hierarchical task models using invariance analysis. In ECAI. 1274–1282.
- 1999. Shop: Simple hierarchical ordered planner. In IJCAI, 968–973.
- 2006. Learning hierarchical task networks by observation. In ICML, 665–672.
- 2010. The LAMA planner: Guiding cost-based anytime planning with landmarks. JAIR 39:127–177.
- 2017. Learning HTN domains using process mining and data mining techniques. In ICAPS Workshop on Generalized Planning.
- 2022. PDDL generators. https://doi.org/10.5281/zenodo.6382173.
- Shrobe, H. E. 2021. Hierarchical problem networks for knowledge-based planning. In ACS.
- 2005. A domain-independent system for case-based task decomposition without domain theories. In AAAI, 234–240.
- 2007. Learning recursive htn-method structures for planning. In ICAPS Workshop on AI Planning and Learning.
- 2009. Learning HTN method preconditions and action models from partial observations. In IJCAI, 1804–1809.
- 2014. Learning hierarchical task network domains from partially observed plan traces. Artificial Intelligence 212:134–157.