HTN-Maker: Learning HTNs with Minimal Additional Knowledge Engineering Required

This web site maintains the source code and domain files for HTN-Maker, HTN-Maker^ND, and the (Q-Maker, Q-Reinforce,Q-SHOP) variant (see descriptions below). HTN-Maker was developed under funding by National Science Foundation under Grant No. NSF 0642882. For a broad view of other publications about HTN learning and activities performed under that effort please visit this web site.

HTN-Maker is an algorithm for learning hierarchical planning knowledge in the form of decomposition methods for Hierarchical Task Networks (HTNs). HTN-Maker takes as input the initial states from a set of classical planning problems in a planning domain and solutions to those problems, as well as a set of semantically-annotated tasks to be accomplished. The algorithm analyzes this semantic information in order to determine which portions of the input plans accomplish a particular task and constructs HTN methods based on those analyses.
Main reference: Hogg, Chad, Munoz-Avila, Hector, and Ugur Kuter. (2008) HTN-Maker: Learning HTNs with Minimal Additional Knowledge Engineering Required. Proceedings of the Twenty-Third AAAI Conference on Artificial Intelligence (AAAI-08). AAAI Press.
PDF
HTN-Maker^ND extends HTN-Maker by allowing the learning of methods in nondeterministic planning domains, in which actions may have multiple possible outcomes. The resulting methods can be used by a planner for nondeterministic HTN domains, such as ND-SHOP
Main reference: Chad Hogg, Ugur Kuter, and Hector Munoz-Avila (2009) Learning Hierarchical Task Networks for Nondeterministic Planning Domains. Proceedings of the Twenty-first International Joint Conference on Artificial Intelligence (IJCAI-09). AAAI Press.
PDF
(Q-Maker, Q-Reinforce,Q-SHOP) We conducted research integrating HTN Learning and Reinforcement Learning. Our formalism associates with HTN methods Q-values that estimate the quality of plans that could be generated using them. This work consists of three algorithms: (1) Q-Maker, which follows a bottom-up procedure starting from input traces to learn HTN methods and initial estimates for their Q-values; (2) Q-Reinforce, which follows a top-down HTN plan generation procedure to refine the Q-value associated with methods; and (3) Q-SHOP, a variant of the HTN planner SHOP that reduces tasks by selecting the appropriate method with highest Q-value
Main reference: Chad Hogg, Ugur Kuter, and Hector Munoz-Avila (2010) Learning Methods to Generate Good Plans: Integrating HTN Learning and Reinforcement Learning. Proceedings of the Twenty-Fourth AAAI Conference on Artificial Intelligence (AAAI-2010). AAAI Press.
PDF

Downloads:

Note: HTN-Maker and variants are released under terms of the GNU general public license as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

HTN-Maker distribution (includes HTN-Maker and variants).
HTN-Maker experiment files.
HTN-Maker^ND experiment files.
Q-Maker, Q-Reinforce,Q-SHOP) experiment files.

Publications:

PDF

Chad Hogg, Ugur Kuter, and Hector Munoz-Avila (2010) Learning Methods to Generate Good Plans: Integrating HTN Learning and Reinforcement Learning. Proceedings of the Twenty-Fourth AAAI Conference on Artificial Intelligence (AAAI-10). AAAI Press.
PDF

Chad Hogg, Ugur Kuter, and Hector Munoz-Avila (2009) Learning Hierarchical Task Networks for Nondeterministic Planning Domains. Proceedings of the Twenty-first International Joint Conference on Artificial Intelligence (IJCAI-09). AAAI Press.
PDF

Chad Hogg, Ugur Kuter, and Hector Munoz-Avila (2009) From Plan Traces to Hierarchical Task Networks Using Reinforcements: A Preliminary Report. Proceedings of the IJCAI-09 Workshop on Learning Structural Information from Traces (STRUCK-09). AAAI Press.
PDF

Hogg, Chad, Munoz-Avila, Hector, and Ugur Kuter. (2008) HTN-Maker: Learning HTNs with Minimal Additional Knowledge Engineering Required. To appear in Proceedings of the Twenty-Third AAAI Conference on Artificial Intelligence (AAAI-08). AAAI Press.
PDF

Hogg, C. & Munoz-Avila, H. (2007) Learning of Tasks Models for HTN Planning. Proceedings of the ICAPS-07 Workshop on AI Planning and Learning (AIPL). AAAI Press.
PDF

Contributors:

Contact about web site: Hector Munoz-Avila

This material is based upon work supported by the National Science Foundation under Grant No. NSF 0642882. Any opinions, findings and conclusions or recomendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation (NSF)

Last updated: Thur. Apr. 9 11:49:51 EST 2010