Learning HTN Methods with Preference from HTN Planning Instances

Learning HTN Methods with Preference from HTN Planning Instances. Xiao, Z., Wan, H., Zhuo, H. H., Herzig, A., Perrussel, L., & Chen, P. In Proceedings of the 2nd ICAPS Workshop on Hierarchical Planning (HPlan 2019), pages 31–39, 2019. A follow-up paper was later accepted at AAAI 2020.

Paper abstract bibtex

The hierarchical task network (HTN) planning technique is used in a growing number of real-world applications. However in many domains, such as the logistics domain, as there exist thousands of cases, it is difficult and time-consuming for humans to specify all HTN methods to cover all desirable plans. This suggests that it is important to learn HTN methods to accomplish the tasks via decomposition. The traditional HTN-method learning approaches require complete executable plans and annotated tasks, which are often difficult to acquire in real-world applications. In this paper, we propose a novel framework to learn HTN methods from HTN instances with incomplete method sets and without annotated tasks. Besides, previous approaches demand total orders on the subtasks in the methods while our approach is capable of learning methods with partial orders. To reduce the number of methods learned, we consider priorities on methods and compute the minimal set of methods based on prioritized preferences. By taking experiments on three well-known planning domains, we demonstrate that our approach is effective, especially on solving new HTN problems.

@InProceedings{HPlan2019paper5,
  author    = {Zhanhao Xiao and Hai Wan and Hankui Hankz Zhuo and Andreas Herzig and Laurent Perrussel and Peilin Chen},
  title     = {Learning HTN Methods with Preference from HTN Planning Instances},
  booktitle = {Proceedings of the 2nd ICAPS Workshop on Hierarchical Planning (HPlan 2019)},
  year      = {2019},
  pages     = {31--39},
  abstract  = {The hierarchical task network (HTN) planning technique is used in a growing number of real-world applications. However in many domains, such as the logistics domain, as there exist thousands of cases, it is difficult and time-consuming for humans to specify all HTN methods to cover all desirable plans. This suggests that it is important to learn HTN methods to accomplish the tasks via decomposition. The traditional HTN-method learning approaches require complete executable plans and annotated tasks, which are often difficult to acquire in real-world applications. In this paper, we propose a novel framework to learn HTN methods from HTN instances with incomplete method sets and without annotated tasks. Besides, previous approaches demand total orders on the subtasks in the methods while our approach is capable of learning methods with partial orders. To reduce the number of methods learned, we consider priorities on methods and compute the minimal set of methods based on prioritized preferences. By taking experiments on three well-known planning domains, we demonstrate that our approach is effective, especially on solving new HTN problems.},
  url_paper = {https://openreview.net/pdf?id=r1e6USBWF4},
  note      = {A follow-up paper was later accepted at AAAI 2020.}
}

Downloads: 0

{"_id":"gwhFvka67zJrKoRnf","bibbaseid":"xiao-wan-zhuo-herzig-perrussel-chen-learninghtnmethodswithpreferencefromhtnplanninginstances-2019","author_short":["Xiao, Z.","Wan, H.","Zhuo, H. H.","Herzig, A.","Perrussel, L.","Chen, P."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["Zhanhao"],"propositions":[],"lastnames":["Xiao"],"suffixes":[]},{"firstnames":["Hai"],"propositions":[],"lastnames":["Wan"],"suffixes":[]},{"firstnames":["Hankui","Hankz"],"propositions":[],"lastnames":["Zhuo"],"suffixes":[]},{"firstnames":["Andreas"],"propositions":[],"lastnames":["Herzig"],"suffixes":[]},{"firstnames":["Laurent"],"propositions":[],"lastnames":["Perrussel"],"suffixes":[]},{"firstnames":["Peilin"],"propositions":[],"lastnames":["Chen"],"suffixes":[]}],"title":"Learning HTN Methods with Preference from HTN Planning Instances","booktitle":"Proceedings of the 2nd ICAPS Workshop on Hierarchical Planning (HPlan 2019)","year":"2019","pages":"31–39","abstract":"The hierarchical task network (HTN) planning technique is used in a growing number of real-world applications. However in many domains, such as the logistics domain, as there exist thousands of cases, it is difficult and time-consuming for humans to specify all HTN methods to cover all desirable plans. This suggests that it is important to learn HTN methods to accomplish the tasks via decomposition. The traditional HTN-method learning approaches require complete executable plans and annotated tasks, which are often difficult to acquire in real-world applications. In this paper, we propose a novel framework to learn HTN methods from HTN instances with incomplete method sets and without annotated tasks. Besides, previous approaches demand total orders on the subtasks in the methods while our approach is capable of learning methods with partial orders. To reduce the number of methods learned, we consider priorities on methods and compute the minimal set of methods based on prioritized preferences. By taking experiments on three well-known planning domains, we demonstrate that our approach is effective, especially on solving new HTN problems.","url_paper":"https://openreview.net/pdf?id=r1e6USBWF4","note":"A follow-up paper was later accepted at AAAI 2020.","bibtex":"@InProceedings{HPlan2019paper5,\n author = {Zhanhao Xiao and Hai Wan and Hankui Hankz Zhuo and Andreas Herzig and Laurent Perrussel and Peilin Chen},\n title = {Learning HTN Methods with Preference from HTN Planning Instances},\n booktitle = {Proceedings of the 2nd ICAPS Workshop on Hierarchical Planning (HPlan 2019)},\n year = {2019},\n pages = {31--39},\n abstract = {The hierarchical task network (HTN) planning technique is used in a growing number of real-world applications. However in many domains, such as the logistics domain, as there exist thousands of cases, it is difficult and time-consuming for humans to specify all HTN methods to cover all desirable plans. This suggests that it is important to learn HTN methods to accomplish the tasks via decomposition. The traditional HTN-method learning approaches require complete executable plans and annotated tasks, which are often difficult to acquire in real-world applications. In this paper, we propose a novel framework to learn HTN methods from HTN instances with incomplete method sets and without annotated tasks. Besides, previous approaches demand total orders on the subtasks in the methods while our approach is capable of learning methods with partial orders. To reduce the number of methods learned, we consider priorities on methods and compute the minimal set of methods based on prioritized preferences. By taking experiments on three well-known planning domains, we demonstrate that our approach is effective, especially on solving new HTN problems.},\n url_paper = {https://openreview.net/pdf?id=r1e6USBWF4},\n note = {A follow-up paper was later accepted at AAAI 2020.}\n}\n\n","author_short":["Xiao, Z.","Wan, H.","Zhuo, H. H.","Herzig, A.","Perrussel, L.","Chen, P."],"key":"HPlan2019paper5","id":"HPlan2019paper5","bibbaseid":"xiao-wan-zhuo-herzig-perrussel-chen-learninghtnmethodswithpreferencefromhtnplanninginstances-2019","role":"author","urls":{" paper":"https://openreview.net/pdf?id=r1e6USBWF4"},"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"hierarchical-task.net/bibtex/hplan-publications.bib","dataSources":["T9LQLt3D2MDhrCfjh"],"keywords":[],"search_terms":["learning","htn","methods","preference","htn","planning","instances","xiao","wan","zhuo","herzig","perrussel","chen"],"title":"Learning HTN Methods with Preference from HTN Planning Instances","year":2019}