Online Learning of Action Models for PDDL Planning
Abstract
The automated learning of action models is widely recognised as a key and compelling challenge to address the difficulties of the manual specification of planning domains. Most state-of-the-art methods perform this learning offline from an input set of plan traces generated by the execution of (successful) plans. However, how to generate informative plan traces for learning action models is still an open issue. Moreover, plan traces might not be available for a new environment. In this paper, we propose an algorithm for learning action models online, incrementally during the execution of plans. Such plans are generated to achieve goals that the algorithm decides online in order to obtain informative plan traces and reach states from which useful information can be learned. We show some fundamental theoretical properties of the algorithm, and we experimentally evaluate the online learning of the action models over a large set of IPC domains.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The symbolic analogue of learning an action model inside a planning loop is classical action-model learning.
“The generate-test-repair loop is Reflexion-style iteration [Shinn et al., 2023, Madaan et al., 2023, Chen et al., 2024], and the symbolic analogue of learning an action model inside a planning loop is classical action-model learning [Lamanna et al., 2021, Yang et al., 2007, Aineto et al., 2019].”
From OPINE-World: Programmatic World Modeling with Ontology-error-Prioritized Interactive Exploration for ARC-AGI-3 (2026) · cites this paper as Lamanna et al., 2021
Classical action-model learning shares the planner-in-loop structure with symbolic synthesis in place of language-model synthesis.
“Classical action-model learning shares the planner-in-loop structure with symbolic synthesis in place of language-model synthesis [Lamanna et al., 2021].”
From OPINE-World: Programmatic World Modeling with Ontology-error-Prioritized Interactive Exploration for ARC-AGI-3 (2026) · cites this paper as Lamanna et al., 2021
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
No references could be resolved yet.
Research works printed but not linked: 17
- Michal Certicky. Real-time action model learning with online algorithm 3SG. Applied Artificial In- telligence, 28(7):690-711, 2014.
- Stephen Cresswell, Thomas Leo McCluskey, and Margaret Mary West. Acquiring planning domain models using LOCM. Knowledge Eng. Review, 28(2):195-213, 2013.
- García-Martínez and Borrajo, 2000] Ramón García- Martínez and Daniel Borrajo. An integrated approach of learning, planning, and execution. J. Intell. Robotic Syst., 29(1):47-78, 2000.
- Gil, 1994a] Yolanda Gil. Learning by experimentation: In- cremental refinement of incomplete planning domains. In ICML, 1994.
- Gil, 1994b] Yolanda Gil. Learning new planning operators by exploration and experimentation. In AAAI, 1994.
- Kitano and Tadokoro, 2001] Hiroaki Kitano and Satoshi Ta- dokoro. Robocup rescue: A grand challenge for multia- gent and intelligent systems. AI Mag., 22(1):39-52, 2001.
- Leonardo Lamanna, Alfonso Gerevini, Alessandro Saetti, Luciano Serafini, and Paolo Traverso. On-line learning of planning domains from sensor data in pal: Scaling up to large state spaces. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 11862-11869, 2021.
- Kira Mourão, Luke S. Zettlemoyer, Ronald P. A. Petrick, and Mark Steedman. Learning STRIPS operators from noisy and incomplete observa- tions. In UAI, 2012.
- Cyrill Stachniss, John J. Leonard, and Sebastian Thrun. Simultaneous localization and map- ping. In Springer Handbook of Robotics, Springer Hand- books, pages 1153-1176. Springer, 2016.
- R. S. Sutton and A. G. Barto. Re- inforcement learning -an introduction. Adaptive compu- tation and machine learning. MIT Press, 1998.
- Thomas J Walsh and Michael L Littman. Efficient learning of action schemas and web- service descriptions. In AAAI, 2008.
- Xuemei Wang. Planning while learning oper- ators. In AAAI, 1996.
- Joseph Z. Xu and John E. Laird. Instance-based online learning of deterministic relational action models. In AAAI, 2010.
- Joseph Z. Xu and John E. Laird. Com- bining learned discrete and continuous action models. In AAAI, 2011.
- Joseph Zhen Ying Xu and John E. Laird. Learning integrated symbolic and continuous ac- tion models for continuous domains. In AAAI, 2013.
- Zhuo and Kambhampati, 2013] Hankz Hankui Zhuo and Subbarao Kambhampati. Action-model acquisition from noisy plan traces. In IJCAI, 2013.
- Zhuo et al., 2010] Hankz Hankui Zhuo, Qiang Yang, Derek Hao Hu, and Lei Li. Learning complex action mod- els with quantifiers and logical implications. Artif. Intell., 174(18):1540-1569, 2010.
Cited by
Sorted by citations
Courtis, D., Li, W., & Sanner, S. (2026) · arXiv · doi:10.48550/arxiv.2607.01531
- Authors
- Lamanna, L., Saetti, A., Serafini, L., Gerevini, A., & Traverso, P.
- Affiliations
- Department of Information Engineering, University of Brescia, Italy
- Fondazione Bruno Kessler (FBK), Trento, Italy
- Venue
- Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence
- Published
- 2021-08
- Pages
- 4112-4118
- Cited by
- 8
- References
- 0
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
