---

WLPlan

Plan is a library for generating graph representations and feature embeddings of PDDL planning problems and states for machine learning tasks. WLPlan currently supports both classical and numeric planning problems.

The main pipeline in WLPlan consists of (1) converting planning problems and states into graphs, and (2) synthesising feature embeddings by running a variant of the Weisfeiler-Leman (WL) algorithm on the resulting graph.

The entire pipeline of generating feature embeddings from PDDL planning problems can be performed in a matter of steps with WLPlan:

feature_generator = init_feature_generator(domain)
feature_generator.collect(dataset)
X = feature_generator.embed(dataset)

Detailed documentation for WLPlan can be found in the official website available here.

Installation

Python Interface

The Python interface can be installed simply with

pip install wlplan

The PyPI release only supports python>=3.10. Alternatively, you can also install the package from the source code with the install.sh script.

C++ Interface

The C++ interface can be installed in your project by running

./cmake_build.py <path/to/installation>

and adding the following to the root CMakeLists.txt file of your project

list(APPEND CMAKE_PREFIX_PATH "<path/to/installation>")
find_package(wlplan)
...
target_link_libraries(<your_project> PRIVATE wlplan)

References

Academic Publications

Some academic publications which use WLPlan are listed as follows.

• Daniel Höller. Learning Heuristic Functions for HTN Planning. In ICAPS 2025 Workshop on Bridging the Gap Between AI Planning and Reinforcement Learning (PRL), 2025.
• Dillon Z. Chen. Symmetry-Invariant Novelty Heuristics via Unsupervised Weisfeiler-Leman Features. In ICAPS 2025 Workshop on Heuristics and Search for Domain-independent Planning (HSDIP), 2025.
• Dillon Z. Chen. Weisfeiler-Leman Features for Planning: A 1,000,000 Sample Size Hyperparameter Study. In 28th European Conference on Artificial Intelligence (ECAI), 2025.
• Mingyu Hao, Dillon Z. Chen, Felipe Trevizan, and Sylvie Thiébaux. Effective Data Generation and Feature Selection in Learning for Planning. In 28th European Conference on Artificial Intelligence (ECAI), 2025.
• Rebecca Eifler, Nika Beriachvili, Arthur Bit-Monnot, Dillon Z. Chen, Jan Eisenhut, Jörg Hoffmann, Sylvie Thiébaux, and Florent Teichteil-Königsbuch. An Operator-Centric Trustable Decision-Making Tool for Planning Ground Logistic Operations of Beluga Aircraft. In 28th European Conference on Artificial Intelligence (ECAI), 2025.
• Dillon Z. Chen and Sylvie Thiébaux. Graph Learning for Numeric Planning. In 38th Conference on Neural Information Processing Systems (NeurIPS), 2024.
• Dillon Z. Chen, Felipe Trevizan, and Sylvie Thiébaux. Return to Tradition: Learning Reliable Heuristics with Classical Machine Learning. In 34th International Conference on Automated Planning and Scheduling (ICAPS), 2024.

Bibtex

The academic reference for WLPlan is given by the bibtex entry

@article{chen-wlplan-2024,
  author       = {Dillon Z. Chen},
  title        = {WLPlan: Relational Features for Symbolic Planning},
  journal      = {CoRR},
  volume       = {abs/2411.00577},
  year         = {2024},
}