WLPlan Documentation

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)

Get Started

• Installation
• Introduction by Example

Tutorials

• Graphs and NetworkX
• Graph Neural Networks

Package Reference

• wlplan
• wlplan.data
• wlplan.feature_generator
• wlplan.graph_generator
• wlplan.planning