Graph Learning for Numeric Planning

TL;DR

This paper introduces graph learning models tailored for numeric planning, leveraging relational structures and novel graph kernels to improve efficiency, generalization, and heuristic learning in numeric planning tasks.

Contribution

It presents a new graph kernel for continuous and categorical attributes and optimization methods for heuristic learning, advancing numeric planning solutions.

Findings

Graph kernels outperform graph neural networks in efficiency and generalization.

The models achieve competitive coverage with domain-independent numeric planners.

Proposed methods are data-efficient and interpretable.

Abstract

Graph learning is naturally well suited for use in symbolic, object-centric planning due to its ability to exploit relational structures exhibited in planning domains and to take as input planning instances with arbitrary numbers of objects. Numeric planning is an extension of symbolic planning in which states may now also exhibit numeric variables. In this work, we propose data-efficient and interpretable machine learning models for learning to solve numeric planning tasks. This involves constructing a new graph kernel for graphs with both continuous and categorical attributes, as well as new optimisation methods for learning heuristic functions for numeric planning. Experiments show that our graph kernels are vastly more efficient and generalise better than graph neural networks for numeric planning, and also yield competitive coverage performance compared to domain-independent numeric planners. Code is available at https://github.com/DillonZChen/goose