Generalized Planning With Deep Reinforcement Learning

TL;DR

This paper explores using Deep Reinforcement Learning combined with Graph Neural Networks to develop generalized planning policies that can effectively solve significantly larger problem instances than those used during training.

Contribution

It introduces a novel approach integrating Deep Reinforcement Learning and Graph Neural Networks for generalized planning, capable of scaling to larger problem instances.

Findings

Policies generalize to larger instances

Deep RL with GNNs outperforms traditional methods

Effective transfer from small to large problems

Abstract

A hallmark of intelligence is the ability to deduce general principles from examples, which are correct beyond the range of those observed. Generalized Planning deals with finding such principles for a class of planning problems, so that principles discovered using small instances of a domain can be used to solve much larger instances of the same domain. In this work we study the use of Deep Reinforcement Learning and Graph Neural Networks to learn such generalized policies and demonstrate that they can generalize to instances that are orders of magnitude larger than those they were trained on.