Language Models For Generalised PDDL Planning: Synthesising Sound and Programmatic Policies

TL;DR

This paper introduces LMPlan, a novel approach using language models to generate sound, generalised policies for PDDL planning problems, outperforming traditional planners and recent LM methods in complex scenarios.

Contribution

It presents a method for synthesising provably sound policies directly from language models without external verification, capable of handling large PDDL problems.

Findings

LMs can generate policies solving more PDDL problems than traditional planners.

Policies are sound relative to the PDDL domain without external checks.

LMs sometimes plan effectively with meaningless symbolic representations.

Abstract

We study the usage of language models (LMs) for planning over world models specified in the Planning Domain Definition Language (PDDL). We prompt LMs to generate Python programs that serve as generalised policies for solving PDDL problems from a given domain. Notably, our approach synthesises policies that are provably sound relative to the PDDL domain without reliance on external verifiers. We conduct experiments on competition benchmarks which show that our policies can solve more PDDL problems than PDDL planners and recent LM approaches within a fixed time and memory constraint. Our approach manifests in the LMPlan planner which can solve planning problems with several hundreds of relevant objects. Surprisingly, we observe that LMs used in our framework sometimes plan more effectively over PDDL problems written in meaningless symbols in place of natural language; e.g. rewriting (at dog kitchen) as (p2 o1 o3). This finding challenges hypotheses that LMs reason over word semantics and memorise solutions from its training corpus, and is worth further exploration.