Scalable Task Planning via Large Language Models and Structured World Representations

TL;DR

This paper presents a scalable task planning approach that combines large language models with structured world representations to efficiently handle complex, large-scale environments, demonstrated through simulations and real-world robot experiments.

Contribution

It introduces a method to leverage LLMs for pruning planning state spaces, improving scalability in task planning for large environments.

Findings

Effective pruning of irrelevant states using LLMs

Significant reduction in planning complexity

Successful real-world robot validation

Abstract

Planning methods struggle with computational intractability in solving task-level problems in large-scale environments. This work explores leveraging the commonsense knowledge encoded in LLMs to empower planning techniques to deal with these complex scenarios. We achieve this by efficiently using LLMs to prune irrelevant components from the planning problem's state space, substantially simplifying its complexity. We demonstrate the efficacy of this system through extensive experiments within a household simulation environment, alongside real-world validation using a 7-DoF manipulator (video https://youtu.be/6ro2UOtOQS4).