Abstract:Today's classical planners are powerful, but modeling input tasks in formats such as PDDL is tedious and error-prone. In contrast, planning with Large Language Models (LLMs) allows for almost any input text, but offers no guarantees on plan quality or even soundness. In an attempt to merge the best of these two approaches, some work has begun to use LLMs to automate parts of the PDDL creation process. However, these methods still require various degrees of expert input. We present NL2Plan, the first domain-agnostic offline LLM-driven planning system. NL2Plan uses an LLM to incrementally extract the necessary information from a short text prompt before creating a complete PDDL description of both the domain and the problem, which is finally solved by a classical planner. We evaluate NL2Plan on four planning domains and find that it solves 10 out of 15 tasks - a clear improvement over a plain chain-of-thought reasoning LLM approach, which only solves 2 tasks. Moreover, in two out of the five failure cases, instead of returning an invalid plan, NL2Plan reports that it failed to solve the task. In addition to using NL2Plan in end-to-end mode, users can inspect and correct all of its intermediate results, such as the PDDL representation, increasing explainability and making it an assistive tool for PDDL creation.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is a key challenge in the fields of natural language processing and artificial intelligence planning: how to automatically generate a complete PDDL (Problem Domain Definition Language) description from a brief natural language description and generate an effective plan accordingly. Specifically, the paper introduces a system named NL2Plan, which is able to: 1. **Reduce human intervention**: The traditional PDDL modeling process is cumbersome and error - prone, requiring domain experts to write manually. NL2Plan, by using large - language models (LLM), automatically extracts necessary information from brief natural language descriptions to generate complete PDDL descriptions, thus reducing the need for human intervention. 2. **Improve plan quality**: Although direct use of LLM for planning can accept almost any input text, it cannot guarantee the quality or correctness of the plan. NL2Plan combines the flexibility of LLM and the precision of classical planners. It generates PDDL descriptions through LLM and then solves them with classical planners, thereby improving the quality and reliability of the plan. 3. **Achieve domain - independence**: Most existing methods still require a certain degree of domain - specific expertise when generating PDDL descriptions, while NL2Plan is a domain - independent system that can be applied in multiple different planning domains without the need for domain - specific adaptation or adjustment. 4. **Enhance interpretability**: NL2Plan not only generates the final plan but also allows users to check and correct intermediate results, such as PDDL representations, increasing the transparency and interpretability of the system, making it an auxiliary tool to help users better understand the task - modeling process. Through these features, NL2Plan aims to overcome the limitations of existing methods in the natural - language - to - PDDL conversion process and provide a more efficient, reliable, and user - friendly solution.

NL2Plan: Robust LLM-Driven Planning from Minimal Text Descriptions

LLM+P: Empowering Large Language Models with Optimal Planning Proficiency

Leveraging Environment Interaction for Automated PDDL Translation and Planning with Large Language Models

On the Planning Abilities of Large Language Models : A Critical Investigation

Query-Efficient Planning with Language Models

Leveraging Pre-trained Large Language Models to Construct and Utilize World Models for Model-based Task Planning

On the Planning Abilities of Large Language Models (A Critical Investigation with a Proposed Benchmark)

On the Limit of Language Models as Planning Formalizers

Planning in the Dark: LLM-Symbolic Planning Pipeline without Experts

Learning to Plan with Natural Language

Exploring and Benchmarking the Planning Capabilities of Large Language Models

Chasing Progress, Not Perfection: Revisiting Strategies for End-to-End LLM Plan Generation

Large Language Models as Planning Domain Generators

LLMs Can't Plan, But Can Help Planning in LLM-Modulo Frameworks

TIC: Translate-Infer-Compile for accurate "text to plan" using LLMs and Logical Representations

Robust Planning with LLM-Modulo Framework: Case Study in Travel Planning

Describe, Explain, Plan and Select: Interactive Planning with Large Language Models Enables Open-World Multi-Task Agents

LLM-Assist: Enhancing Closed-Loop Planning with Language-Based Reasoning

What's the Plan? Evaluating and Developing Planning-Aware Techniques for Language Models

SELP: Generating Safe and Efficient Task Plans for Robot Agents with Large Language Models