LLM-as-BT-Planner: Leveraging LLMs for Behavior Tree Generation in Robot Task Planning

TL;DR

This paper introduces LLM-as-BT-Planner, a framework that uses large language models to automatically generate behavior trees for robot task planning, reducing manual effort and improving robustness in assembly tasks.

Contribution

It presents four in-context learning methods and evaluates fine-tuned models for BT generation, advancing the use of LLMs in robotic behavior planning.

Findings

Enhanced BT generation success rate with in-context learning

Fine-tuned LLMs outperform baseline models in task planning

Framework applicable in both simulated and real-world settings

Abstract

Robotic assembly tasks remain an open challenge due to their long horizon nature and complex part relations. Behavior trees (BTs) are increasingly used in robot task planning for their modularity and flexibility, but creating them manually can be effort-intensive. Large language models (LLMs) have recently been applied to robotic task planning for generating action sequences, yet their ability to generate BTs has not been fully investigated. To this end, we propose LLM-as-BT-Planner, a novel framework that leverages LLMs for BT generation in robotic assembly task planning. Four in-context learning methods are introduced to utilize the natural language processing and inference capabilities of LLMs for producing task plans in BT format, reducing manual effort while ensuring robustness and comprehensibility. Additionally, we evaluate the performance of fine-tuned smaller LLMs on the same tasks. Experiments in both simulated and real-world settings demonstrate that our framework enhances LLMs' ability to generate BTs, improving success rate through in-context learning and supervised fine-tuning.