LLMs struggle to outperform engineered systems in robotics planning, study finds

A new study published on arXiv explores the integration of Large Language Models (LLMs) with Task and Motion Planning (TAMP) for robotics. Researchers developed 16 algorithms to substitute key TAMP components with LLMs, conducting 13,750 zero-shot experiments across three domains. The findings indicate that LLM-based planners generally show lower success rates and longer planning times compared to traditional engineered systems. Specifically, providing geometric details led to more task-planning errors than pure PDDL descriptions, and direct LLM approaches outperformed reasoning-based LLM variants in most scenarios. AI

IMPACT This research suggests that while LLMs show promise, current engineered systems remain superior for complex robotics planning tasks, indicating areas for future development.

RANK_REASON The cluster contains a research paper detailing experiments and findings on LLMs for robotics planning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• Jorge Mendez-Mendez
• Large Language Models
• PDDLStream
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →