Llm-Basiertes Aufgabenplanungsrahmenwerk für einen Roboter zur Ausführung Domänenspezifischer Benutzungsfälle (dsus)

Details

Veröffentlichungs-Nr.

EP4755593

Anmeldetag

20. November 2025

Veröffentlichung

10. Juni 2026

Rechtsraum

EP

IPC

B25J9/16

Abstract

Applying large language models (LLMs), even with very large token-sized prompts, does not achieve the task planning performance required for a domain-specific industrial use case (DSU). The method and system disclosed overcome the obstacles of a robotic task planner for DSUs by introducing a task planning framework. Central to the framework is a robotic system ontology that organizes the components of the robotic system in a coherent and systematic manner. This ontology empowers the planning framework to efficiently capture a contextual representation of a DSU using the LLM. Additionally, the research introduces a LLM-tuning regimen referred as chain of hierarchical thought (CoHT), specifically crafted to complement the robotic system ontology. Integrating these two components enables enhancing accuracy, robustness, and throughput of the robot in a cost-effective manner. Furthermore, provided is an empirical methodology to generate LLM-tuning datasets size for a guaranteed performance, leveraging a heuristics-based method.

Anmelder

Firma

Tata Consultancy Services Limited

Land

🇮🇳 Indien