Human-robot interaction using retrieval-augmented generation and fine-tuning with transformer neural networks in...

Area of Science:

• Robotics
• Artificial Intelligence
• Manufacturing Systems

Background:

• Modern manufacturing relies on Artificial Intelligence (AI) for automation and Human-Robot Interaction (HRI).
• Traditional robotic systems often lack flexibility and dynamic response capabilities in complex group working conditions.

Purpose of the Study:

• To propose a novel framework integrating Retrieval-Augmented Generation (RAG) with fine-tuned Transformer Neural Networks for advanced robotic decision-making.
• To enhance robotic flexibility, real-time responsiveness, and human-robot collaboration in manufacturing.

Main Methods:

• Development of a model combining RAG for knowledge acquisition and Transformers for optimization.
• Implementation of regret-based learning to enable robots to learn from past errors and improve future decisions.
• Validation through a numerical case study comparing the proposed system against conventional robotic systems.

Main Results:

• The proposed RAG and Transformer-based framework demonstrates improved robotic decision-making and flexibility.
• Regret-based learning contributes to predictable performance improvements in robotic systems.
• The system effectively addresses challenges in scalability, fine-tuning, multimodal learning, and ethical considerations in AI-driven robotics.

Conclusions:

• The developed framework offers a dynamic and intelligent approach to human-robot interaction in manufacturing.
• This research paves the way for advancements in Industry 5.0, intelligent manufacturing, and collaborative robotics.
• The study provides a clear implementation strategy for AI-based human-robot manufacturing systems.