Enabling Ergonomic Human-Robot Interaction Through Large Language Modeling

Recent advancements in robotics, particularly in the development of collaborative robots (cobots), have significantly reshaped human-machine interaction by fostering more intelligent, adaptive, and human-centered systems. Traditionally regulated under the European Machinery Directive 2006/42/EC, robotic technologies now demand an updated framework due to the increasing autonomy and self-evolving behavior of modern machines. In response, Regulation (EU) 2023/1230 introduces revised Essential Health and Safety Requirements (EHSRs), with a notable focus on ergonomics through EHSR 1.1.6. This updated requirement broadens the scope of ergonomic considerations to include not only physical strain but also cognitive and psychological impacts within dynamic collaborative environments. Of particular relevance is point (g), which emphasizes that autonomous systems must engage in appropriate human interaction through verbal and non-verbal communication and transparently convey their intentions. This evolution in regulatory thinking aligns with the capabilities of Large Language Models (LLMs), whose advanced natural language understanding and generation offer substantial ergonomic benefits. LLMs facilitate context-sensitive, human-like communication, enhancing transparency, predictability, and usability in human-robot interaction. When integrated into multimodal interfaces, such as speech, gestures, and facial expressions, LLMs enable robots to better adapt to diverse users and interaction contexts. This paper aims to present a correlation table mapping the outputs of a robotic system based on Large Language Models to the new ergonomic requirements outlined in EHSR 1.1.6. Future research will compare LLM-based systems with traditional rule-based models, empirically assessing perceived anthropomorphism, system intelligence, and user cognitive workload. This approach represents a promising direction for the development of ergonomic, intelligible, and user-centered robotic technologies.