LLM-Powered Personalized Feedback System for Local Muscular Fatigue Monitoring Using Electromyography during Construction Material Handling
Document Type
Conference Proceeding
Publication Date
1-1-2025
Abstract
Wearable health monitoring technologies empowered by AI and ML are transforming occupational health management. EMG sensors provide continuous muscle activity tracking, helping effectively prevent work-related fatigue. This advancement is vital in construction, where physical strain and accidents threaten worker safety. However, existing data-driven EMG models deliver objective measurements but lack actionable and predictive capabilities. We introduce an intelligent feedback system using a GPT-powered RAG framework that integrates EMG data with expert fatigue-assessment knowledge to deliver personalized, real-time recommendations tailored to tasks and individual conditions. Beyond traditional monitoring, it dynamically assesses muscle fatigue and generates proactive, context-aware guidance to mitigate risks. We validated the system in simulated exoskeleton-assisted material handling within realistic construction environments. Performance was evaluated with RAG metrics-context relevance, answer relevance, and groundedness-and results show improved and effective insights compared to conventional methods. This paradigm establishes a new standard for personalized, AI-driven occupational health in construction.
Publication Source (Journal or Book title)
Computing in Civil Engineering 2025 Resilient Robotic and Educational Systems Selected Papers from the ASCE International Conference on Computing in Civil Engineering 2025
First Page
750
Last Page
758
Recommended Citation
Gautam, Y., Habibnezhad, M., Liu, Y., & Jebelli, H. (2025). LLM-Powered Personalized Feedback System for Local Muscular Fatigue Monitoring Using Electromyography during Construction Material Handling. Computing in Civil Engineering 2025 Resilient Robotic and Educational Systems Selected Papers from the ASCE International Conference on Computing in Civil Engineering 2025, 750-758. https://doi.org/10.1061/9780784486443.082