Advancing Back-Support Exoskeleton Control: A Biomechanical Approach

Musculoskeletal disorders, particularly low-back pain (LBP), are major concerns in physically demanding jobs. This thesis investigates active back-support exoskeletons' role in mitigating low-back load, a significant LBP contributor, by refining their control mechanisms for lifting activities. Chapter 2 presents a novel control strategy utilizing trunk angular acceleration for lifting assistance, decreasing L5/S1 disc spinal compression by up to 16%. This approach provides a comfortable alternative to traditional trunk angle-based controls, showing comparable effects on spinal compression force, lumbar moment, and muscle activation. Chapter 3 details a regression model for estimating L5/S1 joint moments, simplifying sensor requirements while accurately estimating lumbosacral joint active moments. This model aims to facilitate exoskeleton control design by predicting low-back load, demonstrating high accuracy in L5/S1 estimations. Subsequent chapters focus on applying this regression model. Chapter 4 examines calibration optimization in real-world settings, identifying efficient calibration methods for workplace exoskeleton control integration. Chapter 5 aims to simplify the model for easier workplace application by reducing measurement tool requirements. This effort maintains model accuracy while ensuring feasibility for real-world use. Chapter 6 assesses muscle fatigue's impact on low-back load estimation, finding minimal accuracy deviation in fatigued conditions, highlighting the model's robustness. The discussion addresses challenges in implementing these models in workplace exoskeletons, including calibration, EMG integration, user support perception, and fatigue effects. Preliminary trials show promising tailored support from the model-integrated exoskeletons. This research significantly progresses exoskeleton control optimization for reducing occupational low-back load, marking a vital step towards developing effective, user-centric exoskeletons. Future studies will continue exploring these models' real-world applications.