| dc.description.abstract | Robotic lower-limb wearable assistive technologies have shown promise as both human augmentation and rehabilitation devices; however, accounting for unexpected disturbances (such as a trip or stumble) is a crucial gap in their development. This dissertation explores how stumble recovery behaviors might be implemented into lower-limb exoskeletal and prosthetic devices, and to what extent they might be effective in reducing the likelihood of falls.
In order to investigate stumble recovery behaviors, a stumble perturbation system was designed, built, and validated to realistically and safely study the stumble responses of various populations in the laboratory. Next, the biomechanical dataset from these validation experiments was leveraged to elucidate the healthy adult stumble recovery strategy selection process with a machine learning approach. Using insights from these works, three key elements of a stumble recovery controller were designed and tested in a powered knee exoskeleton: stumble detection, stumble recovery strategy identification, and stumble recovery strategy assistance. Detection and identification algorithms yielded high detection and classification accuracies when tested with real-time exoskeleton sensor data from a five-participant stumble experiment. In a separate experiment, knee exoskeleton assistance was shown to improve recovery limb kinematics and reduce fall risk for three healthy adults whose recovery was impaired due to leg weights attached to the shank. In parallel, the sound-side stumble responses of six transfemoral prosthesis users wearing their prescribed passive prostheses were characterized to identify the main stumble recovery deficiencies of this population and suggest potential interventions. One of the interventions suggested, a powered knee prosthesis, was tested for sound- and prosthetic-side stumbles with a subset of these transfemoral prosthesis users. Overall, the powered prosthesis intervention served to reduce fall risk and compensation techniques.
Ultimately, this work uses novel experimental techniques and comprehensive biomechanics analyses to propose, implement, and evaluate robotic knee prosthesis and exoskeleton interventions for stumble recovery. Results suggest that robotic knee assistance at both the recovery and support limb are promising approaches for improving stumble recovery and reducing fall risk for wearers. | |
