Control Approaches for Exosuits in Clinical Populations during Diverse Activities

Abstract

Individuals with musculoskeletal or neurological diseases often experience mobility challenges, including reduced walking efficiency and limited automaticity. For example, gait patterns of individuals post-stroke are hemiparesis, characterized with slowness and asymmetry; more than half of individuals with Parkinson's Disease (PD) experience Freezing of Gait (FoG), characterized with brief, episodic reduction or absence of forward progression. Despite existing solutions such as rehabilitation and passive assistive devices, these impairments elevate injury risk and diminish quality of life for clinical populations. Wearable electromechanical devices have been developed to restore mobility, compensate for impairments, and enhance rehabilitation outcomes for clinical populations. Specifically, joint-targeting devices have achieved notable success in clinical populations to reduce physical effort, increase practice volume, and retrain regular biomechanics. Nevertheless, controllers of joint-targeting devices have been tuned for continuous, steady walking, relying on gait phase estimation or gait event detection algorithms only reliable under supervised, controlled conditions. Pathological gait poses additional challenges in controllers with atypical and varying joint kinematics, short and intermittent walking bouts, and impaired endurance. This thesis seeks to address the challenge in assisting pathological gait in real world settings by designing and evaluating control systems that are robust in diverse activities. First, we deployed a unilateral hip flexion exosuit for inpatient gait retraining in clinics. We expanded the use of exosuit to non-ambulatory patients and non-walking repetition practices during actual rehabilitation sessions by implementing assistance triggered by pysical therapisyt (PT). The exosuit eliminated the needs for manual limb support from PTs and increased participants' gait endurance or speed. We also executed a community Robotic Exosuit Augmented Locomotion program to support community walking practices with a semi-active ankle exosuit. Four participants in the chronic stage of stroke independently used the ankle exosuit for 4 weeks. Two participants significantly increased their unassisted paretic propulsion and daily steps following the program. Second, we developed a turning-specific controller to provide assistance during transitional activities and reduce FoG for individuals with PD This is an advancement from a case study of a single individual performing straight-line walking to a full study of nine participants performing turning and freeze-provoking activity. We developed a controller for a bilateral hip flexion exosuit to accommodate asymmetry and reduced range of motion during turning. The turning controller successfully assisted various activities, including individual-specific hotspots, and led to a significant reduction in freeze severity and improvements in turn quality. Third, we further innovated the controller for individuals with PD and established a control strategy combining adaptive oscillators and machine learning models to accommodate complex freeze-provoking trajectories at home. We introduced a phase-coupling term in adaptive oscillators to greatly improve the reliability and adaptability of gait phase estimation in transitional activities. As gait further deteriorated due to FoG, we switched to a kinetic-based gait phase realized by a machine learning (ML)-estimated weight distribution, along with ML-estimated FoG probability to determine FoG in real time. This control strategy was validated in complex at-home hotspots, demonstrating reliable assistance across various gait representations and a successful reduction in freeze severity and improvement in functional outcome. Together, this thesis advances control systems for wearable assistive devices that consistently assist pathological gait patterns, validated in real-world deployment in clinics and communities for individuals post-stroke, and versatilely adapt to diverse activities and complex trajectories at home, demonstrating a life changing reduction in FoG for individuals with PD. We removed several key barriers and established a practical path to implement exosuits from laboratory to clinics and home.