Wearable Systems for Advancing Human Mobility: From Estimating Energy Expenditure and Personalizing Robotic Assistance

Abstract

Physical inactivity and mobility challenges remain major global health issues, affecting millions worldwide. Addressing them requires accurate physical activity monitoring and providing effective, adaptable interventions such as robotic assistance. This thesis presents the design and evaluation of three wearable systems to understand and improve mobility in everyday environments. First, we introduce OpenMetabolics, a biomechanically-informed activity monitor that employs a smartphone in a pants pocket to measure leg motion and accurately estimate energy expenditure. Using a data-driven machine learning model and a novel pocket motion artifact correction algorithm, OpenMetabolics reduces cumulative estimation error by approximately half compared to existing tools during real-world activities. Second, we quantify how sensing modality and use-case configuration affect the accuracy of commercial wearable devices in free-living conditions. By systematically evaluating smartwatches during real-world outdoor activities, we demonstrate that activity-specific tracking combining GPS, IMU, and heart rate data achieves near laboratory-grade accuracy for low-intensity activities like walking, while generic monitoring modes and wrist-only signals yield substantially larger errors. Third, we present a unified control framework that uses biomechanical simulation to generalize and personalize robotic assistance across diverse activities of daily living. Leveraging a reconfigurable assistive device, this approach utilizes portable sensors and simulation to automatically generalize assistance, followed by human-in-the-loop optimization to personalize the assistive forces. This method significantly reduces target muscle excitation during upper- and lower-limb activities, including sit-to-stand, stair climbing, and drinking, effectively offsetting the equivalent of years of age-related muscle weakness for older adults. Together, these accurate, accessible, and adaptable wearable systems could provide scalable tools to monitor health outcomes and effectively improve everyday human mobility.