RESEARCH
Design of Personalized, Adaptive Control Strategies for Assistive Robots
Assistive robots (like protheses and exoskeletons) have the potential to meaningfully transform human mobility. Yet, fundamentally, we do not yet know how to apply robotic assistance to the human body in order to promote meaningful clinical improvements or achieve targeted physiological goals.
We work towards closing this gap through the design and experimental evaluation of personalized, adaptive control strategies for assistive robotic devices. We design control systems with the human body “in the loop” and seek to identify objectives for assistance that maximize an individual user’s mobility. We work closely with people who may benefit from assistive devices, including adults, children, and toddlers with or without disabilities.
Interest Areas: Controls, Neuromechanics, Physiological Sensing, Human-in-the-Loop Optimization
Quantifying User Preference and Perception using Robotic Exoskeletons
Optimizing exoskeleton assistance to maximize a user’s individual mobility goals is a challenge, because users may prioritize many different objective simultaneously (like energy cost, comfort, or stability). Understanding these quantities can illuminate key features of exoskeleton assistance needed to inform future control paradigms.
We study how individuals wearing robotic exoskeletons perceive and interact with robotic assistance, towards the goal of developing intuitive, effective control strategies for users with and without disabilities.
We also investigate how different biofeedback paradigms can guide individuals in learning how to use assistive tehcnologies, and how different modalities change the way users interact with the devices.
Related Publications: [Pub 1] [Pub 2] [Pub 3]
Interest Areas: Controls, Human-Robot Interaction, Neuromechanics, Psychophysics
Leveraging Machine Learning and Wearable Sensors to Quantify Human Movement
Real-time estimates of energy expenditure (or other physiological metrics) can be used for a variety of health-related applications, such as at-home rehabilitation monitoring or studying assistive device use in the community. For assistive devices, these measurements are necessary for adaptive control strategies to respond to real-time changes in the user’s physiology. Yet, obtaining real-time measurements of physiological signals is notoriously challenging.
We utilize machine learning and wearable sensors to build data-driven predictive models of relevant physiological metrics outside the lab environment. We seek to investigate both black-box machine learning methods and causal models that provide physiolgical intuition and interpretability.
We leverage a wide range of sensors, like electromyography (EMG), inertial measurement units (IMUs), GPS, heart rate monitors, or pulse oximeters, to obtain signals from the human body and design inteligent methods that measure and describe human movement.
Related Publications: [Pub 1] [Pub 2] [Pub 3] [Pub 4]
Interest Areas: Machine Learning, Data Science, Physiological Sensing
Increasing Early Access to Powered Mobility & Accessible Computer Use
Photos shared with parental permission.
For toddlers with disabilities, interacting with technology can provide access to mobility, communication, and play. These devices often enable their frst experiences of their own self-initiated actions. But, children must be able to interact with these technologies in a developmentally appropriate way in order to efectively use them.
We study how early access to powered mobility impacts development, language, and movement, and how young children learn to use a joystick interface for mobility and accessible computer use.
Interest Areas: Neuromechanics, Human-Computer Interaction, Pediatric Rehabilitation