Lower limb assistive robotic devices, such as active prosthesis, orthoses, and exoskeletons have the potential to restore function for the millions of Americans who experience mobility challenges due to injury and disability. Since individuals with mobility challenges have an increased energetic cost of transport, the benefit of such assistive devices is commonly assessed via the reduction in the metabolic work rate of the individual who is using the device. Currently, metabolic work rate can only be obtained in a laboratory environment, using breath-by-breath measurements of respiratory gas analysis. To obtain a single steady state data point of metabolic work rate, multiple minutes of data must be collected, since the signals are noisy, sparsely sampled, and dynamically delayed. In addition, the user has to wear a mask and bulky equipment, further restricting the applicability of the method on a larger scale. We propose an improved way to obtain such estimates of metabolic work rate in real-time.
Aim 1 will determine salient signal features and characterize the dynamics of sensing metabolic work rate from a variety of physiological sensor signals.
Aim 2 will use advanced sensor fusion and machine learning techniques to accurately predict instantaneous energy cost in real-time from multiple physiological signals without relying on a metabolic mask.
Aim 3 will use the obtained real-time estimates to optimize push-off timing for an active robotic prosthesis. The resulting methods will enable an automated and continuous evaluation of assistive robotic devices that can be realized outside the laboratory and with simple wearable sensors. This automated evaluation will enable devices, such as active prostheses, orthoses, or exoskeletons, that can self-monitor their performance, optimize their own behavior, and continuously adapt to changing circumstances. This will open up a radically new way of human-robot- interaction for assistive devices. It will greatly increase their clinical viability and enable novel advanced controllers and algorithms that can improve device performance on a subject specific basis.
A common way of evaluating assistive robotic devices, such as active prostheses or exoskeletons, is by measuring the reduction in effort that they bring to an individual walking in them. The proposed project will develop ways to perform this evaluation automatically and in real-time by the device itself, which will be used in the future to develop prostheses and exoskeletons that automatically adapt themselves to their users. This project supports the NIH's stated mission of reducing disability by improving patient outcomes with new prosthetic and orthotic devices.
