Content area
As humans age, maintaining stability during activities of everyday living becomes more difficult. Performance on balance-related tasks diminishes, based on both clinical and analytical metrics. This decline is due in part to age-related, joint level changes in torque production, which are in turn correlated with increased fall risk. Thus, there is great excitement in the clinical and research community regarding robotic mobility assistive devices that can supplement torque and power production at different joints. While there has been progress demonstrating that assistive devices may lower metabolic cost during walking, less is known about how these devices may affect stability, particularly during feet-in-place activities such as sit-to-stand or standing. There is therefore a need to incorporate age-related neuromuscular decrements and powered mobility assistance into a single framework for stability analysis. Such a framework will not only aid in our understanding of the mechanics underlying robot-assisted movement stability, but it can also guide the design of perturbation experiments on older adults by suggesting conditions under which stability is most likely to be affected.
In this dissertation, we develop a method to characterize the effect of robotic assistance on human stability from a formal methods perspective: instead of analyzing individual control strategies for a given model, we compute a set representing all states from which it is feasible to maintain balance. We call this set the stabilizable region. In contrast to prior work, we provide mathematical guarantees that the stabilizable region includes all possible recoverable states. We also guarantee that it is always possible to remain within the region, which is an important safety property known as invariance.
To validate our method, we first compute stabilizable regions for nominal full-strength models of young adults. Using a published dataset of perturbed sit-to-stand motion, we show that the boundaries of these regions accurately detect when the participant can no longer use feet-in-place strategies to maintain stability. Next, to understand how old age and assistive devices affect stabilizability, we compute stabilizable regions for sex- and age- adjusted models of adults wearing ankle exoskeletons. Lastly, we analyze how ‘imperfect’ ankle exoskeletons may affect feasible standing balance. We consider devices that provide lower- and higher-than-necessary torques relative to the optimal amount required to maximize static stability. We relax our assumption that the ankle exoskeleton controller is perfectly tuned to the user, and compute the stabilizable region when the mass and relative center-of-mass height of the user are under- or over-estimated. We also approximate a delay in the exoskeleton’s actuation torque, quantifying how the stabilizable region changes over a range of device delays.
Our analysis shows that common ankle exoskeleton control strategies can improve stability in many nominal conditions that are likely to be encountered in daily life. However, they may hinder stability under certain conditions and can amplify existing, joint-level torque deficits related to aging. When the exoskeleton model is adjusted to account for realistic defects such as poorly tuned parameters and device delay, stability is hindered even further. Thus, our results suggest that a great deal more research is needed to enable the safe and effective widespread use of these devices, particularly in clinical populations. Even exoskeletons with mild imperfections using seemingly well-understood control strategies, may have a counterintuitive and undesired effect on standing balance in older adults, particularly those with impaired strength.