Collaborative Research: Detecting Gait Phases with Raised Metabolic Cost using Robotic Perturbations and System Identification for Enabling Targeted Rehabilitation Therapy

能够轻松行走与独立和生活质量密切相关。衰老伴随着行动能力的显著下降。现有的治疗方法和疗法依赖于对步行努力的呼吸测量。这些呼吸测量只能量化行走的平均努力程度。由于这一局限性，现有的治疗方法和疗法有时无法针对最需要帮助的步行运动阶段。这个项目将使用数据驱动的方法和模型来克服在衡量步行努力能力方面的限制。获得这一新信息将使评估治疗如何影响运动的不同阶段成为可能。数据驱动的方法最初将使用由计算机生成的数据集来开发，该数据集是通过对特定运动阶段的物理诱导变化来建立步行模型的。例如，向前拉力将应用于模型的腰部，以诱导变化，这些变化可以被用来检测步行努力的波动。这种计算机行走模型提供了对行走所需工作量的完整测量，这将用于验证数据驱动的方法。接下来，将使用真实人类行走实验的测量结果来验证新的数据驱动方法。在这些人体实验中，连接到参与者腰部的机器人绳索将施加拉力，以诱导变化，这些变化将被用来检测不同运动阶段的努力。在最后的研究中，这些方法将被用来确定年轻人和老年人步行所需的努力有何不同。通过对年轻人和老年人进行人体实验，将在运动的每个阶段表征努力的差异。该项目的成果将有助于创造改善所有运动阶段的强化治疗和辅助装置。在整个项目中，研究人员将通过Osher终身学习研究所为老年人提供关于步行的力学和健康方面的课程，以及数据科学和数字工程。该项目的目标是利用新的数据驱动方法来表征老年人和年轻人步态周期各阶段代谢成本的差异。该项目将结合基于系统识别和机器人扰动的新颖、数据驱动的方法来表征无法直接测量的信号的时间分布，例如代谢成本。第一个目标将在模拟步态数据中产生代谢成本的时间分布。将基于加权回归、神经网络和自动编码器开发新的数据驱动方法，以从生物力学信号中识别代谢成本时间分布。最初，这些方法将在预测步行模拟中创建，从该模拟中完全知道新陈代谢时间分布，以便可以在开发过程中对新方法进行评估。第二个目标是在人体实验中评估不同的时间分布估计方法。在第一个目标中创建的方法将通过使用机器人扰动的人体实验进行测试。使用数据驱动的方法来检测摆动和推离变化的能力也将通过人体实验进行调查，人体实验中使用弹性脚踝绳索或附加质量来直接改变步态周期。第三个目标将描述老年人和年轻人之间步态周期各阶段的成本贡献的差异。第一个子任务将通过应用数据驱动的方法来使用年轻人和老年人的测量数据来计算瞬时成本，从而表征代谢成本中特定阶段的差异。第二个子任务将确定数据驱动的时间剖面估计方法在不同人群中的普适性。这项研究将通过提供使用现有技术无法测量的动态代谢成本时间分布来改变步态分析。对这一新信息的获取将导致多种生物力学应用的改进，包括(1)运动损伤的诊断，(2)定向辅助设备的处方，以及(3)定向康复练习。该项目由残疾和康复工程计划(DARE)和既定的激励竞争性研究计划(EPSCoR)联合资助。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Being able to walk easily is strongly associated with independence and quality of life. Aging is accompanied by a significant reduction in mobility. Existing treatments and therapies rely on respiratory measurements of walking effort. These respiratory measurements can only quantify the average effort of walking. As a result of this limitation, existing treatments and therapies sometimes fail to target the phases of the walking motion that need the most assistance. This project will use data-driven approaches and models to overcome limitations in the ability to measure the effort of walking. Access to this new information will enable evaluating how therapies affect different stages of motion. The data-driven methods will be initially developed using a dataset generated by a computer walking models with physically induced changes to specific stages of motion. For example, forward-pulling forces will be applied to the waist of the model to induce changes that can be leveraged to detect the fluctuations in walking effort. This computer walking model provides access to a complete measure of the effort required for walking, which will be used to validate the data-driven methods. Next, the new data-driven methods will be validated using measurements from real human walking experiments. In these human experiments, pulling forces will be applied by a robotic tether connected to the waist of the participant to induce changes that will be used to detect the effort of the different motion stages. In the final studies, the methods will be used to determine how the effort required for walking differs in younger and older adults. The differences in the effort will be characterized in each stage of motion using human experiments with both younger and older adults. The outcomes of this project will help lead to the creation of enhanced treatments and assistive devices that improve all stages of motion. Throughout this project, the investigators will provide courses for older adults on the mechanics and health aspects of walking and data science and digital engineering through the Osher Lifelong Learning Institute.The goal of this project is to leverage new data-driven approaches to characterize differences in metabolic cost of phases of the gait cycle in old versus young adults. The project will combine novel, data-driven approaches based on system identification and robotic perturbations to characterize the time profile of signals that cannot be measured directly, such as metabolic cost. The first objective will produce the time profile of metabolic cost within simulated gait data. Novel data-driven approaches will be developed based on weighted regression, neural networks, and autoencoders to identify the metabolic cost time profile from biomechanical signals. Initially, these methods will be created in a predictive walking simulation from which the metabolic time profile is fully known, such that the new methods can be evaluated during their development. The second objective will evaluate different time profile estimation approaches in human experiments. The methods created in the first objective will be tested using human experiments with robotic perturbations. The capacity of using the data-driven methods to detect changes in swing and push-off will also be investigated using human experiments where elastic ankle tethers or added mass are used to introduce direct changes to the gait cycle. The third objective will characterize the differences in cost contributions of the phases of the gait cycle between older and younger adults. The first subtask will characterize the phase-specific differences in metabolic cost by applying the data-driven methods to compute the instantaneous costs using measured data from younger and older adults. The second subtask will determine the generalizability of the data-driven time-profile estimation approaches across different populations. This research will transform gait analysis by providing access to dynamic metabolic cost time profiles, which cannot be measured using existing techniques. Access to this new information will lead to improvements across multiple biomechanics applications, including (1) diagnosis of motion impairments, (2) prescription of targeted assistive devices, and (3) targeted rehabilitation exercises. This project is jointly funded by the Disability and Rehabilitation Engineering Program (DARE) and the Established Program to Stimulate Competitive Research (EPSCoR).This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.