CAREER: A Task-Invariant Customization Framework for Lower-Limb Exoskeletons to Assist Volitional Human Motion

该学院早期职业发展（CAREER）奖将支持研究，以推进下肢外骨骼控制和优化方面的知识，为人类用户提供跨运动任务的定制帮助。传统的定制范式通常旨在优化特定运动任务的预定义扭矩曲线的参数，这无法适应人类每天进行的连续变化的活动。流行的方法还强调最大限度地减少与人类运动相关的缓慢收敛的能量消耗，这对身体健全的人很重要，但对患有病理性步态的人来说可能不是高度优先的。这个项目的目标是提供一个范式转换，从特定于任务的，缓慢收敛的定制任务不变，快速定制。这项研究将促进下肢外骨骼的主动学习和适应，以大幅降低每年近100万美国人新中风的步态康复成本。整个步态康复过程中的关键参数，如体重支撑比，将根据中风受试者的训练进度自动定制，否则将由治疗师和工程师团队进行调整，通常持续数小时。这项研究还将通过降低控制参数定制的相关成本，促进外骨骼在制造现场，仓库，战场和其他相关场景中的使用。综合教育计划将有助于培养下一代可穿戴机器人研究人员，并激励K-12学生进一步攻读STEM学位。人类主题研究和展览将促进公众对可穿戴技术的认识，特别是在传统上代表性不足的群体中。该项目的目标是：1）研究在连续变化的运动任务和环境中的不变辅助策略，2）构建用于跨运动任务定制外骨骼辅助的完整框架，以及3）理解定制框架如何促进对不同用户的意志运动和活动的辅助适配。利用两层优化结构将通过跟踪虚拟参考模型的期望能量或质心动量来快速确定内环中的任务不变辅助策略，同时基于基于人类绩效的成本函数来更新外环中的参数。PI和他的团队将进行的研究将调查改变人体能量和质心动量的任务不变辅助策略，构建一个完整的人在回路定制框架，快速定制这两个量，并通过对人类受试者进行各种运动任务的实验来验证其功效。该奖项反映了NSF的法定使命，并通过使用基金会的学术价值和更广泛的影响审查标准。

This Faculty Early Career Development (CAREER) award will support research that advances knowledge in the control and optimization of lower-limb exoskeletons in providing their human users with customized assistance across locomotor tasks. Conventional customization paradigms often aim at optimizing parameters of pre-defined torque profiles for specific locomotor tasks, which cannot accommodate the continuously varying activities humans perform every day. Prevalent approaches also place emphasis on minimizing slow-converging energy expenditures associated with human locomotion, which is important for able-bodied persons but might not be of high priority for individuals with pathological gaits. The goal of this project is to provide a paradigm shift from task-specific, slow convergent customization to task-invariant, rapid customization. This research will facilitate active learning and adaptation of lower-limb exoskeletons to drastically reduce the cost of gait rehabilitation for nearly a million Americans who sustain a new stroke every year. Key parameters throughout gait rehabilitation such as body-weight support ratio will be automatically customized based on stroke subject's training progress, which will otherwise be tuned by a team of therapists and engineers that usually lasts for hours. This research will also promote the use of exoskeletons for able-bodied users in manufacturing sites, warehouses, battlefields, and other relevant scenes by reducing the associated costs in control parameter customization. The integrated education plan will help cultivate the next-generation wearable robot researchers and motivating K-12 students to further pursue STEM degrees. The human subject studies and exhibits will promote the awareness of wearable technologies among the public, especially among traditionally underrepresented groups.The goals of this project are to: 1) investigate invariant assistive strategies in continuously varying locomotor tasks and environments, 2) construct a complete framework for customizing exoskeleton assistance across locomotor tasks, and 3) understand how the customization framework facilitates assistance adaptations to different user’s volitional motion and activities. Utilizing a two-layer optimization structure will rapidly determine task-invariant assistive strategies in the inner-loop through tracking the desired energetics or centroidal momentum of a virtual reference model, meanwhile updating its parameters in the outer-loop based on human performance-based cost functions. The research the PI and his team will conduct will investigate task-invariant assistive strategies that alter human body energetics and centroidal momentum, construct a complete, human-in-the-loop customization framework that rapidly customizes these two quantities, and validate its efficacy via experiments on human subjects across various locomotor tasks.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.