Collaborative Research: Predicting and Optimizing User Comfort for Lower-limb Exoskeletons through Mutual Motor Adaptations

像外骨骼这样的辅助机器人设备可以用来增强人类的能力，帮助进行体力劳动，并改善康复。用户感知的舒适性对于辅助机器人设备的广泛采用至关重要，但舒适性的定义和测量仍然难以捉摸。该项目的研究目标是系统地定义、建模和优化人类用户在使用动力腿外骨骼行走时感知到的舒适度。该项目有三个目标：第一个目标是根据下肢外骨骼机器人行走时记录的生物信号开发一个舒适模型。第二种是基于用户舒适度的口头报告来优化外骨骼控制参数。第三个优化外骨骼控制舒适，无需直接用户报告。如果成功，该项目将导致新一代外骨骼设备的出现，这些外骨骼设备将与它们设计服务的人类更加兼容。这将使大量有步态障碍的人受益，从而促进国家的健康和福利。这项工作的广泛影响包括为宾夕法尼亚州中部和德克萨斯州达拉斯市未被充分代表的少数民族提供新颖的、实际操作的外展活动。该项目朝着机器人控制器的发展迈出了重要的一步，这些控制器可以不断适应每个用户，并最大限度地减少用户的不适。这将通过使用神经网络模型来分析和建模一组选定的用户的生物信号（例如，代谢成本、心率、肌肉激活、运动学和动力学）来开发一种新的舒适预测器，然后通过创建智能控制器来实现，通过人在环强化学习来最大化用户的舒适度。计划使用两个装置（膝关节和髋关节装置和踝关节装置）进行人体受试者实验，以验证舒适性预测器可以与优化方法一起有效使用。如果成功，该项目将使大量有步态障碍的人以及那些需要机器人帮助进行高强度体力劳动的人受益。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Assistive robotic devices such as exoskeletons can be used to enhance human capabilities, help with physically-intense labor, and improve rehabilitation. User-perceived comfort is critically important for the wide-spread adoption of assistive robotic devices, yet the definition and measurement of comfort remains elusive. The research objective of this project is to systematically define, model, and optimize the comfort perceived by human users while walking with powered leg exoskeletons. The project pursues three objectives: The first develops a model of comfort based on biosignals recorded during walking with lower limb exoskeleton robots. The second optimizes parameters of exoskeleton control based on verbalized reports of user comfort. The third optimizes exoskeleton control for comfort without direct user reporting. If successful, the project will lead to a new generation of exoskeleton devices that are more compatible with the humans they are designed to serve. This would benefit a large population of people with gait impairments, thereby advancing the national health and welfare. Broader impacts of this work include novel, hands-on outreach activities serving underrepresented minorities in central Pennsylvania and in Dallas, Texas.This project takes significant steps towards the development of robotic controllers that adapt continuously to each user and minimize user discomfort. This will be achieved by using neural network models to analyze and model a selected set of the users' biological signals (e.g., metabolic cost, heart rate, muscle activation, kinematics, and kinetics) to develop a novel comfort predictor, and then by creating intelligent controllers that maximize the user comfort via human-in-the-loop reinforcement learning. Human subject experiments are planned using two devices (a knee and hip device, and an ankle device) to verify that the comfort predictor can be used effectively with the optimization method. If successful, the project could benefit a large population of people with gait impairments and those requiring robotic assistance with physically-intense labor.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.