NRI: FND: COLLAB: An Open-Source Robotic Leg Platform that Lowers the Barrier for Advanced Prosthetics Research

该项目的目标是为研究人员提供一个功能齐全的标准化开源机器人腿研究平台，而无需从头开始开发每个组件的巨大负担。其结果将是一个强大的和廉价的测试台，可以很容易地制造，组装和控制。机器人假肢的设计和商业化面临的最大挑战之一是控制策略-也就是说，计算机化的指令集规定了机械装置每个部件的努力程度和时间安排。开发控制策略的挑战来自于主动机器人肢体必须完成的许多不同功能。一个重要的功能是检测截肢者执行不同移动活动的意图，例如在水平表面上行走与上下楼梯。另一个重要的功能是协调假肢的努力和运动模式，以模仿健康的人体。有许多研究人员独立地研究更好的控制算法，以解决诸如此类的挑战。为了有价值，这些算法必须通过实验进行测试和验证。世界各地的研究小组为此目的创造了各种专门的机器人腿设计，代表了大量的时间和精力投入。获得合适的研究平台所需的资源是新研究人员需要克服的一个重大障碍。此外，既定研究人员使用的设计差异很大，这阻碍了研究小组之间对新控制策略的比较。该项目产生的可访问的标准化腿平台将降低进入门槛，允许新的研究人员研究机器人腿的控制，明确比较不同的控制方法，并普遍推进该领域。最后，从长远来看，该项目改进的假肢设计将有益于截肢者的生活。该项目的总体研究目标是：1）确定低成本，高性能，开源机器人膝关节和踝关节系统的机电设计; 2）了解如何将单独的假肢控制策略结合起来，以使截肢者步态受益，以及3）评估和比较截肢者实验中的控制器。该方法采用了一种新的设计方法，采用可选择的系列弹性和高扭矩电机技术，以实现低成本的高性能。开源架构中的可互换控制模块使研究人员能够在系统的低，中，高层次上研究新的控制方法，即电机驱动，关节控制和人类意图识别。特别是，基于反射的方法和基于阶段的方法将被实现为中级控制模块，而高级意图识别模块将使机器人腿能够在不同的用户活动之间自动切换。在所有情况下，将新的机器人腿与这些算法一起使用将使测试能够在真实世界的场景中进行，而不是局限于实验室。该项目的结果将降低进行研究的障碍，并使不同的控制方法与标准化的腿硬件之间的公平比较成为可能。最后，拟议的工作将通过培训、推广和传播对学生和社区产生影响。

The objective of this project is to provide researchers with access to a fully capable and standardized open-source robotic leg research platform, without the immense burden of developing each component from scratch. The outcome will be a robust and inexpensive test bed that can be easily manufactured, assembled, and controlled. One of the greatest challenges to the design and commercialization of robotic prosthetic legs is the control strategy -- that is, the computerized instruction set that specifies the effort level and timing for each component of the mechanism. Challenges in developing control strategies stem from the many different functions that an active robotic limb must accomplish. One important function is to detect the amputee's intention to perform different mobility activities, such as walking on a level surface versus ascending or descending stairs. Another important function is to coordinate the pattern of effort and movement of the prosthetic limb in order to emulate the healthy human body. There are many researchers working independently on better control algorithms to address challenges such as these. To be of value, these algorithms must be tested and validated experimentally. Research groups around the world have created a variety of specialized robotic leg designs for this purpose, representing a significant investment of time and effort. The resources required to obtain a suitable research platform represent a substantial obstacle for new researchers to overcome. Furthermore, the vast difference in designs used by established researchers hinders the comparison of new control strategies across research groups. The accessible, standardized leg platform resulting from this project will lower barriers to entry, allowing new researchers to study the control of robotic legs, to unambiguously compare different control approaches, and to generally advance the field. Finally, the improved prosthetic leg designs arising in the long term from this project will benefit the lives of amputees. The overall research goals of this project are 1) to identify an electromechanical design for a low cost, high performance, open-source robotic knee and ankle system; 2) to understand how separate prosthesis control strategies can be combined to benefit amputee gait, and 3) to evaluate and compare resulting controllers in amputee experiments. The approach utilizes a novel design methodology employing selectable series elasticity and high-torque motor technology to achieve high performance at low cost. Interchangeable control modules in the open-source architecture allow researchers to investigate new control methods at low, mid, and high levels of the system, that is, motor drive, joint control, and human intent recognition, respectively. In particular, a reflex-based approach and a phase-based approach will be implemented as mid-level control modules and a high-level intent recognition module will enable the robotic leg to automatically switch between different user activities. In all cases, having the new robot leg available together with these algorithms will enable testing in real-world scenarios, rather than being confined to the laboratory. The results of this project will lower the barrier for conducting research and enable fair comparison across different control approaches with standardized leg hardware. Finally, the proposed work will impact students and the community through training, outreach, and dissemination.