New motion control tools for robotics and rehabilitation

这项研究计划涉及算法的发展，使机器人执行复杂的运动，并证明了稳定性。本研究的工具以非线性控制理论和力学为基础。机器人曾经被下放到制造工厂的地板上，今天，机器人正在我们的家庭、医院和整个社会中蔓延。随着机器人技术的进步，人们越来越需要控制算法，使机器人能够安全地执行复杂的任务。如何让行走的机器人爬楼梯或跳跃？如何让像鱼一样的机器人在陆地上游泳或爬行？传统的机器人运动控制方法是将运动控制问题划分为两个概念上独立的子问题：运动规划和轨迹跟踪。在过去的15年里，机器人运动控制出现了一种新的范式，它能够产生惊人的自然运动，并具有优越的鲁棒性。这里讨论的范例是虚拟约束。这个想法是通过对机器人关节的约束来隐式地表示期望的运动（例如，双足行走）。这种约束在物理上并不存在，但它们是通过反馈控制来执行的，因此才有了“虚拟”这个形容词。虽然虚拟约束已经成为行走机器人控制的一个重要范例，但它们在其他机器人领域的适用性仍未得到充分探索。提出的研究是由这样一个猜想驱动的，即虚拟约束范式实际上适用于机器人的大量运动控制问题，并且它是诱导高度复杂的生物启发运动的良好候选者。在过去的几年里，我和我的合作者已经开发了一种虚拟约束理论，用于欠驱动度为1的机器人，即一个驱动器少于其自由度的机器人。虽然到目前为止我们获得的结果令人鼓舞，但仍有许多工作要做，这是我目前申请发现基金的主题。我打算在未来五年进行的研究主要有两个重点：理论和实验。在理论方面，我打算为欠驱动程度大于1的机器人开发一种新的虚拟约束理论。这将使虚拟约束范式在各种各样的机器人运动问题中变得可行。在实验方面，我将与医疗康复领域的研究人员展开合作，研究虚拟约束在假肢设备和外骨骼中的应用。这项研究的长期目标是开发一套全面的工具，用于生物启发机器人的高级运动控制，能够产生看起来自然的复杂运动，并且具有对环境不确定性的卓越鲁棒性。

This research proposal concerns the development of algorithms to make robots perform complex motions with proven stability. The tools of this research are grounded in nonlinear control theory and mechanics.Once relegated to the floors of manufacturing plants, today robots are spreading in our homes, hospitals, and the society at large. As the technology of robotics advances, there is an increasing need for control algorithms making robots perform complex tasks with proven safety. How to make a walking robot climb stairs or jump? How to make a fish-like robot swim or a snake robot crawl on land?The traditional approach to controlling the motion of robots was to partition the motion control problem into two conceptually separate sub-problems: motion planning and trajectory tracking. Over the past fifteen years, a new paradigm has emerged for motion control of robots which is capable of producing strikingly natural motions with a superior degree of robustness. The paradigm in question is that of virtual constraints. The idea is to represent a desired motion (e.g., bipedal walking) implicitly by means of constraints on the robot joints. Such constraints do not physically exist, but they are enforced by means of feedback control, hence the adjective "virtual". While virtual constraints have come to represent a prominent paradigm for the control of walking robots, their applicability in other areas of robotics remains largely unexplored. The proposed research is driven by the conjecture that the virtual constraint paradigm is in fact applicable to a large class of motion control problems for robots, and it is a good candidate to induce biologically inspired motions of great complexity. Over the past few years, my collaborators and I have developed a theory of virtual constraints for robots with degree of underactuation one, i.e., robots with one actuator less than their number of degrees-of-freedom. While the results we obtained so far are encouraging, much work remains to be done, and this is the subject of my current Discovery Grant application. The research I intend to perform over the next five years has two main thrusts: theoretical and experimental. On the theoretical side, I intend to develop a new theory of virtual constraints for robots with degree of underactuation greater than one. This will allow the virtual constraint paradigm to become viable in a wide variety of robot locomotion problems. On the experimental side, I will initiate a collaborative thrust with researchers in the area of medical rehabilitation, investigating the use of virtual constraints for prosthetic devices and exoskeletons.The long-term objective of this research is the development of a comprehensive set of tools for advanced motion control of biologically inspired robots, capable of producing complex motions that look natural, and which are endowed with a superior degree of robustness against environmental uncertainties.