Feedback Control Design for Bipedal Robots

研究步行机器人的动机来自不同的社会学和商业利益，从希望在危险的职业中取代人类，到恢复残疾人的运动，在不适合轮式运动的地形中使用腿运动的可能性，以及以拟人或动物方式操作的机器的心理吸引力。反馈控制是实现双足运动的一个重要组成部分，由于机器人机构的自由度、与环境接触的间歇性和单向性以及驱动不足，使得双足机器人的控制系统设计面临许多挑战。当今技术最先进的机器人行走是基于启发式控制原理，该原理施加平足行走运动，以避免处理伴随足部旋转的驱动不足（即，脚趾滚动和脚跟罢工）。Grizzle教授的研究将发展以拟人化步态行走的分析基础。通过不依赖于运动学，显著改进的反馈控制器将被设计为允许在步行的站立阶段期间脚的自然旋转，同时仍然保证步行运动的稳定性。超过一定的速度，步行就不再有效（甚至不可能）。本文将从理论和实验两个方面对双足机器人的行走运动控制这一极具挑战性的问题进行研究。这些结果将大大提高设计能够执行类似人类任务的机器的能力。研究的另一部分将着眼于控制一个带有被动膝关节（主动阻尼，但其他情况下不驱动）的机器人，类似于用现代假肢行走。Grizzle教授积极从事本科生和研究生教育。NSF对这项研究提案的支持是通过建造一个机器人原型来加强密歇根大学的研究和教育基础设施。机器人为学生提供实验设施和许多项目。这项研究补助金提供了一个博士生的培训，和本科生的研究经验（REU）补充这项补助金将提供两个本科生有意义的研究经验。特别注意的是，从代表性不足的群体中招募至少一名这样的学生。Grizzle教授还通过向本科生团体和高中生做关于双足步行机器人的演讲，更广泛地分享了工程研究生涯的兴奋。基于他在机器人运动控制方面的研究成果，ProfessorGrizzle正在与神经病学和康复领域的同事建立研究关系。他的目标是将对机器人研究中产生的双足姿势和运动的更深入理解，转移到为中风或脊髓损伤的一类患者开发改进的康复疗法上。该研究的研究成果在Grizzle教授的网站上维护;见上面的URL。

The motivation for studying walking robots arises from diverse sociological and commercialinterests, ranging from the desire to replace humans in hazardous occupations, to the restoration ofmotion in the disabled, the possibility of using legged locomotion in terrain that is unsuitable forwheeled locomotion, and the psychological appeal of machines that operate in anthropomorphic oranimal-like ways. Feedback control is an integral part of achieving bipedal locomotion.Control system design for biped robots poses numerous challenges due to the many degrees offreedom in the mechanisms, the intermittent and unilateral nature of the contact conditions withthe environment, and underactuation. The most technologically advanced biped robots today arewalking on the basis of a heuristic control principle that imposes a flat-footed walking motion inorder to avoid dealing with the underactuation that comes with foot rotation (i.e., toe roll andheel strike). Professor Grizzle's research will develop the analytical foundations of walking with ananthropomorphic gait. By not relying on heuristics, significantly improved feedback controllers willbe designed that allow the natural rotation of the foot during the stance phase of walking, whilestill guaranteeing stability of the walking motion. Beyond a certain speed, walking is no longerefficient (or even possible). The extremely challenging problem of controlling a running motionof a bipedal robot will be investigated, both theoretically and experimentally. These results willgreatly enhance the ability to design machines that can perform human-like tasks. Another portionof the research will look at controlling a robot with a passive knee (active damping but otherwiseunactuated), analogous to walking with a modern prosthetic.Professor Grizzle is actively engaged in undergraduate and graduate student education. NSFsupport of this research proposal is enhancing the research and education infrastructure at theUniversity of Michigan through the construction of a biped robot prototype. The robot affordsexperimental facilities and numerous projects for students. This research grant provides for thetraining of one PhD student, and a Research Experience for Undergraduates (REU) supplement tothis grant will provide a meaningful research experience for two undergraduate students. Specialattention is being paid to recruiting at least one of these students from an underrepresented group.Professor Grizzle is also more broadly sharing the excitement of a research career in engineeringby making presentations on bipedal walking robots to undergraduate student groups, and highschool students. On the basis of his research findings in biped robot locomotion control, ProfessorGrizzle is forging research relationships with colleagues in neurology and rehabilitation. His goalis to transfer the deeper understanding of bipedal posture and locomotion that is resulting fromthe study of robots, to the development of improved rehabilitation therapies for a class of patientswho have suffered strokes or spinal cord injuries.The research findings of this grant are maintained on Professor Grizzle's web site; see above forthe URL.