NRI: Collaborative Research: Versatile Locomotion: From Walking to Dexterous Climbing with a Human-Scale Robot

该项目旨在赋予腿式机器人在各种地形中导航的技能。 这种能力需要在搜索和救援，建筑和探索地球和其他行星上的远程环境等应用中使用机器人。这个多学科团队由杜克、斯坦福大学、加州大学圣巴巴拉分校、喷气推进实验室和Motiv Robotics的研究人员组成，他们将开发一种机器人，可以爬上从平地到悬垂悬崖的各种表面。使用一系列传感器，独特的手和复杂的算法，机器人将动态地采用步行，爬行，攀爬和摆动策略来穿越各种各样的地形。在这项研究的过程中，该团队希望实现人类规模的攀岩机器人首次演示的里程碑。这项研究还有望深入了解人类和动物运动中的认知和生物力学过程。虽然攀岩是这项工作的理想试验场，但该项目进行的基础研究更多地是为了实现一个通用目标，即为机器人提供适应性地在各种地形中航行的身体和认知技能。它采用灵巧的攀爬方法，使用非步态的协调接触序列来移动身体，就像灵巧的操作使用手指和手掌的接触来移动物体一样。 它将应用优化，机器学习，生物灵感和控制理论的原理，在几个领域做出智力贡献，如机器人手设计，规划算法，平衡策略和运动性能测量。 在这项研究的过程中，新的抓手，基于传感器的规划策略，反应机动，和运动指标将被开发。

The project aims to give legged robots the skills to navigate a wide variety of terrain. This capability is needed to employ robots in applications such as search-and-rescue, construction, and exploration of remote environments on Earth and other planets. The multidisciplinary team, composed of researchers at Duke, Stanford, UC Santa Barbara, JPL, and Motiv Robotics, will develop a robot to climb a variety of surfaces ranging from flat ground to overhanging cliffs. Using an array of sensors, unique hands, and sophisticated algorithms, the robot will dynamically adopt walking, crawling, climbing, and swinging strategies to traverse wildly varied terrain. During the course of this research, the team hopes to achieve the milestone of the first demonstration of a human-scale rock climbing robot. The research is also expected to lead to insights into cognitive and biomechanical processes in human and animal locomotion.Although rock climbing serves as an ideal proving ground for the work, this project conducts basic research to address more a general-purpose goal; namely, to provide the physical and cognitive skills for robots to adaptively navigate varied terrain. It takes a dexterous climbing approach, which uses non-gaited, coordinated sequences of contact to move the body, much as dexterous manipulation uses contact with the fingers and palm to move an object. It will apply principles from optimization, machine learning, bioinspiration, and control theory to make intellectual contributions in several domains, such as robot hand design, planning algorithms, balance strategies, and locomotion performance measurement. Novel grippers, sensor-based planning strategies, reactive maneuvers, and locomotion metrics will be developed during the course of this research.