NRI: Rethinking Multi-Legged Robots: Passive Terrain Adaptability through Underactuated Mechanisms and Exactly-Constrained Kinematics

长期以来，腿式交通工具一直吸引着我们，因为它可以像动物一样敏捷地移动，优雅而迅速地到达轮式交通工具无法到达的地方。 然而，尽管取得了许多成功，但目前的腿式机器人大多无法穿越崎岖的地形，并且机械复杂，需要复杂的硬件和控制，并且操作时间短。该项目的工作将研究利用自适应电机和传动装置的多足机器人的开发，以降低复杂性和成本，同时还允许腿适应崎岖的地形。这些腿式机器人的设计将通过OpenRobotHardware.org免费提供，并将能够通过快速原型技术轻松廉价地制造。该项目采用了一种新的腿式机器人方法，提出了一种新的架构，通过应用欠驱动机构来减少封闭运动链与地面的过度约束-为机器人配备腿，被动适应地形粗糙度的大变化，对身体的干扰力最小。由于这种被动的适应性，机器人能够在粗糙的地形上找到一个稳定的立足点，而不需要任何感知或规划：腿只是落在适当的位置。此外，机构中的适应性还使得能够被动地调节由腿施加到主体的力和扭矩，以便最小化将倾向于使车辆不稳定或导致其失去航向的干扰。这种能力将大大简化腿式机器人系统的控制，大大提高功率效率及其对地形变化的鲁棒性，并降低生产和维护成本。

Legged vehicles have long intrigued us with the promise of moving with animal-like agility, gracefully and swiftly going where wheeled vehicles cannot. Despite numerous successes, however, current legged robots are mostly unable to traverse rough terrain, and are mechanically complex, requiring elaborate hardware and control, and having short operation-times. Work on this project will investigate the development of multi-legged robots that utilize adaptive motors and transmissions, to reduce the complexity and cost, while also allowing the legs to adapt to rough terrain. These legged robot designs will be made freely available through OpenRobotHardware.org, and will be able to be easily and inexpensively fabricated with rapid prototyping techniques.The project takes a new approach to legged robots by proposing new architectures that reduce over-constraint in the closed kinematic chains with the ground by applying under-actuated mechanisms - equipping robots with legs that passively adapt to large variations in terrain roughness with minimal disturbance forces to the body. As a result of this kind of passive adaptability, the robot is able to find a stable footing on rough terrain without any sensing or planning: the legs simply fall into place. Furthermore, adaptability in the mechanism also enables the forces and torques applied by the legs to the body to be passively regulated so as to minimize disturbances that would tend to destabilize the vehicle or cause it to lose its heading. This ability will drastically simplify the control of legged robot systems, greatly increase power efficiency and their robustness to terrain variations, and decrease production and maintenance costs.