NRI: Collaborative Research: Exploiting Granular Mechanics to Enable Robotic Locomotion

我们需要机器人将我们的触角延伸到肮脏和危险的环境中。要做到这一点，移动机器人必须能够在杂乱的非结构化地形中移动。传统的移动机器人还没有开始展示生物在自然地形上的多功能性。这是因为移动机器人是在干净的实验室地板上制造出来的，它们的模型主要是在干净的实验室地板上验证的，而生物有机体已经进化到能够适应异质、肮脏和不可预测的环境。这种现实世界复杂地形的一个重要例子，尽管它无处不在，但经常被我们的社区所忽视，包括在沙漠、灾难现场、容器和洞穴中常见的松散颗粒材料。因此，要想在肮脏的环境中创造出更高水平的移动性，需要机器人技术和物理学的同时进步，尤其是在与预期行为相关的相互作用方面。提出的工作是建立在几何力学，入侵的颗粒物理和来自沙漠栖息蛇的生物灵感的基础上。我们使用几何力学，这是一个将微分几何原理应用于经典力学问题的领域，为受生物启发的机器人设计步态。我们将几何工具的优点应用到颗粒环境中：即使在这些数学上“混乱”的系统中，我们也可以开始有效地分析步态。这项工作的关键概念是，具有复杂、非线性低级物理的系统通常表现出更“清晰”的高级运动，通常由运动学关系近似。这种高级运动控制器的发展将有助于我们发现颗粒介质中运动的基本生物学原理的能力。因此，我们将为颗粒材料开发计算效率高的分析工具，并将开发技术来研究颗粒介质表面系统的运动。

We need robots to extend our reach into dirty and dangerous environments. To do so, mobile robots must be able to locomote in messy unstructured terrains. Conventional mobile robots have not begun to display the multi-functionality of organisms that inhabit natural terrains. This is because mobile robots have been created in, and their models mainly validated on, clean hard laboratory floors, whereas biological organisms have evolved to contend with heterogeneous, dirty and unpredictable environments. One important example of such real world complex terrain, often overlooked by our community despite its ubiquity, involves loose granular materials commonly found in deserts, disaster sites, containers, and caves. Therefore creation of the next level of mobility to traverse dirty environments requires simultaneous advances in both robotics and physics, particularly regarding the interactions associated with desired behaviors. The proposed work is built on a foundation of geometric mechanics, granular physics of intrusion and biological inspiration from desert-dwelling snakes. We use geometric mechanics, a field that applies principles from differential geometry to problems in classical mechanics, to design gaits for biologically inspired robots. We bring the benefits of the geometric tools to bear on granular environments: in even these mathematically "messy" systems, we can begin to efficiently analyze gaits. The key concept in this effort is that systems with complicated, nonlinear low-level physics often exhibit much "cleaner" high-level motion, often approximated by a kinematic relationship. Development of such high-level motion controllers will be aided by our ability to discover basic biological principles of locomotion in granular media. We will therefore develop computationally efficient analysis tools for granular materials and will develop techniques to study the locomotion of systems on the surface of granular media.