NRI: Liquid-Solid Metal for Embodied Intelligence in Semi-Soft, Human-Collaborative Robots

这项国家机器人计划（NRI）拨款支持的研究将为新型半软机器人的设计和控制提供新的知识，促进科学进步，提高国家工业和医疗保健目标。灵活的机器人能够在受限的环境中导航。因此，它们有可能彻底改变各种机器人任务，包括工业检查，搜索和救援行动以及微创手术。现有的柔性机器人要么是软的（由橡胶弹性元件制成），这使得它们安全，要么是硬的（由刚性支撑结构制成），这使得它们能够举起重物。该奖项支持的团队旨在创造一种新型的半软机器人，利用这两种模式的优势。研究人员将继续进行基础研究，以提供设计和控制机器人所需的知识，这些机器人的内部骨骼可以从液态金属过渡到固态金属。一个半软机器人将能够压缩自己通过小开口，然后超越，以操纵它的环境。这种机器人有可能造福美国经济和社会，在基础设施、制造业、灾难应对和医学方面有广泛的应用。该研究涉及机械工程、数学建模、控制理论、计算机科学和机器人运动规划等多个学科。研究方法和相关的推广活动将有助于扩大代表性不足的群体对机器人技术的参与，并对工程和计算机科学教育产生积极影响。通过低熔点金属合金的相变，半软机器人将拥有超越现有机器人设计的能力，使机器人能够从非常软的（通过狭窄的开口变形），非常坚硬（以便机器人可以举起重物并与环境进行有力的互动）。 这些半软机器人还将具有弯曲的控制肌腱，使它们具有比目前使用直控制肌腱的机器人更复杂的形状的潜力。 然而，在实现有效、可控的半软机器人的潜力之前，科学障碍仍然存在。本研究将探讨半软体机器人之设计、建模、控制与运动规划之整合。更具体地说，研究团队将（1）进行有限元建模，以了解各种低熔点合金结构设计的机械性能及其与弯曲肌腱驱动的相互作用;（2）开发基于Cosserat Rod和机器学习的机器人机械模型;（3）研究分辨率和模型预测控制的使用;（4）开发快速运动规划算法和用户界面，以实现对机器人的安全和准确的监督人类控制;以及（5）在模拟肺手术中验证这些系统和概念。该奖项反映了NSF的法定使命，并通过使用基金会的智力价值进行评估，更广泛的影响审查标准。

This National Robotics Initiative (NRI) grant supports research that will contribute new knowledge to the design and control of a novel semi-soft robot, promoting both the progress of science and enhancing national industrial and healthcare objectives. Flexible robots are able to navigate through restrictive environments. As such, they have the potential to revolutionize a variety of robotic tasks including industrial inspections, search-and-rescue operations, and minimally invasive surgery. Existing flexible robots are either soft (made of rubbery elastic elements), which makes them safe, or hard (made of rigid support structures), which makes them able to lift heavy loads. The team supported by this award seeks to create a new kind of semi-soft robot that leverage the benefits of both paradigms. The researchers will pursue fundamental research to provide needed knowledge in the design and control of robots with internal skeletons that can transition from liquid to solid metal. A semi-soft robot will be able to compress itself to pass through small openings, and then stiffen beyond, to manipulate its environment. Such a robot has the potential to benefit the U.S. economy and society with broad applications in infrastructure, manufacturing, disaster response, and medicine. This research involves several disciplines including mechanical engineering, mathematical modeling, control theory, computer science, and robot motion planning. The research approach and associated outreach activities will help to broaden participation of underrepresented groups in robotics and positively impact engineering and computer science education.Semi-soft robots will possess capabilities beyond existing robot designs via the phase transition of low melting point metal alloys, enabling the robot to control its stiffness from very soft (to deform through narrow openings), to very stiff (so that the robot can lift heavy loads and interact forcefully with its environment). These semi-soft robots will also have curved control tendons, endowing them with the potential to take on more complex shapes than current robots, which use straight control tendons. However, scientific barriers remain before the potential of effective, controllable semi-soft robots is realized. This research will investigate the integration of the design, modeling, control, and motion planning of semi-soft robots. More specifically, the research team will (1) perform finite element modeling to understand the mechanical properties of a variety of low melting point alloy structure designs and their interaction with curved tendon actuation; (2) develop Cosserat Rod- and machine learning-based mechanical models of the robot; (3) investigate the use of resolved rates and model predictive control; (4) develop fast motion planning algorithms and a user interface that enables safe and accurate supervisory human control of the robot; and (5) validate these systems and concepts in simulated lung surgery.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.