NRI: FND: Collaborative Navigation, Learning, and Collaboration in Fluids with Application to Ubiquitous Marine Co-Robots

作为地球的命脉，海洋塑造并调节着全球气候模式，维持着化学和温度的完美平衡，使地球上所有的生命形式得以生存和繁衍。然而，由于缺乏足够的海洋表面以下观测数据，目前对全球海洋活动和海洋健康的认识极其不足。有待探索的海洋面积与地下区域现有传感器的数量之间的差距仍然惊人地巨大，使大多数海洋尚未被探索。小型自主水下航行器正在成为持久和普遍的海洋传感和监测的重要组成部分。准确的定位对于这些车辆进行智能传感和控制以及用户正确解释车辆的测量结果至关重要。然而，水下定位是出了名的具有挑战性，因为海洋对射频信号不透明，使得基于卫星的定位系统在水下不可用。为此，该项目将产生新颖的算法，使海洋机器人团队能够通过利用洋流作为定位参考，持续协作地导航未开发的海洋体积。该项目将从根本上增加移动机器人在流体环境中的占地面积和自主性。该项目的成果将有利于几个相关的研究领域，包括海洋学、海洋生态学和气象学。此外，该项目将为STEM（科学、技术、工程和数学）学生，特别是夏威夷原住民，创造独特的机会，让他们认识到机器人的巨大潜力，获得海洋机器人的经验，并参与跨学科的研究活动。该项目将产生一系列可扩展的算法，使移动机器人团队能够在最少的基础设施支持下协作导航和采样流体环境。这些新算法通过无处不在的海洋协作机器人（co-robots）的应用实例得到了验证。研究目标包括：(1)利用背景流动力学知识提高惯性导航长期性能的协同流辅助导航算法；㈡一种基于移动机器人现场流动观测的物理信息、数据驱动的流体动力学学习方法；（iii）一种基于流体的同步定位和映射（fluid-SLAM）方案，该方案可以实现并行的流辅助导航和流动力学学习，以及（iv）一种用于协作机器人团队的分散协作流体slam算法。协同流辅助导航和流动力学学习算法的现场评估将在夏威夷附近的海洋环境中进行。单机器人和协同机器人流体slam算法将在由纳米四旋翼飞行器组成的室内协同机器人试验台的模拟场景中进行评估。由此产生的协同机器人算法将从根本上提高移动协同机器人在不确定和非结构化环境中分布式感知和协同学习的适应性和鲁棒性。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

As the lifeblood of Earth, the ocean shapes and regulates global weather patterns, maintaining the perfect balance of chemistry and temperature to allow all Earth's life-forms to survive and thrive. Nonetheless, the current understanding of global ocean activities and ocean health is extremely inadequate due to the lack of sufficient observation data below the ocean surface. The gap between the large ocean volumes to explore and the number of existing sensors in subsurface regions remains astonishingly huge, leaving the majority of the oceans unexplored. Small-size autonomous underwater vehicles are becoming essential elements in persistent and pervasive ocean sensing and monitoring. Accurate localization is of utmost importance for these vehicles to perform intelligent sensing and control as well as for the users to properly interpret the vehicles' measurements. However, underwater localization is notoriously challenging since the ocean is opaque to radio frequency signals, rendering the satellite-based positioning systems unavailable underwater. To this end, this project will result in novel algorithms that enable teams of marine robots to persistently and collaboratively navigate the under-explored ocean volumes by utilizing ocean flows as localization references. This project will fundamentally increase the footprint and autonomy of mobile robots in fluid environments. The project outcomes will benefit several pertinent research areas including oceanography, marine ecology, and meteorology. Furthermore, the project will create unique opportunities for STEM (science, technology, engineering, and mathematics) students, especially Native Hawaiians, to recognize the great potentials of robotics, gain experience with marine robots, and participate in cross-disciplinary research activities.The project will result in a series of scalable algorithms that enable teams of mobile robots to collaboratively navigate and sample fluid environments with minimal infrastructural support. These novel algorithms are instantiated with the application of ubiquitous marine collaborative robots (co-robots). The research objectives include (i) a collaborative flow-aided navigation algorithm that improves the long-term inertial navigation performance by utilizing the knowledge about the dynamics of background flows; (ii) a physics-informed, data-driven fluid dynamics learning method based on in-situ flow observations by mobile robots; (iii) a fluid-based simultaneous localization and mapping (fluid-SLAM) scheme that enables concurrent flow-aided navigation and flow dynamics learning, and (iv) a decentralized cooperative fluid-SLAM algorithm for teams of co-robots. Field evaluations of the cooperative flow-aided navigation and flow dynamics learning algorithms will be conducted in ocean environments near Hawaii. The single-robot and co-robot fluid-SLAM algorithms will be evaluated in simulated scenarios using an indoor co-robot testbed consisting of a fleet of nano-quadrotors. The resulting co-robot algorithms will fundamentally advance the adaptability and robustness of mobile co-robots in distributed sensing and collaborative learning in uncertain and unstructured environments.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.