KDI: Coordinated Motion of Natural and Man-Made Groups

无论是一群优雅地游过水面的鱼，一群奔跑在草地上的鹿，一群飞过地平线的鸟，还是一群围绕着蜂巢盘旋的蜜蜂，人们不禁想知道这些和其他类似的自然群体是如何协调自己并如此平静地移动的，通常没有明显的领导者或任何形式的集中控制。他们必须使用什么样的信号？如果有的话，水流、漩涡或其他局部环境扰动在这一过程中起什么作用？是否存在协调的群体运动的普遍原则，如果有的话，它们可能是什么？它们如何被用来设计一群自主潜水器，或者一组地面或空中的移动的机器人，以协调的方式共同完成一项有用的任务？从广义上讲，这些都是本研究将解决的问题。更具体地说，我们建议对理解鱼群如何协调集体运动以及自主潜水器群如何被设计成以协调的方式移动的联合问题进行跨学科攻击。我们建议通过实验、理论和计算机模拟来解决这些问题。在理论层面上，我们的目标将是开发模型，这是生物学上合理的，此外，操作策略将使工程系统的协调组运动。我们的最终目标不是导出分析模型，也不是开发计算机模拟，这在视觉上是令人满意的，而是要发现潜在的概念，群体运动的协调可能取决于。例如，我们希望开发一些概念，一方面可以帮助解释鱼类如何保持它们的间距，另一方面可以作为设计各种类型的人造自动驾驶车辆分组的指导方针。这项研究的实验重点将在两个不同但平行的方向上进行。与收集世界上第一个数据库记录长期的目标，在一个大的学校个别鱼的三维运动，我们将采取实时序列的立体视频图像的实际鱼群生活在一个大的{1000加仑或更多}坦克。我们将描绘学校在各种条件下对各种刺激的反应。使用先进的图像/视觉处理技术，我们将把这些数据转换成一种适合于验证所提出的行为和分析学校教育模型的有效性的形式。该数据库连同全部文件将通过万维网提供给科学和工程界。为了在合理的现实环境中尝试候选组协调策略，我们将开发一个实验测试设施，由一所学校组成，大约有十几个相同的微型水下航行器。这些实验预计将对生物学和工程学的角度都有帮助;我们将能够测试在鱼缸中不可能实现的假设生物学模型的机器人版本。这个人造学校的实验将在普林斯顿大学的一个奥运会大小的游泳池中进行。我们的第一个目标将是为这些车辆配备仪器，使它们能够在没有远程控制的情况下自主运行。然后，我们将尝试各种机动，包括编队，巡航，分叉，避开障碍物和改变群体形状。我们将实施各种生物启发的目标寻求策略，如觅食以及与生物学和许多工程应用相关的梯度攀登任务。我们提出的研究将由实验和理论生物学家以及计算机视觉，控制系统和机器人专家组成的跨学科团队进行。我们认为，通过这种多管齐下的攻击，包括理论分析，计算机模拟，并与自然和人造群体的实验，我们可以达到的自然和人造群体的协调运动的基本机制的理解。

Be it a school of fish swimming gracefully through the water, a herd of deer running through a meadow, a flock of birds flying across the horizon, or a swarm of bees circling about a hive one cannot help but wonder how these and other similar natural groupings, coordinate themselves and move so flawlessly, often without an apparent leader or any form of centralized control. What kinds of signaling must they use? What role if any, do currents, vortices, or other local environmental disturbances play in this process? Are there universal principles of coordinated group motion and if so what might they be? How might they be used to design a school of autonomous submerged vehicles, or a group of ground-based or airborne mobile robots to collectively accomplish a useful task in a coordinated manner? In broad terms, these are the questions to which this research will be addressed.More specifically, we propose to mount a cross-disciplinary assault on the joint problems of understanding how fish schools coordinate their collective motion and how schools of autonomous submerged vehicles might be designed to move in a coordinated manner. We propose to address these problems experimentally, theoretically and by means of computer simulation. At the theoretical level, our aims will be to develop models, which are biologically plausible, and, in addition, operating strategies will enable coordinated group motion of engineered systems. Our ultimate goal is not to derive analytical models nor to develop computer simulations, which are visually satisfactory, but rather to discover the underlying concepts upon which the coordination of group motion might depend. Thus, for example we would hope to develop concepts which, on the one hand might help to explain how fish maintain their spacing and, on the other might serve as guidelines for the design of groupings of man-made autonomous vehicles of all types.The experimental thrust of this research will proceed in two distinct but parallel directions. With the objective of gathering the world's first data-base documenting the long-term, three-dimensional motions of individual fish within a large school, we will take timed sequences of stereo video images of actual fish schools living in a large {1000 gallon or more} tank. We will image the responses of schools to various stimuli and under various conditions. Using advanced image/vision processing techniques, we will transform this data into a form suitable for verifying the validity of proposed behavioral and analytical schooling models. This database together with full documentation will be made available to the science and engineering communities via the World Wide Web.With the objective of trying out candidate group coordination strategies in a reasonably realistic setting, we will develop an experimental test facility consisting of a school of approximately a dozen, identical miniature underwater vehicles. These experiments are expected to be useful for both the biological and the engineering perspectives; we will be able to test robotic versions of hypothesized biology models not possible in a fish tank. Experiments with this man-made school will be carried out in an Olympic-sized swimming pool, which is available at Princeton University. Our first goal will be to instrument these vehicles so that they can function autonomously, without remote control. We will then experiment with a variety of maneuvers including group formation, cruising, bifurcating, avoiding obstacles, and changing group shapes. We will implement various biologically inspired goal-seeking strategies such as foraging as well as gradient climbing tasks relevant both to biology and to numerous engineering applications.The research we are proposing will be carried out by a cross-disciplinary team consisting of experimental and theoretical biologists together with experts in computer vision, control systems and robotics. It is our view that by means of this multi-pronged assault consisting of theoretical analysis, computer simulation, and experimentation with both natural and man-made groups, we can attain an understanding of the underlying mechanisms governing the coordinated motion of natural and man-made groups.