EAGER: Local Control Strategies for Predicting Emergent Behavior and Cooperative Control in Real Time under Minimal Communication

这个早期概念探索研究补助金项目的目标是在三个领域开展探索工作，灵感来自鸟类、蝙蝠和昆虫等鸟类物种在极其杂乱和复杂的环境中优雅而快速地飞行的非凡能力。(I)研究蝙蝠和其他鸟类的大型数据库，以确定一系列关键实验，(Ii)使用收集的信息来指导在几个特定场景中创新新的控制原理，包括编队飞行和搜索任务，(Iii)研究纯粹的局部控制，以解释保护生物学家研究的几个选定的群体行为，并将进化思想作为一个平台，以(A)帮助更好地理解涌现行为，(B)综合基于数据的简单算法，以控制现实约束下的集体运动。要遵循的具体方法将涉及(A)将集体视为连续的，(B)用于无碰撞运动的同伦方法，以及(C)导致有限维搜索的逐点优化，从而消除了求解复杂的两点边值问题的需要。这项研究的成功完成将表明，鸟类和其他生物的启发可能会导致导航和控制模式的转变。这项工作的一个关键影响预计是更好地理解视觉和声纳是如何被生物系统用来在高度复杂、无结构和杂乱的环境中导航的。目前使用机器学习思想的图像处理技术对于实时计算来说太难了，需要新的方法。如果成功，这项研究可以回答这样的问题：如何将简单的纠错思想应用于基于视觉的导航？基于生物学原理，能否在保证收敛时间的情况下开发并行算法？此外，还将开发一个使用生物启发控制的异类移动代理平台，该平台可用于激励大学预科和本科生从事工程职业。结果将通过发表论文、会议报告和将以生物启发的控制设计组织的研讨会来传播。

The objective of this Early-Concept Grants for Exploratory Research (EAGER) project is to initiate exploratory work in three areas inspired by the remarkable ability of avian species such as birds, bats and insects to gracefully and rapidly fly through extremely cluttered and complex environments.: (i) research the large data base on bats and other avian species to identify a set of key experiments, (ii) use information gathered to guide the innovation of new control principles in a few specific scenarios including formation flying and search missions, (iii) investigate purely local control to explain a few selected swarm behaviors studied by conservation biologists and use the evolving ideas as a platform to (a) help better understand emergent behavior, and (b) synthesize simple data based algorithms for controlling the motion of collectives under realistic constraints. Specific approaches to be followed will involve (a) treating collectives as continua, (b) homotopy methods for collision free motion, and (c) pointwise optimization leading to finite dimensional searches that eliminates the need for solving complex two-point boundary value problems. The successful completion of this research will show that avian and other biological inspirations could lead to a shift in paradigm in navigation and control. A key impact of this work is expected to be a better understanding of how vision and sonar are used by biological systems to navigate in highly complex, unstructured and cluttered environments. The current state of the art of image processing using ideas of machine learning is simply too difficult for real time computations and new approaches are needed. If successful this research can answer questions such as: How can simple error correction ideas be adopted for vision based navigation? Can parallel algorithms be developed with guarantees of convergence time that are based on biological principles? Additionally, a heterogeneous mobile agent platform where bio-inspired controls are used will be developed which can be used to inspire pre-college and undergraduate students to take up careers in engineering. The results will be disseminated through publication of papers, conference presentations and a workshop to be organized in bio-inspired control design.